 

 

 

 

 

2..
3.5.1! 3;

.1 . L. i
.3334“.

2 .. i

L

..

it a

3 .
:. ....
HI
. . kahuna.
. . ..

 

7H I53“;

3 LIBRARY
100 c, Michigan State
University

This is to certify that the
dissertation entitled

LEARNING MECHANISMS IN PROCESS IMPROVEMENT
INITIATIVES

presented by

JOHN DAVID HANSON

has been accepted towards fulﬁllment
of the requirements for the

degree In Business Adminstration

 

 

mints

Major Profe r’s Si ‘Ture
IL - I “I ‘ O S

 

Date

MSU is an Afﬁrmative Action/Equal Opportunity Institution

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:lClRC/DateDue.indd-p.1

LEARNING MECHANISMS IN PROCESS IMPROVEMENT INITIATIVES
By

John David Hanson

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Marketing and Supply Chain Management

2006

ABSTRACT
LEARNING MECHANISMS IN PROCESS IMPROVEMENT INITIATIVES
By
John David Hanson

This study is concerned with the process by which organizational knowledge is
created. Knowledge is widelyoacknowledged to be important as a resource or
competence of an organization, and this importance is seen to be increasing with
time. One result of this is a diverse and complex literature on Knowledge
Management that deals with activities such as production, absorption,
memorization, sharing and transferring of knowledge. While these activities are
relevant to knowledge at an individual level, they are less meaningful in the
context of organizational knowledge — that is, knowledge that is uniquely a

property of the organization as opposed to its membership.

This study ﬁnds that organizational knowledge is a unique combination of two
distinct elements and as such is not a thing that can be transferred. Instead,
there is a need to focus on the process of creation of knowledge. Existing
literature has been primarily content-based and fails to provide explanation of the
mechanisms involved. There have been repeated observations that the process

aspect is poorly understood and in need of further research.

This study seeks to provide explanation of the mechanisms of knowledge
creation by taking a process view. To do so, a case study approach was taken,

drawing data from three short-cycle process improvement initiatives (Kaizen

G‘I'ETTISI'
in an 6):
can Ede
dff'eren
process
The SIL
knowfe-

leaning

(J
._J

ﬂ)

(3
r—

Some '
disks,-

V

Chime

 

events). The analysis is both inductive and qualitative to support theory-building
in an exploratory context. Deriving insight from the inductive approach requires
considering the full range of evidence. To ensure a comprehensive view, three
different perspectives were used, corresponding to the central entities in the

process: knowledge itself, the individuals involved and the organization.

The study provides a grounded framework for understanding and explaining the
knowledge creation process. The result is a prototypical theory of organizational

learning that is elaborated by four propositions:

0 Organizational knowledge is a 2-part construct consisting of performance

routines coupled with an underlying logic of action.

0 The logic of action underlying a performance routine must be understood

as an element of the culture of the organization.

0 Learning, as in modiﬁcations to organizational knowledge, takes place

independently in the two components, and by different mechanisms.

0 Learning takes place through the actions of individuals who must reconcile

their paradigmatic and narrative views of the situation.

Some implications of these ﬁndings and directions for future research are
discussed with particular attention being directed to the role of organizational

culture and values.

Copyright by
JOHN DAVID HANSON

2006

Writing ‘
yaamc
ccntribu
this Opp.

beenins

Iwould I
chairper;
better dc
to Brian

Shown b

Finally, ;
Hanson.

Support

 

 

 

ACKNOWLEDGEMENTS

Writing the doctoral dissertation often seems to be the world's Ioneliest endeavor,
yet a moment’s reﬂection reveals that there are many people who have
contributed directly or indirectly to the completion of the task. I would like to take
this opportunity to acknowledge the efforts of those whose contributions have

been instrumental to the realization of my goals.

I would like to thank my entire committee, but special thanks are reserved for the
chairperson, Steven Melnyk, whose constant critiques have made this a much
better document than it would otherwise have been. Special thanks are also due
to Brian Pentland who has been a constant source of encouragement, and as

Shown by the reference list, a source of intellectual inspiration as well.

Finally, a Special place in these acknowledgements is reserved for Patricia
Hanson, my wife and companion of twenty-ﬁve years, without whose unswerving

support this entire venture would not have been possible, or even worth doing.

LEtof
Latof

Chapt

 

TABLE OF CONTENTS

List of Tables

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

....i.x

List of Figures x
Chapter One: Introduction 1
The Importance of Knowledge 1
Knowledge versus Organizational Knowledge 4

Focus on Routines 4

The Research Question 6
Positioning of the Research Question 7
Focus of the Research Question 8
Importance of the Research Question 9

The Manager’s Problem 10

The Researcher’s Problem 11

The Research Approach 13
Research Design 13
Qualitative Data Analysis 15
Research Protocol 16
Summary 17
Chapter Two: Review of Relevant Literature 19
Introduction 1 9
Process versus Content: Lessons from the Literature ........ 20

Part I: Development of Concepts and Constructs 23
Epistemology 24

Deﬁnitions of Knowledge 24

Typologies of Knowledge 28

Psychology 30

Cognition 30

Learning and Unlearning 33

Information Processing 34

Small Group Dynamics 35

The Firm as a Cultural and Institutional Entity 37

The Firm as an Economic Entity 41

Summary of Key Concepts and Constructs 44

Part II: Positioning of the Research 47
Learning Curves 49

Absorptive Capacity 53

Explaining Heterogeneity in Organizations 55

Chapter Summary 58
Chapter Three: Research Framework and Methodology," 59

 

vi

 

 

Char

Cha

Introduction ................................................................................................................ 59
Research Goal: Theory Building ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 6 0
Research Framework .............................................................................................. 6 4
Research Method _____________________________________________________________________________________________________ 6 7
Research Protocol ___________________________________________________________________________________________________ 6 8
Experimental Design ............................................................................................... 72
Case Selection ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 75
Sample Size ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 78
Research Subject: An Overview ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 80
Entering the Field: Data Collection ...................................................................... 83
Veriﬁcation of Data ___________________________________________________________________________________________________ 85
Analysis of Data and Conclusions ________________________________________________________________________ 86
Validation of Conclusions ....................................................................................... 87
Summary .................................................................................................................... 88
Chapter Four: Results ......................................................................................................... 90
Introduction ................................................................................................................ 90
Part I: Review of the Research Protocol _____________________________________________________________ 95
Knowledge Classiﬁed as Explicit or Tacit .............................................. 95
Transformations of Knowledge Within and Between Explicit
and Tacit Forms ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 97
Knowledge Classiﬁed by Location ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 100
Actions by Individuals to Retrieve or Store Knowledge by
Location ....................................................................................................... 102
Processing Of Knowledge and/or Information by Individuals
(Narrative Versus Paradigmatic) ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 103
Individual Cognition __________________________________________________________________________________ 105
Power and/or Knowledge as Moderators of Group Activity ,,,,,,,,,,,,, 107
Shared Logics of Action (Existence Of and Changes To) ,,,,,,,,,,,,,,, 108
Changes to Performative Aspects of Routines __________________________________ 112
Changes in Situated Performance ________________________________________________________ 112
Autonomous Learning .............................................................................. 114
Reactive Learning _____________________________________________________________________________________ 114
Management Inﬂuence ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 1 15
Summary of Findings ............................................................................... 116
Part ll: Development of Propositions ................................................................ 119
Tentative Propositions _____________________________________________________________________________ 120
Summary of Chapter Four ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 126
Chapter Five: Discussion and Conclusions ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 128
Introduction _____________________________________________________________________________________________________________ 1 28
Overview of the Findings ..................................................................................... 128
The Research Question _______________________________________________________________________________________ 137
Applicability and Limitations ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 139
Suggestions for Future Research ...................................................................... 143
Concluding Remarks ............................................................................................ 150

vii

References _ _ 153

 

Appendix I: Validation QuestiOnnaire ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 170
Appendix II: Case Report #1 ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 171
Appendix III: Case Report #2 ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 2 07
Appendix IV: Case Report #3 ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 2 39
Appendix V: Presentation to Expert Panel ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 2 65
Appendix VI: Validation Survey Results ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 2 68
Appendix VII: Notes of Discussion with Expert Panel ______ __ ..................... 269

 

Appendix VIII: Derivation of the Learning Curve from an Expected
Utility Model (5) ,,,,,,,,,,,,,,, _ 272

 

viii

LIST OF TABLES

Table 1: Research Protocol ............................................................................................... 70
Table 21 Summary Description of Cases ........................................................................ 91
Table 3: Qualiﬁcations of Validation Panel _____________________________________________________________________ 94
Table 4: Summary of Findings re: Research Protocol ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 116

Figure
Figure
Figure
Figure
Figure
Figure

Figure

 

LIST OF FIGURES

Figure 1: Articles on Knowledge Management ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 1
Figure 2: Model of a Theory ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 61
Figure 31 Research Methodology ..................................................................................... 64
Figure 4: Schematic of Research Framework ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 67
Figure 5: Process Improvement Event ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 73
Figure 6: Schematic of Relationships in Propositions _______________________________________________ 120
Figure 6: (Repeated) ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,, 129

helm

’

This re
their 0;

khgwfe

Few w:
is also
the ecc
increas
Mukhe:
PCpular

giowth

‘mq

 

 

 

CHAPTER ONE: INTRODUCTION
The Importance of Knowledge:

This research addresses the question of how organizations create knowledge in
their operations. In that context it is appropriate to begin by establishing that

knowledge is important, and to suggest why it is of interest to study its creation.

Few would challenge the premise that knowledge is important to the ﬁrm. There
is also a corresponding belief that the importance of knowledge to the ﬁrm and to
the economy at large is increasing with time, attributable primarily to the
increasing technical complexity of our world (Drucker, 1993; Bettis & Hitt, 1995;
Mukherjee, Lapre & Wassenhove, 1998). This belief can be seen in the
popularization of the term “Knowledge Management” and the year-by-year

growth in the number of articles published in that area (Figure 1).

 

Figure 1: Articles on Knowledge
Management

700
600
500
400
300
200
100

 

1997 1998 1999 2000 2001 2002 2003 2004 2005

(Mmber of refereed journal arﬁcles in the Ptoquest database containing
the term "Know ledge Management" in the title or abstract.)

 

 

 

The me
not he
BSEDE;
lieu/c
fundar
source
the fee
reasor
CenUa
source
stxenr
Leona:
1994; It
lander

Nahap;

AS note
generat
eiherc

 

and he
above
thelat:

be tea.

The recognition that knowledge constitutes an important resource of the ﬁrm is
not new (it can be seen clearly in Penrose, 1959 for example), but increased
attention to this fact has coincided with the emergence of the Resource-Based
View of the ﬁrm (RBV) in the ﬁeld of strategy. Arguing the economic
fundamentals of the RBV, Williamson (1981) makes the case that knowledge is a
source of economic rent. The basis for this argument is that knowledge is one of
the few resources whose use incurs no opportunity cost. Extending this
reasoning, Grant (1996) proposed the Knowledge-Based View (KBV) of the ﬁrm.
Central to this view is the argument that knowledge is a ﬁrm’s primary, if not only,
source of sustainable competitive advantage. Many researchers have made
statements to this effect (Prahalad & Hamel, 1990; Barney, 1991; Nelson, 1991;
Leonard-Barton, 1992; Cyert, Kumar & Williams, 1993; Henderson & Cockburn,
1994; Nonaka, 1994; Bates & Flynn, 1995; Nonaka & Takeuchi, 1995; Kogut &
Zander, 1996; Miller, 1996; Spender, 1996a; Davenport & Prusak, 1998;
Nahapiet & Ghoshal, 1998).

As noted above, the economic foundation of the KBV is that knowledge can
generate rents because its use incurs no opportunity cost. This can be true for
either of two reasons; the use of knowledge in one application may not preclude
its Simultaneous use in another, or the knowledge may be difﬁcult to replicate
and hence have no market value as an item of trade. The arguments cited
above for, the role of knowledge in creating competitive advantage hinge more on
the latter point than the former. To the extent that the knowledge of a ﬁrm cannot

be readily copied, it represents a source of sustainable competitive advantage.

'Th

TEL

book
vafue
come
that n
create
Consis

geﬂEra

We be;
Mafia 9.
Critltai

glOl'rtr v

a broa'

COhe‘”J

 

mem C:

The ideal Situation for a ﬁrm would be to have knowledge that was readily

reusable within the ﬁrm, but impossible to copy externally.

From a managerial perspective, it is of interest to assess the relative magnitude
of knowledge as one of the resources of a ﬁrm. A simple check on this
magnitude can be made by examining the price-to-book ratio of publicly-traded
ﬁrms as a proxy for the ratio of their total worth to the value of their tangible
assets. This ratio ﬂuctuates somewhat, but on August 4, 2005, the Morgan
Stanley Capital International (MSCI) US Broad Market Index (statistics available
at: hgpzllﬂagshiﬁxanguard.comNGApp/hnwlFundsHoldings?FundId=0085§FundIntExt=lNT) had a price-to-
book ratio of 2.7:1, suggesting that tangible assets make up less than 40% of the
value of a typical ﬁrm. Admittedly, the accounting system tends to be
conservative in its valuation of assets, and there are things other than knowledge
that make up the missing 60%, such as brand names and de facto monopolies
created by historical accident or legislative action. Nonetheless, this result is
consistent with the above citations which argue that the principal value-

generating resource of a ﬁrm is its aggregate store of knowledge.

We have previously noted the rapid growth in the literature on Knowledge
Management which is consistent with the both the view that knowledge is a
critical resource of the ﬁrm and that its importance is growing with time. This
growth was explicitly forecasted by Drucker (1993). Knowledge management is
a broad term lacking a universal deﬁnition, but was described by Amin and
Cohendet (2000) as consisting of the activities of production, absorption,

memorization, sharing and transfen'ing of knowledge. It should be apparent by

this point that, if we are to make much progress beyond sweeping
generalizations, we will have to be much more precise about what we mean by
knowledge. Wacker (2004) noted that common language or “dictionary”
deﬁnitions are rarely useful for research because they are insufﬁciently precise.
This is very much the case in this research, and we will develop more precise
deﬁnitions in Chapter Two. What we must establish at this point however, is the
distinction between organizational knowledge and all other forms that we

encounter.

Knowledge versus Organizational Knowledge:

A key issue of concern in the strategy literature is the appropriability of rents
(Collis & Montgomery, 1998). In the context of knowledge, it becomes an
important distinction whether knowledge is uniquely a property of the
organization (which would then capture the rents), or a property of the individuals
who happen to make up the organization at the time (and who would capture at
least a portion of the rents through above-normal salaries). In the work that
follows, the term organizational knowledge will be used to mean a form of
knowledge that is uniquely a property of the ﬁrm; something that the ﬁrm can
claim as its own and is robust against disturbances such as employee turnover.
In the framing of the research question it is organizational knowledge that will be

central to the research.
Focus on Routines:

Having established our interest in organizational forms of knowledge, we turn our

attention to the subject of routines. We do so because many researchers have

conch}
agent
2002}
as org
Simor
existe
reter t
opera

(1958

Since
organ?
and P.
action..-
Bacda

Ofanc

resista

 

concluded that routines are the means by which knowledge is embedded in an
organization (Levitt & March, 1988; Cohen, 1991; Amundson, 1998; Orlikowski,
2002). While no single researcher can claim priority on the concept of routines
as organizational knowledge, the idea is clearly present in the work of March and
Simon (1958) who observed that what we are calling organizational knowledge
existed in the form of performance programs, which we now more commonly
refer to as routines. These routines are fundamental to production and
operations management. While drawing attention to routines, March and Simon
(1958) did not describe how they were encoded in organizations or how they

were created.

Since then, progress has been made in understanding the anatomy of
organizational knowledge in the form of routines. Pentland and Rueter (1994)
and Pentland (1995a) characterized routines as consisting of a grammar of
actions that could be used to deﬁne a range of effective behavior. Cohen and
Bacdayan (1994) characterized routines as representing the procedural memory
of an organization, arguing that their inherently tacit nature made them highly
resistant to change and that they therefore represented a source of stability in
organizations. Note that in making the distinction between procedural and
declarative memory, Cohen and Bacdayan (1994), like Cohen and Levinthal
(1990), are drawing an analogy between the psychology of individuals and the
properties of an organization. The validity of this approach will be considered in
Chapter Two. Leonard-Barton (1992) also characterized routines as sources of

(potentially unwanted) stability by equating core capabilities with “core rigidities.”

some
perfc
aspe
they

know

are Cl
M
TMBp

honiu

devote

Feldman (2000) countered this view by suggesting that routines could include the
potential for variation and selection, and therefore evolutionary, incremental
change. More recently, Feldman and Pentland (2003) have decomposed
routines into “ostensive” and “perfonnative” aspects. This somewhat bridges the
two views (routines as sources of stability versus routines as agents of change)
by pointing out that the visible aspects of a routine are but a manifestation of
some underlying rationale. A great deal of change may be possible in the
perforrnative aspects of a routine without challenging the stability of the ostensive
aspect. While it is not fair to claim that these concepts are universally accepted,
they do go a long way towards describing the anatomy of organizational
knowledge. What remains relatively unstudied is the question of how routines

are created or changed.

The Research Question:

The preceding comment is motivated at least in part by the following quotation
from the introduction to a 1996 special issue of Strategic Management Journal

devoted to “Knowledge and the Firm”:

“The surge of interest into organizational capabilities and competences has
directed attention to organizationally embedded knowledge, but has made only
limited progress in understanding its anatomy and creation. ” (Spender & Grant,

1996)

While progress has been made Since then on understanding the anatomy of
organizational knowledge (described above, and in more detail in Chapter Two),

the issue of its creation remains poorly understood. This shortcoming with

respe
to be
disco
some
simp
that e
augrr
by St
preSE
resea

their

ﬁOCkg

The d

Wasnl

and .

respect to the understanding of the creation of organizational knowledge proved
to be particularly salient within the scope of this research. It was ultimately
discovered that every instantiation of organizational knowledge was unique in
some respects and could be seen as an instance of the creation process and not
Simply an application. This point was anticipated by Hayek (1945) who noted
that every circumstance was Speciﬁc and that generalizations must be
augmented with something else to be useful. A similar conclusion was reached
by Starbuck (1992) who stated that “the distinction between creating, applying or
preserving knowledge is very hard to make or defend.” This leads us to the
research question, broadly stated: “How do organizations create knowledge in

their operations?”

Positioning of the Research Question:

The operative word in this question is “how.” This positions the question as one
of process as opposed to content, a critical point of distinction between this
research and much that has gone before. A Shortage of research into the
process of organizational teaming has been noted by many researchers (Adler &
Clark, 1991; Miller, 1996; Amundson, 1998; Hatch & Mowery, 1998; Meredith,
1998). Fahey and Prusak (1998) offer the suggestion that Knowledge
Management Should be thought about in terms of ﬂows of knowledge rather than

stocks of knowledge; also a clear call for attention to the underlying process.

The distinction between process and content points of view in strategy research
was discussed by Huff and Reger (1987), who made a number of observations

and recommendations speciﬁc to process research. A key observation was that

iAnC‘
recc
Stun
resez
alacl
ofas
Organ
lteratt
Study ;

for QUE

Huberr

 

there were divides between those who study ﬁrms and those who study people
and between those who study economics and those who study behavior. One of
their major recommendations was to bridge these divides by importing concepts

from other ﬁelds such as organizational behavior and, psychology.

Another recommendation was to vary the use of research methods, a
recommendation consistent with the observation that process-oriented research
is underrepresented in the literature. AS noted by Meredith (1998), different
research methods lend themselves to different types of question. Consequently,
a lack of attention to process-related questions is consistent with his observation
of a shortage of case study and ﬁeld research. By focusing on the process of
organizational learning (the “how”), this study is a departure from much of the
literature and must make use of different research methods. Speciﬁcally, a case
study approach, using direct observation and qualitative analysis is appropriate
for questions of “how,” particularly when there is a lack of existing theory (Miles &

Huberman, 1994; Yin, 1994; Meredith, 1998).

Focus of the Research Question:

The research question; “How do organizations create knowledge in their
operations?” is a broad one that can be approached from many directions and
interpreted in many ways. At the outset it must be acknowledged that this
research is undertaken and described with a Managerial Interest as deﬁned by
Martin (2002, p. 169). What this means is that the research focuses on

processes that are of interest to managers because the desired outcome is a

base
unde
istha
other

age;

be de'

 

performance improvement as deﬁned by typical measures of productivity

improvement or cost reduction.

A further narrowing of the focus of the research is that it is research in
Operations Management. This is fully consistent with the managerial interest
and is motivated by two factors. One is that there are particular knowledge-
based problems in Operations Management that could beneﬁt from a better
understanding of process. These will be described in the next section. The other
is that Operations Management has been relatively Slow to adopt concepts from
other branches of management (Huff & Reger, 1987; Amundson, 1998), leaving
a gap that this research addresses in formulating the research protocol. This will

be developed further in Chapter Two.

Finally, the focus of this research is on theory-building, meaning theory creation
as opposed to theory-testing. This distinction is noted in part because of the
different treatments of the term theory-building by Hunt (1991) and by Wacker
(1998). This focus follows from the conclusion that there is a lack of existing
theory that is sufﬁciently explanatory, and is consistent with the contention of
Spender (1996b) that organizational knowledge, learning and memory are “three

concepts in search of a theory.”
Importance of the Research Question:

The foregoing has attempted to Show the importance of knowledge and of
organizational knowledge in particular. It has also argued that the process of
creation of organizational knowledge (equivalently: organizational learning) iS

poorly understood. It remains to be argued that an improved understanding of

the l

for ;

ratic
beta
argu
Tobi
a Sig

than

 

3C? '

the process will be of value, although this is clearly the rationale behind the calls

for process-focused research.

The Manager’s Problem:

Recall the Price-to-Book ratio of 2.7 cited above. The existence of a gap
between market and book values highlights an important problem. Price-to-book
ratio is the best proxy we have for Tobin’s Q, which is deﬁned as the ratio
between the market value of a ﬁrm’s assets and their replacement cost. Tobin
argued that this ratio should, on average, be unity (Brainard & Tobin, 1968;
Tobin, 1969). Unless Tobin was seriously mistaken, corporations are made up of
a Signiﬁcant percentage of assets that we simply do not know how to value, other

than through an auction process.

This is a well-known problem that is unlikely to be resolved, in part because of
the conﬂicting demands on the accounting system. The problem for the
managers of these corporations is that they are entrusted with a major asset, and
must manage it appropriately. They must maximize the returns from the
knowledge assets that they have and must acquire or replenish these assets as
necessary. In fact, Drucker (1993) offers the “right deﬁnition of a manager” (sic)
as “one who is responsible for the application and performance of knowledge.”
The problem is this: knowledge assets per se are not measurable (or there
would not be the gap between market and book values), yet many managers
believe that “if you can’t measure it, you can’t manage it.” (This statement has
been attributed to many writers but often, wrongly, to Deming, 1986, who was

actually making the counter-argument, describing the above problem.)

10

The resr
depend
(Ausbn.
moxyrr
main
dtear

mghsr

Anott
saesf
Mdudi
Corpo:
helei
meaSL
this re
DGCar
aept

Impm

The result is that managers and practitioners in Operations Management must
depend on proxies for knowledge. Proxies are problematic for several reasons
(Austin, 1996). The principal problems are that an increase in the value of a
proxy measure does not guarantee an improvement in the true objective, and
that it may also lead to unwanted effects. An example cited by Austin is the use
of tear strength of paper as a proxy for the propensity of the paper to break in a

high-speed printing press.

Another example of a proxy (for innovativeness) might be the percentage of
sales from products introduced in the last ﬁve years (used by many companies,
including the subject of this research; a more public example being Emerson
Corporation in their 2004 Annual Report, p. 6). A manager seeking proxies for
the level of knowledge might consider the training level of the staff (this was a
measure used in one of the divisions of the corporation that was the subject of
this research), or the number of patents awarded (Spender 8; Grant, 1996;
DeCarolis & Deeds, 1999), or R&D spending (Bettis & Hitt, 1995). All of these
are proxies for content with obvious potential for unintended results. An
improved understanding of the process of organizational learning has the

potential to reduce managers’ dependence on proxies for content.
The Researcher’s Problem:

The problems faced by researchers are related but slightly different. At the
strategic level, researchers are interested in issues such as ﬁrm heterogeneity
(“why do ﬁrms differ in the value of their knowledge assets?") and the role of

management in creating that heterogeneity (e.g.: Nelson, 1991; Conner &

11

 

Prahalad, 1996). There is a general premise that better management leads to
better performance (otherwise we should close all business schools), but lacking
useful measures of assets, and hence return on assets, researchers, like
managers, are forced to rely on “noisy" proxies such as share price appreciation
as indicators of managerial performance (Dess & Robinson, 1984; Hitt & Ireland,

1985).

At a more operational level, there is considerable interest in the question of why
ﬁrms differ in their ability to effectively implement concepts such as Total Quality
Management (TQM). The research in this area has been primarily content based
(e.g. Ferdows & De Meyer, 1990; Flynn, Schroeder & Sakakibara, 1995; Powell,
1995; Sakakibara, Flynn, Schroeder & Morris, 1997; Flynn & Flynn, 2004). From
a content perspective, the differences in results represent a puzzle because ﬁrms

adopting (for example) a TQM methodology are homogeneous in this respect.

This is not to imply that these researchers are unaware of the role of process; it
is simply a reﬂection of the fact that researchers are generally obliged to study
ﬁrms from the outside, where issues of process are not readily visible. It is the
lack of visibility into process issues that this study seeks to rectify. While the
results should be of primary interest to researchers in this ﬁeld, it is expected that

they will also be of value to managers and practitioners alike.

The Research Approach:

The foregoing has set out to establish that the process of organizational learning

is not well understood. That being the case, there is no established theory that

12

 

cen'
indl.
and
gror
obs«

undl

T6866
This 5
few

are r.

 

this research seeks to conﬁrm or justify. Rather, this research is aimed at theory

building, operating in the context of discovery as described by Hunt (1991).

Following further with Hunt’s typology of research, purposive theory building
offers the researcher the choice of inductive or deductive methods. Lacking any
central law or hypothesis that is presumed to apply, this research adopts an
inductive approach, comprised of the steps of observing, recording, classifying
and generalizing. This procedure is closely analogous the development of
grounded theory (Glaser & Strauss, 1967), in which theory is developed from
observation only. In this mode, the researcher enters the ﬁeld with “no theory
under consideration and no hypotheses to test” (Eisenhardt, 1989). This ideal
state creates certain problems which will be dealt with in Chapter Three, but the
key point is that the choice of approach has a major inﬂuence on the research

design.

Research Design:

The ﬁrst step in the inductive process of theory building iS observation, so the
research design must be one that provides sufﬁcient opportunity for observation.
This strongly suggests a case study approach, consistent with the
recommendation of Yin (1994) and Meredith (1998) who state that case studies
are well-suited to research questions of the “how and why” variety. Where it is
not the cases themselves that are of interest, but rather some underlying
commonality, the preference is for instnrmental cases as deﬁned by Stake
(1994). Instrumental means that the cases are simply vehicles for studying some

phenomena; the cases themselves are not unusual in any way, nor should they

13

 

be.
imp
Mor

bet

motir
precl

admt

be. Ultimately, three cases were used, consisting of Short-cycle process
improvement projects of the type known as Kaizen events (Melnyk, Calantone,
Montabon & Smith, 1998). The issues of actual case selection and the criteria to

be used will be discussed in Chapter Three.

Case study is not in itself a method (Stake, 1994); it still remains to select the
method for collecting data. For this research, the choice was made to employ
direct observation as opposed to surveys or interviews with informants. This was
motivated by two factors: a) the grounded nature of the research, which
precluded the assumption of speciﬁc constructs and propositions, and b) the
admonition (paraphrasing Melnyk & Stewart, 2000) that one “cannot understand
a process by having participants discuss what they think is going on.” In
particular, it was decided to employ a participant-observer model. This is a
model often employed in ethnographic research (Martin, 2002, pp 48, 210) which

bridges the “insider” and “outsider” views of an organization.

This choice was based on the premise that interaction with the participants
would: a) result in greater insight and b) cause them to lose the sense of being
watched, and therefore behave more normally. The former point was addressed
by Bartel and Garud (2003) when they stated that “to see and understand (the
exchanges that take place), the researcher must become semi-native.” The
latter point was actually validated in one of the case studies when it was
necessary for the project team to collect time-study information on a process.
When the team members observed the production workers in a passive way, the

workers performed in a stylized fashion, Showing great energy but omitting key

14

Ste:

anal

hat

pane

 

steps. It was only when the team started discussing the process with the
workers in a more engaged and informal way that a true picture of the work

process started to emerge.

Qualitative Data Analysis:

Having deﬁned the research approach to this stage, it was evident that the
analysis of data would have to be qualitative in nature. The need to consider a
wide range of possible explanations for the observations is actually one of the
strengths of qualitative analysis (Miles & Huberman, 1994). The exact process of
analysis is what Kaplan (1964) and Hunt (1991) describe as pattem matching.
That is, the recorded observations are scrutinized for repeating themes and

patterns of potential interest.

The human brain is good at seeing patterns in data — perhaps too good. As a
result, this method suffers from problems with the criterion that Hunt (1991)
refers to as intersubjective conﬁnnability. Simply stated, this means that two
researchers may not extract the same patterns from a common data set. This
problem disappears with time as a ﬁeld converges on an accepted set of
validated constructs, but iS unavoidable at the exploratory stage where this
research is positioned. To mitigate this potential threat to validity, the themes
extracted were reviewed with a panel of experts to determine consistency with

their experience.

15

The
the t
rese
mea.

not.

Research Protocol:

It was stated above that the inductive method of theory-building begins with
observations, but the question is: observations of what? If one starts with “no
theory under consideration and no hypotheses to test” (Eisenhardt, 1989), there
is a great danger of recording too much of what ultimately proves to be
uninteresting at the expense of that which, retrospectively, is interesting. A
compromise to the ideal of grounded theory is required and this is ideally done
through the development of a sound research protocol as elaborated by Ellram
(1996). Development of this protocol is one of the central purposes of Chapter

Two.

The research protocol for this study consists of an event checklist (Ellram, 1996),
the usage of which was twofold. Prior to case observations it is a tool for the
researcher to become sensitized as to what to look for. Post facto, it provides a
means of classifying data as representing an instance of some phenomenon, or
not. To fulﬁll these roles, the protocol requires two characteristics. First, it must
not only specify what is to be looked for, but also how it might be recognized.

Second, it must cast as broad a net as is feasible.

The latter task is the more difﬁcult and, as will be seen in Chapter Two, it is
addressed by examining the issue of organizational learning from three
perspectives. These perspectives correspond to the identiﬁable entities in the
learning process: the individual(s) involved, the organization being studied, and
knowledge itself as a distinct entity. Interestingly, these perspectives draw upon

distinct Iiteratures with very little overlap. If one accepts the proposition that the

16

prc

ca 5

ex;

TBC

COl‘.

grai-

WEV

cont

of its

 

process of organizational learning is not well understood, then it must be the
case that none of these streams of literature has, by itself, provided a satisfactory
explanation. Combining the different perspectives in a single protocol not only
improves the breadth of the observations, it also responds to the
recommendation of Huff and Reger (1987) and Amundson (1998) to integrate

concepts from other ﬁelds.

Summaﬂ:

This introduction has sought to establish a number of points, ﬁrstly that
knowledge is important to an organization or ﬁrm and that this importance is
growing. Appropriable knowledge is particularly important and this deﬁnes what
we will refer to as organizational knowledge. Secondly, while the nature and
content of organizational knowledge are at least partially understood, the process

of its creation, organizational learning, is much less well understood.

It is argued that a better understanding of organizational learning will address a
number of problems for researchers and practitioners. In particular, an improved
understanding of the process variables may allow a reduction in dependence on
outcome measures that are but proxies for learning. While models based on
outcome measures (such as a learning curve) may have predictive value, they
fail the test of explanation (Schmenner & Swink, 1998). It is the ultimate goal of

this research to supplement them with more explanatory process models.

The central research question addressed by this research is: “How do
organizations create knowledge in their operations?” The approach taken by the

research to answer this question is inductive theory-building. A logical

17

conse
case :
thhzh

pnhoc
The b

Chapt
onto:
These
ogam

ctthei

Chapte

mettx

Chapte
II‘ITEe c;
summa

Pl0posr

 

consequence of this approach iS the use of a participant-observer model in a
case study setting, with qualitative analysis of data. Potential threats to validity in
this approach are addressed by the development of broadly-based research

protocol and the use of an expert panel for validation of speciﬁc ﬁndings.
The balance of this report is organizes as follows:

Chapter Two is primarily devoted to development of a research framework and
protocol based on a reading of the literature from three different perspectives.
These perspectives correspond to the individuals involved in the process, the
organization in which it takes place, and the knowledge that is created as a result

of the organizational learning process.

Chapter Three describes in more detail the development of the research

methodology, including the research protocol and the case selection criteria.

Chapter Four contains the summaries and analysis of the data recorded from the
three case studies and from the discussions with the expert panel. These
summaries are integrated, following the research protocol and a set of

propositions are derived.

Chapter Five discusses the signiﬁcance and implications of the propositions. In
particular, the implications are starting points for the development of hypotheses
that can be tested in other settings. Limitations with respect to generalizability

are discussed and suggestions for additional research are presented.

18

 

lntrod

_____—
lnintr
knovﬂ
creati
bus 51

pOSﬁh

Ttus c
ohec
learnt
obhec‘
might
obser
thata
tnoce

IthuE
The ﬁt
togehi
ways!

heir

ME}

de,

CHAPTER TWO: REVIEW OF RELEVANT LITERATURE:

Introduction:

In introducing the research question, Chapter One sought to establish that
knowledge is important to the ﬁrm, and that understanding the process of its
creation Should be of interest to both researchers and practitioners. In so doing,
this study was positioned as one focusing on process as opposed to content; a

positioning that sets the requirements for this chapter.

This chapter is devoted to a reading of the literature, and has three speciﬁc
objectives. The ﬁrst is to establish deﬁnitions of organizational knowledge and
teaming that are consistent with and grounded in the literature. The second
objective is to cast a broad net to identify a list of key concepts or constructs that
might be expected to be relevant to the research question and fruitful as areas of
observation. The third objective is to identify and survey the streams of literature
that are constrained by the limited understanding of the organizational learning
process. It is these streams of literature that this research proposes to extend

through improved insight into the process.

The ﬁrst objective falls naturally out of the second, so these will be treated
together. The concepts that are identiﬁed from the literature will be used in two
ways, consistent with the inductive approach to theory-building (Hunt, 1991). In
the inductive approach, one ideally begins with observations. However, there is
a need to establish certain screens for deciding what observations will be of
interest and which will not. Therefore, one use of the concepts will be the

development of a research protocol. In this instance the protocol will take the

19

fr;

pr

Th

reﬁt
lher
the;
of ti
insh
Hovr

requr

Proc-

ID the

 

form of an event checklist (Ellram, 1996) that will ensure that the questions of
interest are investigated uniformly in all cases. The second use will be the
development of a research framework. The concepts represent potential
constructs to be used in theory-building and are the basis for the process of
classiﬁcation in inductive theory-building (Hunt, 1991). As elements of a
framework, they deﬁne the range of relationships that can be expressed as

propositions of a tentative theory.

This chapter will be organized in two main parts. The ﬁrst part will identify the
relevant concepts and deﬁnitions. The second will identify those streams of
literature that are intended to be enhanced by this research. It will be noted that
there was considerable reference to the literature in establishing the importance
of the research question in Chapter One. That work will not be repeated here;
instead, this chapter will be focused on the three purposes described above.
However, before proceeding with those tasks there iS an additional point

requiring elaboration.
Process versus Content: Lessons from the Literature:

In the introduction, this study was positioned as one of process as opposed to
content. A number of references were cited to make the case that this is an
important difference with implications for the research design and that process-
based research was lacking. A recurring theme that will be seen in this literature
review, particularly in Part II, is the call for more research, or at least better

understanding, of the process of organizational learning.

20

pr
pn

the

Somi
Macli
organ
of the
of the
were p
Roos. '
and 8;
do not

Cohte I

 

Ohep‘

With that in mind, the literature was examined for lessons learned with respect to
process-oriented research. The immediate conclusion was that there is very little
process-oriented research. The great majority of research that even addresses
the issue of process in organizational learning is, in essence, attempting to infer
process by examining outcomes. Unfortunately, in the presence of equiﬁnality
(Hambrick, 1984), the results of such inference will be indeterminate. The
majority of the works cited in Part II of this chapter could be cited as examples of
this approach. A noteworthy example is Adler & Clark (1991), cited not as an
object of criticism, but as an instance of an explicit recognition of this problem
and a recommendation to pursue process-based research. The overriding

conclusion is that to understand process, one must observe process.

Some good examples of good process research were noted, Speciﬁcally
MacDufﬁe (1997) and Sinclair, Klepper and Cohen (2000) with respect to
organizational learning in Operations. In the general context of Operations, one
of the most comprehensive pieces of process research ever undertaken was that
of the International Motor Vehicle Program at MIT (lMVP), the results of which
were presented in “The Machine That Changed the World” (Womack, Jones &
Roos, 1991). Other good examples are Ward, Liker, Cristiano and Sobek (1995)
and Spear and Bowen (1999). In looking at these examples, we note that they
do not generate much new theory in spite of the richness and usefulness of their
content. As a result, they tend to be published as books or in more practitioner-

oriented journals, and with the exception of Womack, do not get cited very often.

21

Thel
rese
dLJTE

irnvc

EEvea
bun:
cont
ant:

3T8!

tiavir
learni
thefo
achiet
a Close
(Doerat

descﬁb

There are lessons here about the difﬁculty of doing good process oriented
research. Speciﬁcally, it is difﬁcult to do studies of sufﬁcient depth, breadth and
duration to progress beyond single-sample anecdotes. Note that the IMVP study

involved 55 researchers, took ﬁve years and cost $5M (in 1980's dollars).

Even with those resources to address the problems of breadth and duration,
building theory is a challenge because of the problem of establishing adequate
controls. Organizations are open systems and it is rarely possible to be sure that
all of the relevant factors have been observed. Unless the subjects of the study

are naturally bounded in some way this will always remain as a question.

Having taken up the challenge of observing the process of organizational
learning, these lessons have to be taken into account if the research is to allow
the formation of even tentative theory. In particular the critical issues are to
achieve a degree of repeatability in the research setting while also approximating
a closed system. In effect, what is needed is a virtual laboratory, set in the
Operations world. These issues will be addressed in the research design,

described in Chapter Three.

22

ﬁta
ﬁan
arm
the

SGTT

iAsv
Eﬁdn
VieWI
resuh
inuno

mEnbf

(J

()

C)

0

 

PART I: DEVELOPMENT OF CONCEPTS AND CONSTRUCTS:

A disadvantage of a process point of view relative to a content-based analysis is
that it requires more data. Consideration must be given to all of the factors that
form the mechanism of interest. In seeking to develop a list of relevant concepts
and constructs with which to populate a research framework we are faced with
the challenge of casting a sufﬁciently broad net while maintaining some

semblance of coherence and order.

As was suggested in the introduction, this study will attempt to do just that by
taking a subject-object view of the process. In other words, the process will be
viewed from the perspective of the principal agents and/or objects involved. This
results in three major perspectives from which to observeand interpret a process
improvement step. For each of these points of view, relevant constructs can be
identiﬁed from the literature. These points of view are:
0 The knowledge that is created, used or transformed in the process.
0 The individual person(S) involved in the process.
0 The organization within which the process occurs. The organization view
is two-fold: as an economic entity, it provides the purpose, and therefore
the deﬁnition of process improvement. The organization also deﬁnes a
culture or institution within which the individual(s) operate and learning

takes place.

Interestingly, each of these points of view is supported by distinct and largely
non-overlapping ﬁelds of literature. Speciﬁcally, these ﬁelds are the various

branches of epistemology, psychology and organization theory, respectively.

23

Sur
see:
only
inter:
concr
htdtvh
intere
the inc

CORD€<

With th
constru
to a pro

literaturr

M

We begir
with an In
lshslmpl
(1995) n,
the Datat
“Genera
points 0 I

has Drov

Surveying any of these ﬁelds would be a vast undertaking in its own right, but the
scope of the problem can be substantially reduced by noting that we are really
only interested in the intersections of these areas. For example, we are
interested in the psychology of the individual, but only to the extent that it
concerns the processing and absorption of knowledge or the interaction of the
individual with the organization. On the subject of knowledge itself, the prime
interest is in knowledge in an organizational context. Knowledge in the context of
the individual may be of interest to the extent that it has an identiﬁable

connection to organizational knowledge.

With the ﬁeld narrowed in this way, the following is a search for the relevant
constructs that can be used to develop a research framework that will in turn lead
to a protocol or event checklist. An important additional goal is to derive from the

literature appropriate deﬁnitions of organizational knowledge and learning.

Epistemology:
DreﬁnitionsL of Knowledge:

 

We begin with the study of knowledge itself, or epistemology, and are confronted
with an immediate problem. Does knowledge have an independent existence or
is it simply a descriptive property of some other entity? Nonaka and Takeuchi
(1995) note that Western philosophy has been dominated by the Cartesian Split,
the paradigm of Descartes in which knowledge, and by extension truth, are
independent absolutes which can be discovered and proved. Polanyi (1958)
points out that this assumption underpins the scientiﬁc method as we know it and

has proven to be useful and effective in that context. However, he ultimately

24

conc
uniqt
just/t?
isthe
persc
This p
Amin
memc
the ter
indepe
but the
that Ne
dealby
Persona
not the g
35 a wan
35 8 thing

i"iliortant

To addre.

and Who/9

J

Observe?

starting it

concludes that this view iS false and that knowledge is, in the ﬁnal analysis,
unique to the knower. In this he falls back on a deﬁnition of knowledge as
justiﬁed belief, a deﬁnition that Nonaka and Takeuchi (1995) attribute to Plato. It
is the role of belief and the context of justiﬁcation that makes this an individual or

personal form of knowledge.

This presents a difﬁculty for Knowledge Management as a ﬁeld (described by
Amin & Cohendet, 2000 as consisting of the activities of production, absorption,
memorization, sharing and transferring of knowledge). These terms, and even
the term Knowledge Management itself, imply a reiﬁed view of knowledge as an
independent entity. This clearly conﬂicts with the conclusion of Polanyi (1958),
but the distinction may not be signiﬁcant in practice. For example, we all know
that Newton’s laws of motion are not quite right, but we have achieved a great
deal by accepting them as correct. In a relatively homogeneous society, the
personal nature of knowledge may not be particularly visible or important. It is
not the goal of this study to take sides in this debate, but the discussion serves
as a warning that words can create a trap. If we speak too freely of knowledge
as a thing to be managed (stored, retrieved, applied, etc.), we may fail to see

important issues.

To address issues of knowledge in this research, we need suitable deﬁnitions
and typologies of knowledge so that we can classify and organize our
observations. Many deﬁnitions of knowledge exist with respect to an individual,

starting with “justiﬁed belief" as discussed above. Another example would be

25

orga
leave
(1 99*:
ascﬁt
Teece
organr}
assemt
Know/e
form oft

PIOperty

TO arrive
Pentland
SPEC/float:
This SChJ

Wiiiings (C

 

“the subjective meaning of the known to the knower" (offered as one of many

possible deﬁnitions by Machlup, 1980).

However, our interest is primarily in knowledge as a property of an organization,
and it is not clear that deﬁnitions centered on individuals are relevant when
applied to organizations. Cyert and March (1963) ﬁrst proposed that an
organization could learn independently from its individual members but did not
leave us with a comprehensive deﬁnition of organizational knowledge. Grant
(1996), in proposing the Knowledge-Based View of the ﬁrm, avoided the issue by
ascribing knowledge purely to the individuals comprising the organization.
Teece, Pisano and Shuen (1997) also avoided the issue by arguing that what the
organization possesses is the complementarity of the knowledge of individuals
assembled within the organization. Nonetheless, the term Organizational
Knowledge is very much part of our vocabulary and many feel that it is a distinct
form of knowledge. Speciﬁcally, Hedberg (1981) describes it as being a “robust

property” of an organization.

To arrive at a useful deﬁnition of organizational knowledge, we follow the lead of
Pentland (1992), who drew on the philosophy of American pragmatism, and
Speciﬁcally the work of George Herbert Mead, William James and John Dewey.
This school also rejects the Cartesian view of knowledge and it is from their
writings (speciﬁcally Dewey, 1916 and Dewey & Bentley, 1949) that we derive

' the idea that knowledge is Situated, that is to say: deﬁned only relative to a
Situation. Accordingly, we adopt the deﬁnition of organizational knowledge as

situated performance, which means the ability to deliver a satisfactory

26

pt
(15

beu
the
san

San)

(Dnci
pent
‘wouh
ekunr
SChOi
Ushnt

0f Siiug

 

enter
deft."

”$6

performance in a deﬁned set of circumstances. This is a functional rather than a
descriptive deﬁnition, but is made manageable by the fact that organizations
generally have a deﬁned purpose that deﬁnes the Situation. Note that there are
important parallels between this formulation of knowledge and the treatment of
cognition in the psychology of individuals, which will be addressed in the next

section.

Many equivalent deﬁnitions or statements about organizational knowledge have
been made. Amin and Cohendet (2000) speak of the ﬁrm as “a social institution,
the main characteristic of which is to know well how to do certain things. ” In the
same vein, Orlikowski (2002) equates knowledge with “effective action” as do

Sabherwal and Becerra-Fernandez (2003).

Once we accept the deﬁnition of organizational knowledge as situated
performance we can derive a working deﬁnition of organizational Ieaming which
would be: an increase in situated performance. This deﬁnition lacks a certain
element of volition, so it will be supplemented by one proposed by Argyris and
Schon (1978): the detection and conection of error. Equivalence of the
deﬁnitions can be established by regarding “error” to be any unsatisfactory level

of situated performance.

Even though our primary interest is in organizational knowledge as deﬁned
above, it is clear that knowledge as a property of individual persons may also
enter into the research question. AS noted above, there is a vast range of
deﬁnitions that can be applied to individual knowledge, but there is no particular

need for us to be too precise in deﬁning that term. As long as our focus is

27

(E
kr
tht

[MJI

knc:

In:
The

1991
(proc
expha

ofknc

Many

shape;
Oneof
bctve
Somehi
Ghent!

notde

Ttpoloi
we Wat
COmm
Vehuei

Codii'i

restricted to the activities of individuals that have an impact on organizational
knowledge as deﬁned, we need not be concerned about whether the content of
these activities meets a speciﬁc deﬁnition of knowledge. For classiﬁcation
purposes, multiple forms of individual knowledge can coexist with organizational

knowledge.
Iypologies of Knowledge:

The second step, after observation, in the inductive process of discovery (Hunt,
1991) iS classiﬁcation. While we are primarily interested in classifying the actions
(processes) that lead to changes in organizational knowledge, the task of
explanation is simpliﬁed by the ability to distinguish between the different aspects

of knowledge.

Many typologies of knowledge have been put fonlvard, their dimensions being
shaped by whether one subscribes to the reiﬁed or situated views of knowledge.
One of the most enduring is that of Polanyi (1958) who introduced the concept of
tacit versus explicit knowledge. In his meaning of the term, tacit knowledge is .
something unique to the knower, which by its very nature, cannot be articulated.
Given that this usually implies some sort of motor skills or neural learning, it is

not clear that there is an organizational analog of this concept.

Typologies of knowledge that are speciﬁc to an individual are problematic when
we want to speak about an organization. In an organizational context, it is more
common to adopt a reiﬁed view and classify knowledge according to its form or

venue. Winter (1987 p. 170) offers such a typology based on forms of

codiﬁcation in which one of the categories is tacit knowledge, by which he simply

28

The
the i
setti
by Cl
ofkn
convi
know
more
are nc
tacit k
be am

that in:

Ellister
an 0ng
indepg
”lemc J

Wife 3'

apptc

means “not codiﬁed,” or resident only in communal memory. This is Similar to
how Walsh (1995) describes knowledge in an organization as being deposited in
different “bins.” Using this view of knowledge, it would make sense to analyze
organizational learning by observing the ﬂow of identiﬁable pieces of knowledge

to and from various bins.

The classiﬁcation of knowledge as being either tacit or explicit does depend on
the individual(s) involved, but the concept can be used in an organizational
setting. Nonaka (1994) attempts to do so in a theory that knowledge is created
by conversions, speciﬁcally conversions between or within tacit or explicit types
of knowledge. Using this view, Ieaming would be described in terms of the
conversions that take place. While Nonaka (1994) aligns his deﬁnition of tacit
knowledge with that of Polanyi (1958), Tsoukas (2003) counters that the usage is
more akin to that of Winter (1987), and as a result, the transformations involved
are not truly different in kind. In developing the research protocol, we will discuss
tacit knowledge, and will use a compromise deﬁnition: knowledge which cannot
be articulated by the person involved. We will not attempt to distinguish whether

that inability is a function of the person or the nature of the knowledge.

Epistemology thus gives us two fundamental ways to characterize knowledge in
an organizational context, but both depend on a reiﬁed view of knowledge as an
independent entity. One is classiﬁcation by location (written records, individual
memory, etc.). In such a typology, the Speciﬁc “bins” would be chosen to
correspond to the environment in which activities take place. The alternate

approach is classiﬁcation by type, where the state of knowledge changes through

29

con
tacit
the l

{Mas/t

/\s n:
purpc
reiﬂec
parag'
corres
be disc

circum:

Psychr

The rea
queshor
Organize
Psychol
iSaINO
leasoni
a {Biailv

f’Olh its 2

conversions (tacit to explicit, explicit to explicit, etc). An expanded deﬁnition of
tacit is used here to ensure relevance in the organizational setting. As noted in
the Deﬁnitions section, individual and organizational knowledge can both be

positioned within these typologies.

As noted above, we have adopted a situated deﬁnition of knowledge for the
purposes of this study, but have also acknowledged the potential usefulness of a
reiﬁed view for practical observations. This is the basis of the preceding
paragraph. However, if these typologies are found to be not useful, there is no
corresponding typology of Situated knowledge. In that case, knowledge can only
be discussed as a property of the individual or the organization in a speciﬁc

circumstance. These entities are the subjects of the following sections.

Psychology:

Cognition:

The reader will have noted from the section on epistemology that is an open
question as to whether knowledge is an identiﬁable independent entity in the
organizational learning process. This uncertainty has an exact counterpart in the
psychology of the individual; speciﬁcally in the functioning of cognition. Cognition
is a broad concept, deﬁned by DeFillippi and Ornstein (2003) as “thinking,
reasoning and memory.” AS with knowledge, the question is whether cognition is

a relatively independent property of the individual or whether it is inseparable

from its environment or situation.

30

The
that
impi
was
Sign:
to th
Lyles
(199‘
scien.
led to

by Pyl

Organ
theoris
envhor
sense;
in this d
cliterate:
the Orga
pJana
and De
- only i,
adaptarJ

Occurs J

The proponents of the former view coined the term cognitive science and argue
that human cognition should be understood primarily in terms of (admittedly
imperfect) mathematical processing. A telling quotation is that “the computer
was made in the image of the human” (Simon & Kaplan, 1989). In 1991, a
Signiﬁcant amount of published work in organizational learning could be assigned
to the pure cognitive or “neo-rationalist” school (term used by Easter’oy-Smith &
Lyles, 2003). Examples include: Epple, Argote and Devadas (1991), Huber
(1991), March (1991), Simon (1991). Since then, the attention of cognitive
science has somewhat turned away from organizational learning and has instead
led to important advances in artiﬁcial intelligence and expert systems as signaled

by Pylyshyn (1984).

Organizational learning has become much more the province of the practice-
theorists, who argue that learning is Situated in and inseparable from its
environment. In other words, a person does not “know" something in an absolute
sense; they only “know” it in the context of their environment. The environment
in this discussion is generally taken to be the culture in which the individual
operates. We will have more to say later about culture from the point of view of
the organization. This view has its origins in the same school of American
pragmatism that gave us Situated performance, speciﬁcally Dewey (1916; 1933)
and Dewey and Bentley (1949), who argued that there is no absolute knowledge
— only that which is rooted in experience. As a result, learning is situated, “an
adaptation of a person or group to features of the situation in which learning

occurs” (Rosch & Lloyd, 1978).

31

Mmen
tothei
mganh
commt
Exam;

Balzan

 

More recent years have seen a great increase in the amount of research devoted
to the impact of culture on cognition and its corresponding impact on
organizational learning. Much of this literature has been based on the concept of
communities of practice as the speciﬁc culture in which this takes place.
Examples of research in this stream are: Shaw and Alley (1985), McCabe and
Balzano (1986), Suchman (1987), Lave (1988), Brown, Collins and Duguid
(1989), Greeno (1989), Orr (1990), Rogoff (1990), Brown and Duguid (1991),
Lave and Wenger (1991), Cook and Yanow (1993), Nicolini and Meznar (1995),

Wenger (1998), Bartel and Garud (2003).

In making the case for Situated cognition (and against cognitive science)
Hutchins (1995) counters the above-cited quotation from Simon and Kaplan
(1989) by saying that “cognitive science has tried to remake the human in the
image of the computer.” He goes on to argue (p. 354) that “the understanding of
the individual that has developed without consideration of the cultural process iS

fundamentally ﬂawed.”

The above discussion has related to the role of cognition in the individual, and
the impact of the culture or organization in inﬂuencing it. An open question is
whether organizations can be said to exhibit the same cognitive properties as
individuals. Some researchers argue that this analogy can be made. Cohen
(1991) and Cohen and Bacdayan (1994), for example, cite numerous examples
to argue that organizations and individuals exhibit very Similar modes of
behavior, at least on the surface. The role of cognition has been extended from

the individual to small groups by many researchers (see Walsh, 1995 for a

32

disan
under
eﬂen
chaar
uniere

thSzr
Learn

Asind
oneta
sehch
thatkr
contnm
isalos
aDproa

Ailgote

Oil the
khOWIec
stuahor
ehectiv
80mg f
situate r.

mehee

liter,

discussion of this literature — and its limitations) and to organizations as whole
under the term Absorptive Capacity (Cohen & Levinthal, 1990). While these
extensions appear to be justiﬁed on the basis of their outcome measures, it is not
clear that the underlying mechanisms support this. For process-based research
where it is the mechanisms that are being studied, it is not necessary to make

this assumption in advance.

Learning and Unlearning:

AS individuals and organizations team, so can they forget (“unlearn”). The view
one takes of this mechanism is highly dependent on the view of cognition that is
selected. In the pure cognitive view (cognitive science), one must take a position
that knowledge is good and that it iS cumulative - more is better. Things that
contradict each other cannot both be knowledge. This being the case, forgetting
is a loss of knowledge and, by extension, a loss of performance. This is the
approach taken by Argote, Beckman and Epple (1990), Argote and Epple (1990),

Argote (1999) in modeling a forgetting process.

On the other hand, situated cognition or practice theory suggests that what is
knowledge at one moment may cease to be in the next as the perception of the
situation changes. To the extent that neither individuals nor organizations can
effectively reconcile mutually contradictory paradigms, this view suggests that
some form of unlearning may be a necessary prerequisite for an increase in
situated knowledge as we have deﬁned it. A number of researchers have noted
the need for unlearning: Duncan and Weiss (1979), Weick (1979), Hedberg

(1981), Lyles (1988), Lyles and Schwenk (1992), Schein (1993), Bettis and

33

T85
var
ren
TBS

knc

Stilt
proc
degr
rese
proo
thoui
while
He cl
slory'
Lapre
DIOCe:
disting
Brieﬂy
ttpen'

JS (1813.

Prahalad (1995), Pentland (1995), Bettis and Wong (2003). Many of those
researchers have also noted the difﬁculty of unlearning and have offered a
variety of explanations as to why that should be so. These ﬁndings serve as a
reminder that the process of creating knowledge that is the subject of our
research question may equally involve activities that destroy or displace

knowledge.
Information Processing;

Since transformations or ﬂows of knowledge will depend on information
processing by individuals, the role of processing must be understood. To a large
degree, this dovetails with the discussion of cognition but there is additional
research that sheds some light on the underlying processes. With respect to
processing of information, Bruner (1986, p. 11) describes “two modes of
thought,” which he categorizes as nanative and paradigmatic. He argues that
while both modes can be present, neither is ultimately reconcilable with the other.
He characterizes the difference between them as the difference between a “good
story” and a “logical argument,” respectively. Walsh (1995) and Mukherjee,
Lapre and Wassenhove (1998) make similar arguments about dual modes of
processing, but with somewhat more descriptive terminology. Walsh
distinguishes between top-down and bottom-up processing of information,
Brieﬂy, in the top-down mode, data are interpreted on the basis of prior
experience and knowledge structure (justiﬁed beliefs), while the bottom-up mode

is data—driven and analytical in nature.

lnhg
proc
woul
(199
macl
ques
lWes
nana
moot
gWen
conm
the u:
dmen

cmha

BOhH
the kr
lebhc

There
Ihdi‘rid
behav

emar

 

 

In light of the above discussion on situated learning, the narrative role of
processing is expected to play an important role, for it is in this mode that we
would expect to see the cultural context of Ieaming being transmitted. Wenger

( 1998) suggested that narratives were the means by which communities of
practice Shared their knowledge. Bartel and Garud (2003) asked a similar
question to ours: “how do actors create and apply knowledge in their daily work
lives?” Although they do not answer the question directly, they suggest that
narratives are an important vehicle by which this occurs. They suggest a
process of abduction by which individuals can extract multiple lessons from a
given narrative, based on the nuances of context. It is this transmission of
context that distinguishes narrative processing from paradigmatic. As implied by
the use of the term abduction, the extraction of lessons from a narrative is not
deterministic, but rather, situational. As a consequence, a “good” narrative must

contain some “evaluational reference” (Pentland, 1999).

Both narrative and paradigmatic modes of processing are potentially important in
the knowledge creation process, but their relative importance and their
relationship to each other are matters of interest when considering the roles of

individuals in the organizational Ieaming process.

Small Group Dynamics:

There is one additional point we wish to cover with respect to the role of the
individual in our research setting. The above discussion has signaled that the
behavior of individuals is expected to be inﬂuenced by the surrounding culture -

emanating from either the organization as a whole, or perhaps from some

35

 

CORT
rese
invol

HOTTT

The

theh
noth
ﬂudy
Ourr

hndi

None
Penor
Ofvar
QDUp
thp
Uhkp.
QDUp

Neae!
Hum,(
IMEpS
piOCeg

QTOupgi

hhnr

community of practice within it. There is an additional level of interaction in our
research settings, and that is the interpersonal dynamics of the individuals
involved in the cases being studied. While these may reﬂect broader cultural

norms, there may be other dynamics at work.

The internal dynamics of small groups have been much studied with respect to
their effectiveness. The focus on effectiveness suggests a content focus that is
not the primary interest of the present research. The research question in this
study concerns itself with what these groups do, not how effectively they do it.
Our null hypothesis is that the groups in question will be representative of their

kind in terms of their activities and effectiveness.

Nonetheless, it is useful to be aware of the key determinants of group
performance as reported in the literature as a way to recognize potential sources
of variance in the research settings. Some commonly studied parameters are:
group Size (Cummings, Huber 8 Arendt, 1974; Yetton 8 Bottger, 1983;
Littlepage, 1991; Hwang 8 Guynes, 1994), task type (Kabanoff 8 O'Brien, 1979;
Littlepage, 1991; Straus 8 McGrath, 1994) and power relationships within the
group based on status or recognized expertise (Gruenfeld, Mannix, Williams 8
Neale, 1996; Franz 8 Larson, 2002; Okhuysen 8 Eisenhardt, 2002; Thomas-
Hunt, Ogden 8 Neale, 2003). Some observations about group dynamics are
interesting as they relate directly to the issues of knowledge and knowledge
processing. Stasser and Titus (1987) noted that the greater the degree to which
groups Shared a common base of knowledge, the less unique or divergent

information was brought forward. Several explanations for this phenomenon are

36

poss
corni

CODIt

sown
wouk

proce
Titi

Mostl

mshe

 

possible, but the point of interest for the present research is that group
composition has the potential to change the process of learning as well as the

content.

Weldon and Bellinger (1997) found that groups collectively remembered random

details better than did individual members, but the reverse was true for organized
narratives. This echoes the “two modes of thought” concept attributed to Bruner

(1986) above and suggests that groups may engage in more bottom-up problem

solving, but that individuals might prefer top-down or narrative approaches. This

would be another example of how group composition or size might affect the

process of learning.

The Firm as a Cultural and Institutional Entigy:

Most ﬁrms are examples of institutions in the sense that they have a certain
resilience or stability, and an ability to perpetuate themselves (Scott, 2001; p.
48). When viewed from a cultural/cognitive perspective (Scott, 2001; p. 52), the
reﬂective indicators of an institution are the presence of common beliefs and
shared logics of action. The latter term is more relevant to an operations setting,
and stands as something that can be a component of organizational knowledge
as something that is uniquely a property of the organization. Note however the
juxtaposition of common beliefs as an indicator of culture with the previously
cited deﬁnition of knowledge as justiﬁed belief, suggesting a cultural origin or
explanation for organizational knowledge. In a similar vein, Martin (2002, p. 36)

states that culture can supply the underlying “why” rationale that is the foundation

of knowledge.

37

 

 

Fist
sha/
heir
that
indh
eMS‘
cuht
tocx
lnth
rnea
tenn

ofdt

Insh
sock)
dehni
We OI
TBIIOn
WCrki
WDUp
eliior

CUhUn
Denis.

MUkt

A Short digression is required at this point with respect to the use of the word
shared. When Scott (2001) used the term shared logic, it was not obvious that
he intended it to mean universally agreed upon. A more useful interpretation is
that behavior was broadly inﬂuenced by the logic, regardless of whether
individual members understood it, agreed with it, or were even aware of its
existence. Martin (2002) noted a similar problem in deﬁning what it is about a
culture that is shared. She concluded that even members who violently objected
to certain elements of a culture were still, in the end, deﬁned and inﬂuenced by it.
In the work that follows, we will use the term Shared broadly, without intending it
mean anything beyond having some distributed inﬂuence over a group. A better
term might be dominant logic as proposed by Prahalad and Bettis (1986). (End

of digression.)

In sharp contrast to the Cartesian view of knowledge, a stream of research in
sociology (Berger 8'Luckmann, 1966; Gurvitch, 1971; Holzner 8 Marx, 1979)
deﬁnes knowledge as a sociological construct, a product of the culture in which
we operate. This is consistent with Martin’s view of how culture supplies the
rationale that underpins knowledge. We can see practical evidence of this in the
work of Stasser and Titus (1987) and Brown and Duguid (1991) who Show that
group dynamics within a community tend to suppress dissenting views and
enforce a commonality of what is accepted as knowledge. Suggestions of the
cultural component of knowledge in organizations can also be seen in the work of
Denison and MiShra (1995), Miller (1996), Davenport and Prusak (1998),

Mukherjee, Lapre et al. (1998), Alavi and Leidner (2001). Implicit in a cultural

38

vie'
knr
me

8C

Th

int

cu
wil

tor

of t
nar
The
Che

obs

 

view of knowledge is the idea that Ieaming, in the sense of a change in
knowledge state, must at some level require a change of culture. While
mechanisms of culture change are poorly understood, it is generally

acknowledged to be a difﬁcult thing to bring about (Martin, 2002, p. 347).

The introduction of a cultural dimension raises issues of methodology and
interpretation, summarized by the etic/emic divide (Martin, 2002, p. 36). The
distinction in this divide is that the researcher’s point of view is either within the
culture, or outside of it. Each point of view will fail to see things that the other
will. Most typically, the etic (outside) view will fail to see the signiﬁcance attached
to various symbols or actions, while the emic (insider) view will fail to see that
certain taken-for-granted truths are not universal. In addressing the shortcomings
of the etic view, Bartel and Garud (2003) note that: “to see and understand
narratives, the researcher must become semi-native” (emphasis in the original).
The methodological issues raised by this distinction will be discussed in the
Chapter Three, but the result is the selection of a participant-observer model of

observation (Martin, 2002 pp 48, 210) as a way to bridge that gap.

One of the objectives of this discussion has been a well-grounded deﬁnition of
organizational knowledge. In Chapter One it was Shown that there is a body of
opinion that holds that whatever organizational knowledge is, it can be found
embedded in an organization in the form of routines. Routines will be discussed
in more detail below, under the heading of the ﬂrrn as an economic entity, but

there are cultural aspects of the ﬁrm that apply to routines as well.

39

in t
F38!
set
as;
05%
“Sh;
abo
wha

\Nha

Sc01
ibrc
tenn
Take
orga;
Cons
Suggs
bEyor
inﬁnu
"donnr
2003)
Tosu
jmDo

SUch

In their research on organizational routines Feldman and Pentland (2005) and
Pentland and Feldman (2005) found that, to be understood, routines must be
separated into their performative and ostensive components. The performative
aspect is the familiar, observable operation of the routine in practice. The
ostensive portion on the other hand, is the underlying rationale, equivalent to the
“shared logic of action” as it is described by Scott (2001) (and subject to the
above note about the usage of shared). Recognition of this ostensive aspect is
what permits the observer to understand otherwise unpredictable variations in

what seems to be tightly-deﬁned process.

Scott (2001) notes that in the cultural/cognitive view of the institution, the basis
for compliance with the logic of action (equivalently: the ostensive component as
termed by Pentland & Feldman, 2005) may be a “shared understanding” or
“taken-for-grantedness” (p. 52). This strongly suggests that routines and
organizational knowledge in general will have an important cultural component.

‘ Consistent with the above discussion on the limitations of the emic view, this also
suggests that a shared logic that is based on taken-for-grantedness may be
beyond the reach of question or debate. This may be one of the reasons that
institutions are “resilient” as noted by Denison and Mishra (1995), or that
“dominant logics are extraordinarily resistant to unlearning” (Bettis & Wong,

2003)

To summarize the above discussion: an underlying rationale or shared logic is an
important, and possibly necessary, component of organizational knowledge. As

such, a shared logic becomes an important concept in the development of a

40

ﬂ1e<
logi
cha

cha

The
the l
‘wth

onga

knov

 

theoretical framework for the study of organizational learning. Such a shared
logic is potentially an indicator of a culture or institution. If learning involves
changes to the logic, the cultural and institutional forces that protect it from

challenges must also be understood.

The Firm as an Economic Entity:

The preceding sections have examined the role of knowledge, its interaction with
the cognitive functioning of the individual, and the interaction of the individual
with the culture of an organization. In this section we close the loop from the
organization (now seen primarily as an economic entity — typically a ﬁrm) back to

knowledge.

The ﬁrm as a unit of analysis is central to the literatures of economics and
strategy. From the beginning, knowledge, or know-how of some sort has been
considered as an important property of the ﬁrm. The strategy literature, rooted in
economics, has sought to explain the boundaries of the ﬁrm on the basis of
transaction and coordination costs (Coase, 1937). Penrose (1959) (discussed in
Rugman & Verbeke, 2002) argued that through the process of Ieaming, ﬁrms
could reduce their coordination costs and expand those boundaries. March and
Simon (1958) recognized that this knowledge described by Penrose (1959) was
something speciﬁc to a ﬁrm, and introduced the concept of performance
programs as the form in which it was embedded in the ﬁrm. These are what we
now more commonly refer to as routines. Nelson and Winter (1982) connected

routines with transaction cost economics to suggest that routines were the

41

mel

hﬁe

Thu
Bas
pan
unm
indh
foreg
orni

gene

chk
ﬁnni
Bani
1993
Nona
Dave
isso
olﬁm
bUiol
deum
RBV I

Eben

 

 

mechanisms by which an organization could efﬁciently manage transactions

internally and hence grow.

This notion of ﬁrm-speciﬁc resources is critical to the emergence of the Resource
Based View (RBV) of the ﬁrm in the strategy literature. Knowledge is of
particular importance to this view, at least in its economic aspects, because it is a
unique class of asset. It is imperfectly tradable (Williamson, 1981), and
indivisible in the sense that using it for one purpose does not require the ﬁrm to
forego its use for another (Teece, 1980). Thus, the use of knowledge incurs little
or no opportunity cost, making it one of the few, if not the only resource that can

generate true rents (Mahoney & Pandian, 1992).

If this is true, it is logical to argue that knowledge is the most important asset of a
ﬁrm in terms of creating competitive advantage (Prahalad & Hamel, 1990;
Barney, 1991; Nelson, 1991; Leonard-Barton, 1992; Cyert, Kumar & Williams,
1993; Henderson & Cockburn, 1994; Nonaka, 1994; Bates & Flynn, 1995;
Nonaka & Takeuchi, 1995; Kogut & Zander, 1996; Miller, 1996; Spender, 1996;
Davenport & Prusak, 1998; Nahapiet & Ghoshal, 1998). The role of knowledge
is so central to strategy, that Grant (1996) proposed the Knowledge-Based View
of the ﬁrm (KBV) but in doing so treated knowledge not as a property of the ﬁrm
but of the individuals comprising the ﬁrm. At the same time, he expressly
declined to offer any speciﬁc deﬁnition of knowledge, so in keeping with most
RBV literature, knowledge as a resource remains a highly abstract concept,

albeit a vital one.

42

lRou
rout
leas
soli<
rout
rna<
exp
sew
Aﬂor
have
h~o
pert:
rouﬁ
unot
unde
DEﬁo
Obser
Orga
deter
leev
come
aga

DOir

Routines were mentioned above, with little elaboration. The concept of the
routine is generally attributed to March and Simon (1958), but can be traced at
least to Barnard (1938). Although many researchers have done a great deal to
solidify the concept of the routine, there has been relatively little said about what
routines consist of and how they are created or changed. Progress has been
made on the anatomy of routines by Feldman and Pentland (2003) who
expanded the basic concept of routines to include the potential for variation and
selection, and therefore evolutionary, incremental change.

More recently, Feldman and Pentland (2005) and Pentland and Feldman (2005)
have argued that to be understood, routines must be considered as consisting of
two layers; the ostensive and the performative (plus associated artifacts). The
performative layer is the observable sequence of actions that constitute the
routine as it is practiced in the moment. The ostensive part is the relatively
unobservable underlying purpose or intent of the routine. The need to
understand the ostensive layer was motivated by their observation that the
performative layer tended to vary spontaneously and unpredictably (to an outside
observer) in response to changing conditions. If the routine is the embodiment of
organizational knowledge, the ostensive part is at least as important a
determinant of situated performance as the performative part.

Likewise, if performative and ostensive components are both essential
components of organizational knowledge, then beyond some point,
organizational learning must involve changes to both parts. This is precisely the

point made by Argyris and Schbn (1978) when they coined the term “double loop

43

lea
cor
“the
the
Wh
repl
cult
pert
sire;
emp
DOW!
Schr
hint 5
Using
statin

orche

 

Ieaming” to describe the case where the ostensive part is examined and re-
considered. They did not use the term ostensive; instead they spoke of the
“theory in use" that provided the belief that a certain set of actions would result in
the desired outcome.

Whether we speak of ostensive aspect or theory in use, it is clear that these
represent the same thing as a shared logic of action (Scott, 2001) in the
cultural/cognitive view of the institution. As suggested in the introduction, the
performative and ostensive views tend to fall into different, non-overlapping
streams of literature. The result in Operations Management has been an
emphasis on performative aspects (“tools”), with a resulting lack of explanatory
power. To cite a well-regarded paper as an example, Flynn, Sakakibara and
Schroeder (1995) focus extensively on performative aspects, with results that
hint strongly at the importance of an (unidentiﬁed) underlying ostensive aspect.
Using different terminology, Miller (1996) showed a recognition of this gap by
stating that competitive advantage lies “not in speciﬁc resources or skills, but in

orchestrating themes.”

Summau of Key Concepts and Constructs:

The purpose of Part I of this chapter was to establish a list of key concepts and
constructs that is grounded in the literature and relevant to the research question.
To ensure a sufﬁciently diverse, yet coherent approach to this task, the literature
was examined from three different perspectives corresponding to the key

elements in the organizational Ieaming process. These perspectives are those of

44

theinl
whhht
Hwas
andtl
shear
obser
Opera
invohr
eshﬂﬂ

unden

These
posse:

observ

C)

(.)

(l)

C J
\

C)

 

C)

()

(f)

the individual involved, the knowledge that is processed and the organization

within which the learning takes place.

It was seen that each of these points of view led to different streams of literature
and the result is a blended list of concepts and constructs drawing on the multiple
streams. The blending of different points of view is consistent with the
observation earlier in this chapter that conﬁning the literature search to
Operations Management did not provide adequate explanation of the process
involved in organizational learning. By drawing on a wider base of research to
establish the research framework, the goal of this research is to arrive at a better

understanding of the processes involved.

These concepts do not qualify as constructs in the narrow empirical sense of
possessing validated measurement scales; rather they are areas within which
observations are indicated on a theoretical basis. The summary list is:

0 Knowledge classiﬁed as being explicit or tacit

o Transformations of knowledge within and between explicit and tacit

0 Knowledge classiﬁed by location

0 Actions by individuals to retrieve or store knowledge by location

0 Processing of knowledge and/or information by individuals (narrative

versus paradigmatic)

0 Individual cognition as in pre-existing state of knowledge

0 Power and/or knowledge as moderators of small group activity

0 Shared logics of action (existence of and changes to)

0 Changes to performative aspects of routines

45

Thee

as de

 

 

0 Changes in situated performance

These concepts will be used in developing the research framework and protocol

as described in Chapter Three.

46

The

Myer
tram
gaps
draw.
tMoa

aﬁec

ilcor
resez
forth
toun:
shout
byHu
issues
inano

weib

PART II: POSITIONING OF THE RESEARCH:

The bounds within which this research was conceived and conducted should be
re-emphasized. It was stated in Chapter One that the research question was
framed in the context of Operations Management. As a result, the problems or
gaps in understanding that make the question interesting and important were
drawn from this ﬁeld. While it is hoped that the ﬁndings may ultimately be of
broader interest, the focus on Operations dictates the research design and

affects the positioning of the work within the literature.

A consequence of this focus is an admittedly functionalist perspective to the
research (Bacharach, 1989; Martin, 2002; p. 160). This means that the reason
for the research question is the belief or intent that the answer should be relevant
to understanding and explaining performance in operations. Note that this
should not be confused with the functionalist model of explanation as discussed
by Hunt (1991). The functionalist perspective serves as a screen to separate
issues of interest from those that are not. Although they may be valid concerns
in another context, we are not, for example, interested in the personal growth or

well-being of participants in the organizational learning process.

This approach also deﬁnes the stream(s) of literature in whose footsteps it is
intended to follow. There are three interrelated streams of literature that deal
explicitly or implicitly with the role of learning in the improvement of ﬁrm

performance.

47

Chron
addre'
has w
consn
unenc
(nose
and u

Swnn

A mo
capar
capac
exten
proce

Dredic

 

Chronologically, the ﬁrst of these streams deals with Ieaming curves and
addresses the aggregate effect of organizational learning. Interest in this area
has waned somewhat in the last decade, and it does not appear that this is
considered to be an area for cutting edge research. As the discussion below is
intended to show, this may well be a result of a lack of understanding of the
processes involved. Although learning curves have ample empirical veriﬁcation,
and useful predictive ability, they are lacking in explanatory power (Schmenner &

Swink, 1998).

A more recent stream of literature is centered on the concept of absorptive
capacity as introduced by Cohen and Levinthal (1990). As such it deals with the
capacity for organizational learning and offers a potentially predictive tool. While
extensions of this work have broken down the concept into what could be called
process steps, this concept is still more outcome than process focused and more

predictive than explanatory.

The third stream of literature springs from the researcher’s problem noted in
Chapter One. Speciﬁcally, the problem is: “why do ﬁrms differ in their ability to
successfully implement speciﬁc programs such as Total Quality Management
(TQM)?” When viewed from a content perspective, ﬁrms implementing a
standardized methodology appear to be homogeneous and yet there is
substantial variance in their success. The search for an explanation has
generated a body of research which lacks a convenient label. Much of this
literature is to be found under the general heading of quality, since that has been

a focus of many of the initiatives. (Note that other ﬁelds have their own areas of

48

inlet
inter
Ente

Lu(

For
Hetr

BTEE

SUp

rela

Ove

Iear

interest. For example, researchers in information systems struggle to explain
inter-ﬁrm differences in success of implementation of system initiatives such as
Enterprise Resource Planning (ERP); for examples see: Guimaraes, lgbaria and

Lu (1992), Sum, Ang and Yeo (1997), Petroni (2002), Marble (2003).

For the purposes of classiﬁcation, we will refer to this body of work as “Explaining
Heterogeneity in Operations.” Two distinct camps have emerged within this
area. One seeks to explain the heterogeneity through tradeoffs that result from
the choice of competitive priorities, while the other offers the cumulative or
sequential nature of capabilities as an explanation. The review below suggests
that a process perspective has been lacking in this ﬁeld and that empirical
support for any particular explanation is mixed at best. These facts may be

related.

Overall, it is proposed that greater insight into the process of organizational

learning has the potential to contribute to each of these ﬁelds.

Learning Curves:

In a learning curve view of the situation, the ﬁrm or business unit is the unit of
analysis and the dependent variable is some sort of performance measure,
typically a productivity measure such as direct labor cost per unit of production.
The independent variable is some measure of experience, with cumulative
production volume being the most widely accepted. While the idea that
individuals exhibit Ieaming curves has been known at least since the 19th century

(Argote, 1999, p. 4), the seminal work establishing the existence of learning

49

 

 

CU
pTC

Su

sor
pro
QEr
anc
eXp
Bec
pos

Ope

50m
are

CIeE

curves in an organizational context was that of Wright (1936), a study of airframe

production.

Subsequent studies have thoroughly conﬁrmed the existence of learning curves
in a wide range of organizational activities. Hatch and Mowery (1998) cite at
least seventeen examples in the literature. The classical form always shows
continuing improvement of performance with experience, but with declining
returns to scale. Much research has sought to establish empirically whether the
classical shape of the learning curve is indeed a power function as opposed to
some other form, and whether the correct independent variable is cumulative
production as opposed to time. The answer to both questions has been
generally afﬁrmative (see Lieberman, 1984; Darr, Argote & Epple, 1995). Slight
anomalies in the functional form have been observed, and the range of
experimental error allows for the possibility of other interpretations. Argote,
Beckman et al. (1990) found both cumulative volume and time to be useful when
postulating that knowledge is perishable and that there is a “forgetting” curve that

operates in conjunction with the learning curve.

Having empirically established cumulative production volume as the best ﬁt for
an independent variable, it is tempting to infer a causal relationship. Huber
(1991) speaks of “the positive effect of experience on performance” which
sounds like a statement of causality. Other writers (Argote, 1999, for example)
are more careful to disavow statements of causality, but the inference tends to

creep back into the discussion.

50

pl

ll’ll

au
dis
Mo
cha
the
to ti
on e
Klep
cont.

basi:

The work discussed to this point has been very much content based. Some
progress towards a process view has been made by separate the learning curve
into separate components in order to better understand the mechanisms
involved. Levy (1965) considered the Ieaming curve to be made up of three
components: planned or induced learning, random or exogenous learning and
autonomous learning. The second category has not been much studied, but the
distinction between autonomous and induced learning has persisted (Hatch &
Mowery, 1998; Li & Rajagopalan, 1998; Zangwill & Kantor, 1998). Levy (1965)
characterized the induced learning component as a form of investment, made on
the basis of an expected return, thereby introducing a forward-looking component
to the learning curve. (See also Oi, 1967). The idea that learning is based more
on expectations than experience was also suggested more recently by Sinclair,
Klepper et al. (2000). Interestingly, from the perspective of the process versus
content positioning, this last work is the only one to describe process on the

basis of actually observing people at work.

In an attempt to explain the autonomous component, Fine (1988) introduced the
idea of “quality-based learning,” which is to say, learning from mistakes. In this
view, the ﬂattening of the learning curve would be explained by the reduced
frequency of occurrence of mistakes as their root causes are progressively

eliminated.

Other writers have also made the case that total learning effects are made up of
distinct components. Another partitioning of the curve was proposed by Argote

(1999), who characterized learning as consisting of improvements in: (a)

51

nc
be

e m

Bet
hav
as a
addr
proc.
(198r
ﬁnite
netwc
Effect.
shape

but off

It is We
3' Thon
is not
betwe
beiWe,
is still

’93rna’r

individual proﬁciency, (b) improvements in technology and (c) improvements in
organizational routines. There is some overlapping of terms, so it is important
not to mix typologies. For example, improvements in individual proﬁciency could
be the result of planned training (induced Ieaming) or of self-motivated trial and

error (autonomous Ieaming).

Because empirical tests of learning result in a consistent pattern of results, there
have been attempts to develop models that would come up with the same results
as a way to understand the mechanisms involved. Note that while this literature
addresses process issues it is not process-based. It is attempting to infer
process from an examination of content. The most cited is probably that of Muth
(1986), who modeled learning as a random search for better alternatives from a
ﬁnite set. Huberman (2001) also replicated a power function by starting with a
network model. In this formulation, the learning rate is determined by the
effectiveness in ﬁnding new paths. Both of these models predict the observed
shape of learning curves based on variations of sampling without replacement,

but offer little insight about why they vary or why they occur at all.

It is. well known that Ieaming curves do vary substantially between ﬁrms (Dutton
& Thomas, 1984; Hayes & Clark, 1986), but is also clear that the “learning rate”
is not a ﬁxed property of a ﬁrm. Argote and Epple (1990) found that it can vary
between plants within the same ﬁrm, and Adler and Clark (1991) found it to vary
between departments in the same plant. These observations suggest that there
is still a lack of understanding about the fundamental nature of organizational

learning.

52

In f2
obsr
“exp

revie

Of tr
TOUl'll
take

that\

woul:
ﬁlSl ir
anon
h0n1(
Venke
must I
plOCeg
opEra:
COhEr
COnCei

apply1

In fact, this was the principal point made by Adler and Clark (1991). They
observed that the process of learning was relatively unstudied and offered their
“exploratory” work as encouragement to others to rectify this shortcoming. A
review of the 64 papers citing Adler and Clark (1991) suggests that their advice
has been little heeded, with only a handful concerning themselves with process.
Of those that do, only two (MacDufﬁe, 1997; Sinclair, Klepper et al., 2000) were
found to employ actual observation of the process. This research proposes to
take a process perspective based on direct observation in order to begin ﬁlling

that void.

Absorptive Capacity:

There is a related stream of literature that deals with the’ﬁnn’s capacity to learn.
It seems reasonable that a better understanding of how ﬁrms actually do learn
would also help explain their capacity to learn. The term absorptive capacity was
ﬁrst introduced by Cohen and Levinthal (1990). This is deﬁned as the ability of
an organization to assimilate and use knowledge (often technical knowledge)
from outside its boundaries. This concept has proved to be popular, but it is what
Venkatraman (1989) would call a criterion-speciﬁc construct, meaning that it
must be deﬁned in terms of its outcomes. As such, it is a content view, not a
process view. One result of this is the observation that “deﬁnitions and

operationalizations of this construct vary widely” (Zahra & George, 2002; p. 186).

Cohen and Levinthal (1990) attempted to build a criterion-free base for the
concept by attributing to the organization the same cognitive processes that

apply to an individual. Following this cognitive model, they argue that an

53

orga
IUdhl
leve
DIOF
expe

ques

The
and
Szul
abili‘
mod
case
Wevv
intro.
le-cc
Conle
Resc
they

also

 

organization’s ability to absorb new knowledge from external sources is, like an
individual’s ability to learn new things, shaped by the pre-existing cognition or
level of knowledge. Speciﬁcally, Cohen and Levinthal (1990) state the
proposition that absorptive capacity is a function of the ﬁrm’s related prior
experience. The validity of the analogy to individual cognition is open to

question, but it has been adopted by others (Cohen, 1991).

The concept of absorptive capacity has been developed in multiple directions
and applied in various settings. It was applied internally to the organization by
Szulanski (1996) who coined the term “internal stickiness. ” This refers to the
ability, or rather the lack of it, to transmit knowledge internally. Here the cognitive
model of learning behavior is on ﬁrmer ground; since the recipient is in most
cases an individual, although this is not a requirement of the model. A related
view of the internal Ieaming process is that of Garud and Nayyar (1994), who
introduced the term transfonnative capacity. This is the ability to resurrect and
re-combine existing knowledge within the organization and apply it in a changed
context, which the authors position as a resource within the context of the
Resource Based View of the ﬁrm (RBV). As with Cohen and Levinthal (1990),
they offer no operational deﬁnition of knowledge, so transfon'native capacity must

also remain a somewhat abstract concept.

To make absorptive capacity more concrete, Zahra and George (2002)
decomposed the absorption function into four elements (acquisition, assimilation,
transformation and exploitation). This has the advantage of deﬁning sub-

constructs that are less abstract and facilitates the collection of data for empirical

 

res
be

knc

To

bet
inte
imp
dee
pro.
mar

6X8

To:
lOp

the l

Ex

research but does not fully address the issue of explanation. In addition, as will
be seen below, this view tends to force a narrow view of what constitutes

knowledge.

To summarize a portion of their work, Zahra and George (2002) distinguish
between what absorptive capacity should be, based on a set of predictors and
what it actually is, based on outcome measures. The gap is identiﬁed by an
efﬁciency factor n, which stands as an indicator of the process differences
between ﬁrms. Zahra and George (2002) attribute the variance in n to “social
integration mechanisms” of the different ﬁrms. This clearly acknowledges the
importance of the process of organizational learning, but without offering any
deeper analysis. In a similar vein, Zahra and Nielsen (2002) note that “the
process by which (the integration of extemal and internal sources of
manufacturing capability) occur within technology commercialization should be

examined. ”

To conclude, absorptive capacity in its many conceptualizations offers an ability
to predict outcomes of the organizational learning process, but provides little in

the way of explanation of how and why it takes place.

Explaining Heterogeneity in Operations:

As stated in Chapter One, an issue that continues to preoccupy researchers in
Operations Management is the question of why ﬁrms differ in their ability to
implement certain initiatives. In recognition of its importance to operations, a
great deal of this literature has dealt with quality, but the question is equally

applicable to other ﬁelds. The principal conundrum is this: when best practice is

55

 

rigo
mel

C88

In t
per
ber
Lai
ber
80:
Gre

{88¢

orig
mm.
“W
eXpl;
i”vol
the 3
Sub]
arla:
Reg.

WC»:

rigorously codiﬁed, as in TQM, then ﬁrms implementing the standardized
methodology should experience very little variance in results, yet this is not the

case.

In the area of quality, it is clear that there is substantial variance. A high
percentage of ﬁrms that implement quality programs fail to realize the expected
beneﬁts from doing so (Macdonald, 1996; Choi & Behling, 1997; Krumwiede &
Lavelle, 2000; Nwabueze, 2001). More interestingly, ﬁrms that do realize
beneﬁts often ﬁnd them temporary and revert to their pre-program status (Dale,
Boaden, Wilcox & McQuater, 1997; Buch & Rivers, 2001; Bullington, Easley,
Greenwood & Bullington, 2002). The search for an explanation has led

researchers in different directions.

A popular line of research has been that into cumulative capabilities. The idea
originated with Ferdows, De Meyer, Nakane, Miller and Vollman (1985), and was
more formally proposed as a theory by Ferdows and De Meyer (1990). The key
hypothesis is that it is the order in which certain initiatives are implemented that
explains variance in success. This approach says nothing about the process
involved and deals only with ﬁrm-level outcome measures. The central issue of
the sequence of actions creates a problem that is both difﬁcult to study and
subject to many confounding inﬂuences in the course of any longitudinal
analysis. As a result, empirical support for the theory has been equivocal.
Recently, Flynn and Flynn (2004) concluded that “support for sequential

progression of cumulative capabilities was not evident...”

56

Fly
vie

de

inf
for
co
be
an

org

Flynn, Sakakibara et al. (1995) and Powell (1995) took a related but differing
view; both suggesting that it is not necessarily the order that matters, but the
degree to which different capabilities are built on common “infrastructure” or
intangible factors. In the former work, the observed predictive value of
infrastructure elements such as managerial support is interesting, but cries out
for an explanation of how and why it came to be. In the latter paper, the
conclusion was that TQM could represent a source of competitive advantage
because the intangible factors that made it successful were poorly understood
and hard to imitate. If that is true, it is a clear call for research into the nature of

organizational knowledge and the process by which it can be created.

The body of literature that cumulative capabilities set out to displace was one that
argued that tradeoffs between outcomes were inevitable, and that the choice of
how to make them was a matter of choice of competitive priorities. In Operations
Management, this view is generally credited to Skinner (1966). More recently,
the case for the existence of tradeoffs has been strongly argued by Clark (1996),

Schmenner and Swink (1998) and Boyer and Lewis (2002).

In this view of the world, the apparent absence of tradeoffs is just a different
positioning of the tradeoff frontier. Clark (1996) refers to this as the performance
frontier, and speaks in detail of the role of Ieaming in moving it (p. 48). More
tellingly, Schmenner and Swink (1998) refer to the asset frontier as the limiting
case - the best available performance frontier achievable in an existing situation.
Having argued in Chapter One that organizational knowledge is a major, if not

the largest, asset of the ﬁrm, it is no great stretch to argue that the positioning of

57

the

lea

Ta
cur
tim
UN
is I
res

fosr

Thi:
follc
that
ider
mig?
esta
grot
The
Can

“Elev

for b.

the frontier will have a great deal to do with organizational knowledge and

Ieaming.

Taking an integrative view, Schmenner and Swink (1998) note that tradeoffs and
cumulative capabilities are not necessarily in conﬂict when one introduces the
time dimension to the picture. The time dimension highlights the critical need for
understanding the organizational learning process. To quote Clark (1996): “little
is known about paths for improvement and learning” and: “little systematic
research has been done on. . ..the organizational and managerial processes that
foster Ieaming. ”

Chapter Summagy:

This chapter set out to accomplish three things: a) to position this research as '
following in the stream of literature focused on process improvement, b) to show
that there is a shortage of process-based literature in this stream, and c) to
identify, from multiple streams of literature, the key concepts or constructs that
might be expected to fruitful for study. An additional outcome of point 0) is the
establishment of deﬁnitions of organizational knowledge and learning that are
grounded in the literature.

The next steps are to develop a research framework within which hypotheses
can be stated, and to develop a research protocol that ensures that observations
relevant to the framework are captured. The above list of concepts will be used

for both of these purposes in Chapter Three.

58

.0.

Introd'

’

As pre
“how c
review
under:
in supi
buildin
choice

theory

This w
within

import.
from ll
devote
Observ
hede
threat

SCien.

An in,
@563

The

CHAPTER THREE: RESEARCH FRAMEWORK 8: METHODOLOGY:

Introduction:

As previously noted, this study strives to answer a simple but critical question:
“how do organizations create knowledge in their operations?” The literature
review showed that there have been repeated calls for more research and better
understanding of the process of organizational learning, suggesting a deﬁciency
in supporting theory. Ultimately, this research must address itself to theory-
building. The manner in which this is to be done is a matter of philosophy and
choice. This chapter will begin with a discussion of the philosophical approach to

theory-building that will be taken in this research.

This will be followed by the deﬁnition and development of a research framework
within which potential theories may be stated. In the context of this chapter, it is
important to point out that this framework is derived from the literature, and not
from the observations of this study. The balance of the chapter will then be
devoted to describing the research design and methodology by which ﬁeld
observations are collected and transformed into elements of theory. Throughout
the development of the research design, the central them is the avoidance of
threats to validity, which must always be a concern, but particularly in the social

sciences.

An important part of the research design is the selection of the appropriate unit of
analysis. Since this selection is referenced throughout the discussion of the
research protocol, a brief deﬁnition is offered here to facilitate that discussion.

The unit of analysis ultimately selected is characterized as an event, reﬂecting

59

 

the c

 

the comment of Rosch (1978) that “The unit of the event would appear to be a
particularly important unit for analysis. Events stand at the interface between an
analysis of social structure and culture and an analysis of individual psychology.
It may be useful to think of scripts for events as the level of theory at which we
can specify how culture and social structure enter the individual mind.” For our
purposes, events will consist of a deﬁned set of individuals acting in a bounded

setting to effect a veriﬁable improvement in a production process.
B_e_§earch Goal. Theory-Building:

The present research is directed towards theory-building, placing it in what Hunt
(1991) calls the context of discovery. Within that context Hunt (1991) offers two
choices for purposeful and systematic efforts at discovery. These are the

deductive and inductive methods.

In the deductive mode, the researcher works from the general to the speciﬁc. A
relationship is proposed that is assumed toapply generally. The consequences
of that relationship in a particular situation are then deduced to arrive at a
prediction (hypothesis) that can be tested. The inductive researcher works in the
opposite direction. Speciﬁc observations are collected, classiﬁed and analyzed
to yield a consistent relationship can be discerned that is believed to apply with

some generality.

Within the model of theory proposed by Bacharach ( 1989) (see Figure 2, derived
from his Figure 1, page 499), the difference would be stated this way: the
deductive approach states propositions in order to derive hypotheses that can be

tested. The inductive approach observes outcomes that would verify potential

60

 

 

hypc
purse

inadr

Vl

 

 

hypotheses in order to arrive at more general propositions. This research
pursues an inductive approach to theory-building, driven largely by the perceived

inadequacy of existing theory in this area (see Chapter Two).

 

 

 

 

 

 

 

 

 

 

 

 

PROPOSITIONS
CONSTRUCTS k , - CONSTRUCTS
t 5 t
j i l
VARIABLES - ' - VARIABLES
HYPOTHESES

 

 

 

 

 

 

Figure 2: Model of a Theory

The inductive approach brings with it three inherent difﬁculties that can never be
overcome entirely. The goal of the research design must be to minimize the
threats that these difﬁculties pose to the validity of the conclusions. The ﬁrst
problem, noted by Hunt (1991), Eisenhardt (1989) and Ellram (1996) is that when
one begins with observations, it is not possible to begin with a clean sheet of
paper. It is necessary to establish some a priori hypotheses about what will be
interesting to observe and what will not. This can, of course, bias the results.
The best defense against this sort of bias is a good research protocol, grounded
in the literature (Ellram, 1996). Such a protocol is developed in this chapter,

derived from the results of Chapter Two.

61

Th
hyr
ch:
an.
as
pro

are

Wh
the
the
real
advi
Usel
cont
from

Cilcu

The second difﬁculty lies in the method by which observations (potential
hypotheses) are distilled into propositions. Again, Hunt (1991) offers two
choices, neither of them entirely satisfactory. If the observations are quantiﬁable,
and enough data can be collected, then propositions can be derived statistically,
as in exploratory factor analysis. This is a somewhat weak method, since
propositions derived in this way cannot be falsiﬁed; one can only say that they

are supported or not within certain conﬁdence limits.

When observations are not quantiﬁable or are few in number, both of which are
the case in this research, the remaining alternative is pattern matching, which is
the search for distinctive and recurring patterns in the data (Hunt, 1991). The
realities of the social sciences dictate frequent use of the method, and it is
advocated in this context by Kaplan (1964). While this method can generate
useful insight, Hunt (1991) points out that it suffers from a lack of intersubjective
conﬁrrnability. By this he means that if two researchers extract different patterns
from the same data, there is no test to determine which is better. Under these
circumstances, the best that one can hope for is a degree of consensus. This
research seeks to address this problem through the use of a panel of experts
who were asked to comment on the validity of some of the key patterns

observed.

The third difﬁculty with the inductive approach is to make the resulting
propositions testable. This is typically done (using the model of Bacharach,
1989) by deriving testable hypotheses from the propositions. In the inductive

approach however, this involves reversing the process and should return

62

hypotr
to be 5
same

testab
setting
Having
resear

preser

 

 

hypotheses that mirror the original observations (minus those that were deemed
to be spurious). It is thus improper to both derive and test propositions within the
same research setting. This study will end with propositions; their reduction to
testable hypotheses must remain a matter for future research in different

settings.

Having discussed the research goal at some length, the development of the
research design and methodology will proceed according to the ﬂowchart

presented in Figure 3.

63

Res

[:7

I]

Fill—£7

Verr‘

[THC—l

resear‘
The c,

ftrio/:2,

that t

 

 

 

Research Question

 

 

Literature Review

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(Chapter 2)
Research Goal Theoretical
, Framework
Research Method Protocol
Experimental
Design
Entering the Field
Data Collection
Veriﬁcation of Data
Data Analysis
Conclusions Validation of
(Chapter 5) Conclusions

 

 

 

 

 

Figure 3: Research Methodology

Research Framework:

The term framework is used liberally but somewhat loosely in social science
research, so it is appropriate to deﬁne what it means here and why we need one.
The descriptive deﬁnition of a framework is that it is something more than a
typology and something less than a theory. A typology is a classiﬁcation system

that takes a ﬁeld of relevance and divides into mutually exhaustive and exclusive

64

 

 

 

 

ele
ent
sar
abr
are
nol
lite
rela

oft

All

this

Pro

 

elements (Chrisman, Hofer & Boulton, 1988). These elements may be natural
entities or less tangible concepts often referred to as constructs. (Note that the
same deﬁnition may be applied to a taxonomy; there is considerable debate
about the distinction between these two terms relating primarily to whether they
are empirically or logically derived. We will bypass that debate here and just
note that the elements of our framework are derived from a synthesis of the
literature.) A theory goes beyond a typology by specifying propositions about the
relationships between the entities and constructs in a way that allows the stating

of testable hypotheses.

A framework, then, lies somewhere between these states. Like a typology, it
must divide the ﬁeld of interest into elements, but in addition, that division must
be somehow relevant to how the elements are expected to relate to each other.
It is not necessary at this point to state formal hypotheses about the
relationships, but the classiﬁcation must be such that hypotheses could be stated
in a subject-object fashion. Two examples of the term framework being used in

this way in Operations Management are Jensen (1992) and Kim and Lee (1993).

Properly constructed, the framework serves as a roadmap for the research.
While grounded theory development (Glaser & Strauss, 1967) may be an ideal
state, the process of discovery is greatly facilitated by having some idea of where
to look. Both Eisenhardt (1989) and Yin (1994) note that it is pointless to collect
data without some organizing framework or theoretical lens that determines what
will be recorded and what will be ignored. This suggests an important role for the

framework in the construction of a research protocol. Many kinds of ﬁeldwork

65

den
and
inte
rnal
eve
sigr
The
Wh:
ouu
‘wen
wou
ques
exan
ans“
INoce
tithe
”Mshe

0f Org-

Doss/fJ

demand the use of a protocol to ensure that all of the relevant points are covered
and that each situation is approached the same way. For example, in an
interview situation, a protocol might be a pre-determined set of questions that
make sure the key points are addressed. In this research, it took the form of an
event checklist (was there evidence of a certain condition, and what was its

signiﬁcance?) (Ellram, 1996).

The list at the end of Chapter Two is the basis for the framework in this research.
What makes this list more than a typology is the way it was constructed. At the
outset, attention was given to the principal entities in the research setting. These
were the individual, knowledge and the ﬁrm. It was presupposed that these
would interact with each other in some way that would address the research
question. The framework was then populated from the literature by asking, for
example, “What does it mean for an individual to interact with knowledge?” The
answer to that question led to the inclusion of cognition and information
processing as important elements of the framework. Rather than enumerate all
of the elements of the framework in this fashion, the range of interactions is
illustrated graphically in Figure 4. It can be seen that this is not by itself a theory
of organizational knowledge and learning, but it lays out the elements and

possible relationships from which such a theory could be expressed.

66

 

     
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

KNOWLEDGE Cognition, .
(Tacit/Explicit) Modes of Thought 'ND'V'DUAL
A
l
Creating/Changing Group Level
Routines ’, ......... \ Interactions
:’ ORGANIZATIONAL ‘~.

I ‘.\ KNOWLEDGE ,' I
{’SHARED LOGIC“
ROUT'NES OF ACTION

Situated Taken-for-
Perfonnance Grantedness

 

 

 

 

 

 

 
        
   

  

Cultural
Entity Entity

ORGANIZATION

Figure 4: Schematic of Research Framework

 

Economic

 

 

Research Method:

The choice of research method is governed by the research question, the goal of
the research and the elements chosen for the framework. The focus on theory-
building and on process versus content more or less dictates the need for ﬁeld
work. The use of archival data is inconsistent with the process focus of the
research, since it tends to be almost exclusively content centered. Surveys or
interviews were considered but looking at the elements of the framework, it
became clear that it was not feasible to construct a survey instrument or an
interview protocol that would generate the depth of insight required. For

elements such as a shared logic of action, which may involve compliance on the

67

base;

loran

Asan
the se
reconi
queshr
innah.

consul
Resea

Inther
haten
of the r
andso
DIODerIj
Dﬁtocc

In this r
(and sh
theprot
IOoks s}

beenid

 

 

DIECIUQ‘

basis of taken-for-grantedness (Scott, 2001, p. 52), it may be literally impossible

for an insider to recognize its existence (Martin, 2002).

As a result of the above, the decision was made to employ direct observation in
the setting of a small number of in—depth case studies. This is consistent with the
recommendation of Yin (1994) that case studies are appropriate for research
questions of a “how" or “why” type. The data from such studies will be qualitative
in nature, which is a strength of the method when multiple alternatives must be

considered (Miles & Huberman, 1994).
Research Protocol:

In the context of research, a protocol simply means a set of rules or procedures
that ensures that the collection of observations is both consistent with the intent
of the research and consistent from case to case. The cases will of course differ,
and so will the data collected, but a protocol ensures that common themes are
properly identiﬁed. If interviews were being used for data collection, an interview
protocol would ensure that a core set of questions was asked of all respondents

in the same way.

In this research, more passive observation is used, so the researcher does not
(and should not) directly inﬂuence what the subjects say and do. In this instance,
the protocol takes the form of an event checklist to ensure that the researcher
looks systematically for evidence of all of the constructs or concepts that have
been identiﬁed in advance as relevant. The existence of a protocol does not

preclude observations of factors not previously identiﬁed, but which may emerge

68

35 TI

bkac

Arpr
ther
rnon
char
ques
dent
vahdz
ltenﬂ
conha

made

lnthe
etenr
senst
i0 nOtI
Ofevr
more f
data CI
apple:

Sighih'

 

 

.. i

as relevant in the context of the event. In fact, this is a fundamental building

block of grounded theory development.

A prior section outlines the framework to be employed in this research and it is
the elements of that framework that constitute the event checklist. However,
more needs to be done. It is necessary to deﬁne in advance the traits or
characteristics that the researcher will use to determine whether the element in
question is present or not. These characteristics are listed in Table 1. Since the
elements of the table must be regarded as tentative constructs, there are no
validated measurement scales for them. This list of traits is synthesized from the
literature that suggested the elements, and is justiﬁed on the basis of face or
content validity (Nunnally, 1994). Where this is not entirely obvious, reference is

made to the source literature.

In this study, the protocol is used both before and after data collection. Before
entering the ﬁeld, the protocol must be understood by the researcher to provide a
sensitization to the traits to be watched for. During observation it is not sufﬁcient
to note that a certain trait was “present,” it must be tied to some particular piece
of evidence. Since a single piece of evidence may reﬂect multiple traits, it proved
more feasible to record a chronology and apply the protocol immediately after the
data collection to tabulate evidence of the key factors. An advantage of this
approach is that it improves the chances of recording observations whose

signiﬁcance may not be apparent until viewed retrospectively.

69

 

la_ble 1: Research Protocol

 

 

 

 

FRAMEWORK
IDENTIFYING TRAJLS
ELEMENT
Knowledge Knowledge is assumed to be explicit unless identiﬁed
classiﬁed as otherwise. Evidence of tacit knowledge: a) Elements of

being explicit or

the work process that people do, but can't tell us exactly

 

tacit how, b) Reasons for the work process that people believe
or accept, but can’t explain exactly why.

Transformations When new knowledge is created (use evidence of

of knowledge improvement in situated performance), the idea of a

within and transformation requires identiﬁcation of a precursor or

between explicit

precursors. Unless the precursor can be broadly classiﬁed

 

 

 

and tacit as knowledge, it is not evidence of a transformation per
Nonaka (1994).

Knowledge Again, using a broad deﬁnition of knowledge, can new

classiﬁed by knowledge or a knowledge source be classiﬁed by location

location (personal memory, written records, physical artifact, etc.)?
Does it make a difference to the process?

Actions by Closely related to the above. What actions are taken that

individuals to result in knowledge being somehow stored in or retrieved

retrieve or store
knowledge by
location

 

from speciﬁc locations?

 

7O

 

 

Iaple 1: continued

 

Processing of
knowledge and/or
information by
individuals
(narrative versus
paradigmatic)

This is the top-down versus bottom-up distinction and may
be quite hard to observe. The key question is whether
individuals engaging in a knowledge creating activity such
as problem solving are rearranging and extending known
facts or whether they are working with incompletely
speciﬁed models or principles that are best explained by
way of a story or narrative.

 

Individual
cognition (as in
pre-existing state
of knowledge)

This is also difﬁcult to observe - look for evidence of
failure to see, ﬁnd or understand something attributable to
a lack of sufﬁcient background knowledge. How does this
impact the performance of the group as a whole?

 

 

Power and/or
knowledge as
moderators of
group activity

 

Note instances of differences in rank in members of the
team, particularly cases where there is a reporting
relationship. Did this affect behavior? Was a member of
the group recognized to have superior expertise in any
particular area? Did this affect behavior? (This must be a
subjective judgment since there is no control sample
available for reference.)

 

71

 

 

Laple 1: continued

 

Shared logics of
action (existence
of and changes
to)

This is somewhat difﬁcult. Martin (2002) notes that
“shared” doesn’t mean universally agreed upon and it may
actually be the case that no one agrees with it as in the
“Abilene Paradox” (Harvey, 1988). Rather, is there a
recognized dominant logic that motivates the group? Key
indicator — do individuals demonstrate a sense that this
logic is beyond question, or at least beyond their ability to
inﬂuence? This may be easier to see as an outsider but
there will be no hard evidence, since the insiders typically
cannot see that it is an issue. An important question is
whether any changes to the dominant logic were observed

during the course of learning.

 

 

 

Changes to This is a simple tabulation. What differences are there in
performative the work process as a result of the improvement efforts?
aspects of

routines

Changes in Is the process veriﬁably better as a result of the

situated “improvement” efforts? By what measures?
performance

 

 

 

Experimental Design:

Having decided on a research method of in-depth case studies, it is still
necessary to choose a unit of analysis that will allow the research question to be
investigated and in order to deﬁne appropriate criteria for case selection. Each of
the entities considered in the framework development above represents a

potential unit of analysis, as does the work process or routine that is the subject

72

oft
ind
vvae
act

sta

Tor
Pnn
bour
end.
occu

lhera

 

of the improvement effort. Each of these is also potentially incomplete as
indicated by the web of relationships postulated in Figure 4. What is really
wanted is to capture an event, deﬁned as a bounded set of individuals and
activities within which an operating process of an organization is moved from one

state of knowledge to another, as shown in Figure 5.

      

   
   

 

Knowledge State
t = T1

Process Improvement
Event (T1 < t < T2)

Knowledge State
t = T2

 

 

Figure 5: Process Improvement Event

To constitute a useful unit of analysis, such an event must satisfy several criteria.
Principally, it must have measurable beginning and ending states. This both
bounds the event in time, and allows us to verify that a process improvement (the
evidence of learning deﬁned as an increase in situated performance) actually
occurred. Additional criteria can be identiﬁed from the lessons learned in the

literature review of prior process research (Chapter Two).

To avoid the situation of a set of single-sample anecdotes, it is important to
establish comparability, as in the ability to observe multiple, comparable events.
The intent is to search for similarities (patterns) among events, so we do not want
to mask similarities by trying to compare greatly different types of events.
Speciﬁcally, we are looking for events that conform to some consistent standards

so that we can compare the actions of different (but comparable) people in

73

di

pr

CTI
a l
the
ex
de‘
pu

sta

USU;
Prep
3 SR"
deS

06s

different (but comparable) organizations addressing different (but comparable)

problems.

There are additional criteria. To properly observe the process of knowledge
creation, it is necessary to simulate as, much as possible, a closed system, as in
a laboratory setting. This requires that the event be bounded in scope so that
there is a clear distinction between what is endogenous to the event and what is
exogenous (and that the two do not mix). In practical terms, this implies a
deﬁned group of people working together for a ﬁxed period of time with a deﬁned
purpose. A further consideration is that the event should be continuous from

start to ﬁnish so that unobservable activities are not part of the process.

These are difﬁcult conditions to establish, illustrating why it is difﬁcult to do good
process-oriented research. Such situations do exist however, and the conditions
described above are characteristics of a generic class of short-cycle process
improvement initiatives known as Kaizen events. Kaizen is a Japanese term
usually translated as continuous improvement, and Kaizen events as typically
practiced are short cycle projects (typically about one week in duration) involving
a small, dedicated team focused on improving a single work process. This
description corresponds to what Melnyk, Calantone, Montabon and Smith (1998)

described as “second level, tactical events.”

In their work, the deﬁning traits of these events were listed as: Very Short
Tenn/Finite in Life, Highly Focused, “Creativity before Capital,” Team Oriented,
Action Oriented, Veriﬁable Metrics, and Repetitive. Such events have evolved to

a standardized event structure consisting of six major phases: (1) Training; (2)

74

th

of

3F

Hc

OTC

anr

Pro

inst

inte.

 

Documentation of the process “as is;” (3) Identiﬁcation of potential opportunities
for improvement; (4) An immediate, iterative process of implementing and
verifying the improvements; (5) Presentation of the results to management; and
(6) Preparing the action list for next steps (Melnyk, Calantone et al., 1998 and

see also Stewart & Melnyk, 2000).

Kaizen events were selected as an appropriate “laboratory” within which to
observe the process of organizational knowledge creation. As can be seen from
the description above, they provide a consistent format so that recurring patterns
of behavior can be observed. They also provide the necessary bounding,
approaching a closed system as closely as is possible in an Operations setting.
However, identifying an appropriate unit of analysis for study is only part of the
process of research design. Selection of appropriate cases is equally important

and can support or undermine the experimental design.
Case Selection:

Proceeding to case selection, the nature of the research suggests the use of
instrumental cases (Stake, 1994). In other words, the case itself is of no intrinsic
interest, it is simply a vehicle for examining the phenomenon of interest. The
ideal case should be as typical of a larger population as is feasible. Stake (1994)
points out that the case and the unit of analysis do not have to coincide.
However, for the purposes of this study, that is a natural choice providing that

some additional criteria with respect to feasibility and controls can be satisﬁed.

Adopting Kaizen events as cases goes a long way towards ensuring feasibility for

study by bounding the cases in scale, scope and time, all issues which could

75

pre
will
inte
qua
fear
hav
and

worl

The
The
a rel
saml
envir
This
pr em
little I
not a

varia

A Crit
corp
The

and

 

llprhl

present problems for a researcher or research team. An additional advantage
with respect to feasibility is that the events are scheduled. While it might be
interesting and valuable to study unplanned events such as a response to a
quality problem, this could only be observed through serendipity. A ﬁnal issue of
feasibility that must be addressed in case selection is that the researcher must
have access to the event. Process improvement does not take place in public
and may involve proprietary information. An invitation to observe the true

workings of the event must be secured.

The issue of the controls to be placed on the case selection is more complex.
There is a persistent tension between enhancing the ability to isolate and identify
a relationship and enhancing the ability to generalize it beyond the research
sample. The former objective would call for comprehensive controls of
environmental variance, while the latter would be enhanced by fewer controls.
This study will lean in the direction of improving the identiﬁcation on the general
premise that until it is possible to conﬁrm the existence of a relationship, there is
little reason to be concerned about generalization. As a result, external validity is
not a primary concern of this research, and we will attempt to control for as much

variance as possible.

A critical early choice was whether to seek out Kaizen events from different
corporations or to study events taking place under a single corporate umbrella.
The latter alternative was eventually chosen although it was not a clear choice
and there are advantages and disadvantages. The principal advantages are that

it provides a consistency in the structure of the Kaizen events and that it

76

autr
dis;
anc
ado

cori

conr
perr
vvou

The

COpr
they

was:
resee
Pmdt
quest
whkﬂr

Ihes

Oper.

automatically controls for industry and some ﬁrm effects. The major
disadvantage is that the “organization” is an element in the research framework
and we do not want to remove variance in that element. This issue was
addressed by insisting that all cases come from different operating units of the

corporation with separate reporting structures.

A target corporation (referred to hereafter as the Corporation) that was known to
conduct signiﬁcant numbers of Kaizen events was approached and asked for
permission to observe some of the events. It was agreed that a researcher
would be allowed to attend a number of events acting as a participant-observer.
The signiﬁcance of this role will be discussed in more detail below. Since the
corporation regards the content of these events to be proprietary and because
they wish to remain anonymous, a comprehensive non-disclosure agreement
was signed. As a result, only generic data and terminology will be used in this
research report. This presents no difﬁculty for the research, since precise
product details and measures of improvement are not material to the research
question. An overview of the Corporation will be provided in the next section,
which will show that it represents an instrumental choice of research setting in

the sense of not being exceptional in any way.

Within the corporate family, there was a range of events to choose from, but
many were ruled out according to the criteria above and in the previous section.
To ensure comparability of the project teams and to control for variance due to
functional area, only events focused on process improvement in manufacturing

operations were considered. (This is also consistent with the positioning of the

77

£3t
vvn
var
rne
part

coHe

Ase;
Case
Vwor
separ
thhdr
Case

there

comp:
Sam

The i

anal

research question.) The requirement that the event be continuous ruled out any
event that was conducted in two parts or that represented a continuation of an
earlier event. This two-part structure was fairly widely used in the target
corporation, and while it has certain practical advantages, it compromised the
closed system ideal because the participants cannot be observed in the

intervening periods.

Still further restrictions were placed on the case selection. Only cases occurring
within English-speaking North America were considered to avoid cultural
variance. Also, only operating companies with previous exposure to the Kaizen
methodology were considered so that the process itself would not be seen by the
participants as a novelty. Finally, a six-month window Was placed on data

collection, controlling for ﬂuctuations in the economy.

As a result of this rigorous and purposeful selection process, three research
cases were chosen from different operating companies within the corporation.
Two of these companies were within the plumbing products group (but with
separate reporting structures and very different corporate histories), while the
third was under Specialty Products. The Kaizen events that comprised these
cases took place between December, 2004 and June, 2005 during which time
there were no major shifts in the economic environment experienced by these

companies.

Sample Size:

The issue of sample size is always a difﬁcult one, particularly in case study

analysis with qualitative data, where is no standard for what constitutes “enough"

78

if

th

cc
of
the
exi
daI
cor
whr
furti

Eleg

Afte
suffi
in a

feco‘

II sh
0f sq
dire

mu.

cases. Conventional wisdom suggests that larger is better. In the inductive
mode of discovery (Hunt, 1991), the idea is that more data would convince us of
the universality of the patterns being observed and hence, the generality of the

theories that can explain them.

Kosslyn (1978) points out an interesting counterargument that hinges on the
concept of elegance in a theory (elegance being an un-measurable combination
of parsimony and “naturalness”). This concept is a sort of Dutch auction of
theories. The idea is that the full observations of any one experiment can be
explained by a number of theories, none of which are likely to be elegant. The
data set must be reduced by replicating the experiment and retaining only the
common elements. At some point the common themes are reduced to a point
where a single elegant explanation is possible. Here one should stop because
further simpliﬁcation of the themes will only result in more and more potential

elegant theories that explain less and less of the variance.

After three cases, it was judged that the interpretation of events had stabilized
sufﬁciently that additional research effort would be more productively employed
in a different set of cases: perhaps in non-manufacturing settings as

recommended in the conclusions.

It should also be noted that counting cases is not necessarily the best measure
of sample size. The three cases reported on here represent 21 person-weeks of
direct subject observation, 38 identiﬁable process improvement activities and a

much larger number of actions, statements and other data points.

79

8a

The
div
bui
ove

Co

dee
pro
rete
Cor
mar

derr

 

Research Subject; An Overview:

The corporation that is the subject of this research (“Corporation”) and its various
divisions manufacture sell and install a wide variety of home improvement and
building products, with an emphasis on brand name products. While consisting of
over 50 separate operating companies (referred to as “divisions” by the
Corporation), the Corporation is organized into ﬁve product-based business lines:
cabinets and related products, plumbing products, installation and other services,
decorative architectural products (e.g., paints), and other specialty products. Its
products are sold through a variety of channels, including large (‘big box’)
retailers, distributors, and professional installers (e.g., plumbers). The
Corporation is differentiated within the building products and home improvement
markets for being one of the few ﬁrms that can offer the “one stop shopping”

demanded by large chains such as Menard’s, Lowe’s, and Home Depot.

For 2004, total revenues exceeded $12 billion, with net income totaling over $900
million. Overall, the Corporation is recognized as being a well-managed
organization. For the last ﬁve years, as an example, net income has grown at an
average of 13% per year. The Corporation has developed a reputation for

dependable growth even through weak economic times.

Until recently, the Corporation operated in the style of a holding company that
managed its businesses as a portfolio of investments. Each of the divisions was
managed as an autonomous company. Entrepreneurship was encouraged within
these various divisions, with the unintended result that some of the divisions

providing either the same or similar products found themselves in competition

80

with each other. While this is changing in response to market forces, the
historical legacy of autonomy is strong. Over the last ten years, the Corporation
aggressively acquired proﬁtable and well-managed companies with market
positions that were competitive or complementary to existing divisions or that
enabled the Corporation to expand its product lines in its core markets. While
this strategy too is changing, it was complementary with an autonomous
structure, since the promise of autonomy was often an advantage in making an

acquisition on favorable terms.

The Corporation is changing in response to external pressures, but to understand
some of the dynamics at work in these cases, it is necessary to understand the

history as well as the current challenges. These external challenges come from

three primary areas.

The ﬁrst is a consolidation of major customers. This consolidation is seen both in
the growth of “big box’ retailers such as Home Depot, Lowe’s and Menard’s, and
in the builder market with the growth of Pulte and KB Homes. This has created a
fundamental change in how the home improvement business is conducted. The
most obvious change is a shift in bargaining power from seller to buyer. The
biggest customers now command sufﬁcient market share that there is no
alternative but to do business with them on their terms. This has resulted in not
only in increased price pressures, but also in demands for improved quality and

more ﬂexible delivery schedules.

The second challenge comes from the changing nature of the consumer market

itself. In recent years there has been a trend of “bar-belling” demand in many

81

consumer goods markets. That is to say, there is demand for the very best
products and for the very cheapest, with diminishing interest in the middle
ground. It is precisely this middle ground where the Corporation's major divisions
have gravitated, driven by a focus on proﬁtability. As a result, the company is
under pressure on both the high end and the low end. The low end pressure is
more serious, and is largely the result of the emergence of China as a low-cost
manufacturing center. This emergence has been strongly facilitated by the big-
box customers, with their global sourcing reach and the ability to create private
brands. While some of the Corporation’s divisions are insulated (for now) from
foreign competition, the divisions selected for these case studies are signiﬁcantly

affected.

The third major challenge to the historical strategy of the Corporation comes from
the investor community. In the 90’s, the Corporation (like many others) pursued
a strategy of growth by acquisition. In an era of rising expectations, acquisitions
could be made cheaply with stock and options, and Wall Street rewarded these
companies with high valuations. In such a market, the chief role of existing
divisions was to increase margins through cost reductions: growth was handled
at corporate. This explains the Corporation’s receptiveness to Kaizen event

methodology as a cost reduction tool.

In the current economy, investor preference has shifted to strategies that
emphasize innovation and ﬂexibility. This creates a tension between the old and
new ways of thinking in the Corporation that was visible in all of the cases. If

there is a bright spot for the Corporation, it is that the housing and home

82

 

imi
the

the

re
ol:
ac
pa
be

to

en
the
for.

doi

Kai
0/3

0a

improvement markets have experienced something of a boom in recent years, so
the shifts in strategy are being played out in a more benign market environment

than is often the case.
Entering the Field: Data Collection:

As previously noted, the role of the researcher in these cases was that of a
participant-observer (Martin, 2002). This has a number of implications for the
research and explains why it was not feasible to have multiple researchers
observing events. The choice of the participant-observer role offers a number of
advantages. Principally, it avoids the tension created by the presence of a
passive observer. The other team members quickly overcame the sense of
being watched and interacted normally with the researCher. This was validated
to an extent during some of the cases, particularly Case #2. When the team had
to collect data for standard work calculations, it was necessary to observe line
employees at work. It quickly became clear that they performed differently when
they felt they were being watched; it wasn’t until they were engaged in less
formal conversation later that they provided a true picture of what they were
doing. A secondary beneﬁt of the arrangement was that it was much easier to

interact with the subjects to clarify their intents and meanings.

Kaizen events begin as do most team events — with mutual introductions. This
opportunity was used to brieﬂy explain the purpose of the research to the
participants and to obtain their informed consent. Particular emphasis was made

of the fact that the purpose was not to evaluate them. In the context of an

83

inst
doii
Sin
cor

abc

ReI
of iI
forr
ack
con
plat
beir
worl
vaIic

res;

 

ahr

pr 0

do

instrumental case, it was explained that they were presumed to be normal people

doing a good job, and that is why it was of interest to observe them.

Since this is research involving human subjects, certain legal and ethical
considerations had to be observed in addition to the informed consent noted
above. In accordance with Federal law, approval was obtained from UCRIHS,
(University Committee for Research Involving Human Subjects), the Independent
Review Board of Michigan State University, to conduct the research on the basis
of informed consent (approval: IRB #04-876). To document informed consent, a
form was developed which the participants were asked to sign as
acknowledgement that they were satisﬁed with the steps taken to preserve
conﬁdentiality. These steps include the elimination of any names (persons or
places) from the electronic records of the event. Since the participants were not
being asked to do or say anything that they would not have done in their normal
work, no one expressed any concerns about conﬁdentiality. In the case of the
validation survey given to the expert panel (to be discussed below), the

responses were not only kept conﬁdential, but were anonymous as well.

It is recommended by Yin (1994) to triangulate the data by using multiple
sources. This was done in these cases as all relevant sources were made
available to the researcher. Because of the “how” aspect of the research
question, the principle source of data was observation of what individuals said
and did. This was augmented by observation of the physical layout of the work
process both before and after modiﬁcation. In addition, various sources of

documentary evidence were consulted. These sources included production

br

QU

Dre

schedules and forecasts, previous process ﬂow analyses, bills of material and
cost data. Occasionally, non-team members were brought in for consulting,
providing yet another view. Most of these supplemental sources tended to be
peripheral to the research question, in the sense that they did not address

process issues, but are noted in the transcripts when they were used.
Veriﬁcation of Data:

The data records of the events consisted of a transcript that reﬂected as
accurately and objectively as possible what took place during the event.
Veriﬁcation deals with the question of whether the records are: (a) accurate, and
(b) complete. Accuracy was dealt with in the sense of being “not wrong.” Copies
of the transcripts were sent back to the event facilitators with the request to note

any errors of fact. No such errors were reported.

The question of completeness is more difﬁcult, since no record can ever really be
complete. A more conservative approach is to consider completeness with
respect to the protocol. Even this is not comprehensive since important events
might occur that were not anticipated by the protocol. The protocol is also a
specialized document, familiar only to the researcher. A common approach to
this issue is to use additional researchers (who are familiar with the protocol) so
that notes can be compared after the fact. In the case settings described above,
bringing additional researchers into the events was considered to be out of the
question, so the best insurance for completeness is a clear and comprehensive

protocol.

85

Analysis of Data and Conclusions:

The transcript of the event contains, as far as possible, no interpretation of why
something happened. This was done primarily to separate veriﬁable facts from
the interpretation. The veriﬁcation step above was based on the factual layer,
while the validation applies to the interpretation. To arrive at the interpretation,
once the transcript was complete, it was followed by two layers of analysis.
These correspond closely to open or ﬁrst level coding as described by Miles and
Huberman (1994), and to selective coding (Strauss & Corbin, 1990). As
explained by Ellram (1996), these categories are iterative and interdependent. It
is sometimes advised to use a category in between these called axial or pattern

coding (Miles & Huberman, 1994).

It can be difﬁcult to tell where one technique starts and another one stops. The
differences are essentially these: Open coding is used to establish relevant
constructs from the data, axial coding is to look for relationships between the
identiﬁed constructs, and selective coding is to identify and examine alternate
patterns. Coding is perhaps not quite the right word for the last instance, since it

is more a re-examination of the data with speciﬁc patterns in mind.

In these analyses, the ﬁrst layer is recorded in the form of marginal notes as
described by Miles and Huberman (1994). In these notes are included any
observations of issues that suggest themselves as important, evidence for issues
suggested by the research protocol and some explanation about the signiﬁcance
of what was observed. The concept of marginal notes is important because

these observations must be keyed to speciﬁc pieces of data to retain their

86

validity. Because of the use of the research protocol, it was possible to
effectively combine open and axial coding in this one step. Explanatory and
subjective notes were typically added at the end of each day so that the events
were fresh in the researcher’s memory. Additional notes were added from a

retrospective view of signiﬁcance, corresponding to more purely axial coding.

The second layer of analysis corresponds more to selective coding. In this mode
the transcripts were reviewed and compared to the protocol. The presence or
absence of the traits in Table 1 was noted, along with any evidence conﬁrming or
disconﬁnning their signiﬁcance. In so doing, different explanatory patterns were

being applied to the data in search of supporting evidence (or lack of it).

From these two layers of analysis, some preliminary conclusions were formed.
This is largely a synthesis process and is not easily speciﬁed in advance. An
important part of the synthesis process is an awareness of what was not
observed, along with what was. These conclusions are discussed in Chapters

Four and Five.
Validation of Conclusions:

When conclusions are formed from a small set of cases it is desirable to test their
validity against a larger data set. When the conclusions can be reduced to
precise propositions based on validated constructs, a survey followed by
statistical analysis is the preferred method. At this stage of the research, the
conclusions did not lend themselves to this treatment, so an alternate approach

was employed.

87

A panel of seven experts was assembled from a technical services group in the
parent corporation. The bases for considering them to be experts were: (a)
signiﬁcant industrial expertise, (b) a high proportion of that experience devoted to
continuous improvement activities and (c) involvement in (as facilitator or
participant) a high number of Kaizen or other continuous improvement type

events.

The panelists were assembled and given a twenty-minute summary of the
research project, its scope and objectives and the preliminary conclusions. This
was followed by approximately forty minutes of open discussion. A transcript of
this discussion is attached as Appendix VII and is discussed in Chapter Four. At
the end of the discussion, the panelists were asked to anonymously answer a
short questionnaire on their agreement or disagreement with the conclusions.
The questionnaire is reproduced as Appendix I. The results are given in
Appendix VI and are also discussed in Chapter Four. These results were
incorporated into the propositions that are developed in Chapter Four and in the

conclusions that are stated in Chapter Five.

Summagy:

In order to arrive at an appropriate research design, this chapter began with a
discussion of the inductive approach to theory-building. In this approach, speciﬁc
observations are collected, classiﬁed and analyzed in order to state more general
propositions. This approach demands rigor at all phases to overcome the threats
to validity that are inherent in the method. While speciﬁc details of the research

design were chosen to minimize these threats, the principal reliance is on the

88

 

clc

de

Th

3P

SU.

39

 

 

closely related research framework and research protocol, both of which are
developed in this chapter. Their key features are that they are grounded in the

literature and consistently applied in the ﬁeld.

The detailed design of the research follows from the choice of the inductive
approach. A case study method, employing sustained direct observation and
subsequent qualitative analysis of the ﬁndings was chosen as the approach most
appropriate for generating insight relative to the research question. Case
selection is a critical phase of the research design, and speciﬁc Kaizen process
improvement events were chosen as cases that met the dual needs of theoretical

interest and feasibility for study.

The ﬁeld data are qualitative in nature, such that the analysis must be some form
of pattern matching (Kaplan, 1964). Given the lack of intersubjective
conﬁrrnability in this method (Hunt, 1991), a validation procedure was established
to conﬁrm that patterns observed in the data were consistent with the

experiences of a panel of experts in the ﬁeld.

The actual recording and analysis of the data are discussed in Chapter Four.

89

 

rm

The
relei
Thre
obse
gen:
then

leari

The
indu
som

invo

COI‘:

ilar

CHAPTER FOUR - RESULTS

Introduction:

The purpose of this chapter is to presentthe results of the ﬁeld research that are
relevant in the context of the research framework and protocol derived in Chapter
Three. Broadly stated, these results are expressed as a summary of
observations within the structure of the research protocol that are found to be
generally true across the cases studied in this research. These observations are
then stated in the form of potential propositions in a theory of organizational

learning, following the model of Bacharach (1989) (Figure 2, Chapter Three).

The derivation of these propositions, moving from the speciﬁc to the general, is
inductive, per Hunt (1991). As previously noted, the inductive approach is
somewhat limited in the availability of tools of explanation. Since this study
involves qualitative data, the method used here will be pattern matching as
described in the preceding paragraph (Kaplan, 1964; Hunt, 1991). Because
pattern matching as a tool of explanation suffers from a lack of intersubjective
conﬁrmability (Hunt, 1991), there is an onus on part of the researcher to create
transparency in the process so that the reader can retrace the steps. This

chapter is organized to present the results in a way that maximizes transparency.

The individual case reports are the foundation of this chapter, but in the interest
of clarity, they are not incorporated in the body of the text. Instead, Table 2 is
included in this chapter to provide a summary overview of the cases. The actual

case reports are attached as Appendices II, III and IV. Each of these appendices

90

is laid out in the same way. The introduction contains an overview of the

company setting in which the event took place and summary of what occurred,

including the results achieved by the event.

 

Case #

ype of Kaizen

Table 2: Summagy Description of Cases
m

#1

Second level,

#2

Second level,

#3

.—————-+———-4

Second level,

 

 

 

Event tactical tactical tactical

Company Plumbing Drain cleaning Plumbing products

type/products products - faucets equipment - shower valves

Dates December 14-17, March 14-18, June 27-July 1,
2004 2005 2005

Location Ontario New Jersey Indiana

 

Process Selected
for Improvement

Semi-automatic
brazing carousel.

Steel machining,
fabrication and

Final assembly
and pack-out.

 

 

 

 

 

 

 

 

weldingcell.

Overall Goals of Reduce time to Reduce WIP Streamline

Event change ﬁxtures between weld and material ﬂow
from one paint, reduce labor during part
assembly type to hours. changes, reduce
another. clutter and

improve operator
efﬁciency.

Speciﬁc Targets 50% reduction in 10% improvement Redesign pack-
setup time, in productivity, out station,

4 safety/ergo 90% reduction of Implement

improvements, welded frame safety/ergonomic

1 poke-yoke WIP, reduce ﬂoor improvements,

implementation. space by 10%, Implement
implement 3 mistake-prooﬁng
safety/ergo (poke-yoke).
improvements.

Facilitation Dual, one from Dual, one from Dual, both
corporate, one corporate, one internal, one
internal. internal. newly trained as
Corporate Corporate facilitator, neither
facilitator had facilitator had had experience of
experience of > 50 experience of > 50 more than ~10
events. Internal events. Internal events.
facilitator <5. facilitator <5.

Participants (total) 9 7 8

 

91

 

 

 

 

 

Table 2: Continued

 

 

 

 

Modiﬁcations 17 14 7
Made
Results Achieved 49% setup time Implemented Built new packing
reduction, cyclical painting tables for better
7 safety/ergo schedule and pull ﬂow and
improvements, system - ~75% ergonomics, re-
1 poke-yoke reduction in WIP, organized part
implementation. use of standard stock to match
work and small part sequencing,
batch ﬂow to redesigned
reduce labor assembly ﬁxtures
~14%. with mistake-
prooﬁng.
Comments Important Uncertain that Company was
distinction improvements will more concerned

 

between reduction

 

be sustained in

 

with qualitative

in man-hours for face of than quantitative
setup and nervousness measures
reduction in down- about lack of (reduction in
time — focus was buffer stock. clutter, operator
on the former. testimonials).

 

This is followed by what amounts to selective coding (Strauss & Corbin, 1990),

where the data are systematically reviewed according to each element of the

research protocol for the presence or absence of conﬁrming evidence. This

review also includes observations on certain factors that were not speciﬁcally

considered in the research protocol, but emerged from the actual ﬁeld notes as

being potentially interesting. These observations correspond to axial coding in

the terminology of Miles and Huberman (1994). Finally, each appendix includes

the actual case notes, in the form of a transcript of the event. The transcripts are

augmented with the researcher’s marginal notes which offer explanations of

certain events and serve to tie the conclusions to speciﬁc evidence.

92

 

It will be noted that the case reports and the ﬁeld notes are sparse in identifying
details. This is deliberate in order to protect the subject company from any loss
of proprietary information and also to protect the identity of the individuals
involved, even within the company. Some terms may seem cryptic in the
absence of any deﬁnition or explanation, but typically, these are terms that are
used locally for identiﬁcation of speciﬁc processes or products and are not

considered material to the issues being observed.

Returning to this chapter, the ﬁrst section is a review of the research protocol that
integrates the similar reviews in each of the individual case reports. At this level,
the interest is in the presence or absence of conﬁrming evidence that holds
across a majority of observations, remembering that each case may contain
multiple observations. The additional themes that emerged from the ﬁeld notes

are also considered in this section.

The chapter concludes with the development of a set of statements about
relationships that were observed to exist between the elements of the research
framework (Figure 4, Chapter Three). These relationships are stated in the form

of potential propositions in a theory of organizational learning.

At a preliminary stage, these observed relationships were reviewed and
discussed with an expert panel as described in Chapter Three. The
qualiﬁcations of the panelists as experts in the ﬁeld are given in Table 3. The
notes of the panel discussion are provided as Appendix VII, and the insights
provided by the panel are incorporated in the discussion of the research protocol

that follows. These comments and insights were supplemented by a survey, the

93

purpose of which was to provide empirical validation of the panel’s degree of
agreement with the tentative propositions. The survey itself is attached as
Appendix I and the results are tabulated in Appendix VI.

Table 3: ualiﬁcations of Validation Panel

Years of
Industrial Years of Cl or Kaizen
Panelist Experience Experience in CI Events

1 20 9 120
2 10 6 1 30
3 12 2 30
4 17 10 50
5 16 14 75
6 20 10 100
7 40 30 150
Mean 19.3 11.6 93.6
Std. Dev. 9.9 8.9 43.8

94

Lart I: Review of the Research Protocol:

Knowledge ClassiﬁedyAs Beﬂg_E_xplicit Or Tacit:

The study of epistemology provides us with two major ways to classify and
describe knowledge itself. One is a typology or classiﬁcation according to the
essential nature of the knowledge, speciﬁcally the classiﬁcation of tacit versus
explicit knowledge. The tacit/explicit distinction proved to be a valid one, as clear
examples of each could be observed. The usefulness of the view, particularly
when speaking of organizational knowledge, is less clear, as will be discussed in

the next section.

Two types of tacit knowledge were observed in all of the cases. The ﬁrst type
was an individual skill, such as adjusting brazing burners, bending tubes or re-
ordering a production sequence. While these skills certainly represent
knowledge, they do not meet the deﬁnition of organizational knowledge in the
sense that the organization can only make use of these skills by employing their
owners. A conversion of these skills to some explicit form (see next section)
would make them appropriable by the ﬁrm, but this did not occur in any of the
cases. Given the beneﬁts that these skills confer on their owners, it could be

expected that such conversions would be resisted.

A second form of tacit knowledge was observed that was very much a property of
the organization. This consisted of a pervasive but largely unspoken sense of
priorities and beliefs about “how things should be done around here.” This type
of knowledge does have certain reﬂective indicators (performance metrics for

example), but these indicators are not always mutually consistent. It is the

95

unstated balance of priorities that makes this knowledge tacit. It was very difﬁcult
to observe this type of knowledge, or even to recognize its existence. In this
respect, it is an element of culture and probably best understood in that context

as will be discussed later.

Under this typology, any knowledge that is not tacit is explicit by default.

Certainly any physical changes to observed procedures would qualify as
examples of learning resulting in the creation of explicit knowledge. Thirty-eight
such examples were noted, providing a fairly large sample for observing the
mechanisms at work. The strict count of events should not be taken too
seriously however, since the events are not all equally signiﬁcant and many could

be arbitrarily broken down into smaller elements or aggregated into larger ones.

While this classiﬁcation seems straightforward enough, it proved to be
troublesome in a key aspect. The root of this trouble is noted by Pentland and
Feldman (2005) when they decompose routines into performative and ostensive
components. While the performative aspect of a routine will generally be explicit,
the underlying rationale may or may not be. As an example, the case of the
revision to the part cooling mechanism (Case #1) had performative and ostensive
components that were both explicit. The performative component (the revised
“shower head”) was purely explicit, and the ostensive part (cool the part
sufﬁciently for the operator to handle it and uniformly enough to release freely
from the ﬁxture) was explicit in the sense of being readily articulated and

universally understood.

96

Other examples were mixed. For example, the fabrication of a “supermarket”
table for welded frames in Case #2 is again an explicit performative component,
but it really only constitutes organizational knowledge if it is instituted with the
intent to employ “pull” production. As described above this intent constituted tacit
knowledge in the sense that the organization could not clearly or universally
articulate why it did (or did not) believe that this was the correct approach and

under what circumstances that would be true.

To summarize this section, it can be said that while both tacit and explicit
elements of knowledge could be observed, this distinction by itself was not
sufﬁcient for a clear and unambiguous classiﬁcation scheme. It is truer to say
that anything qualifying as organizational knowledge (as deﬁned in this research)
could contain both tacit and explicit elements in varying proportion. Other
elements of this research protocol (to be discussed below) offer more insightful
ways to classify organizational knowledge. The tacit/explicit distinction was not
discussed explicitly with the expert panel since it did not seem to be an important

building block for propositions.
Lmnsfomptions of Knowledge Within gag Between Explicit and Tacit Forms:

In the view of Nonaka (1994), learning requires transformation of knowledge from
one of these forms to another. Given that there are two classiﬁcations, there are
four potential transformations that can be considered. We will deal ﬁrst with the
three possible transformations involving tacit knowledge, and preface the
remarks by saying that no such transformations were actually observed during

the study.

97

At the individual level, tacit knowledge describes something like a skill or art that
can be passed from one person to another through what Nonaka (1994) refers to
as socialization. Our results contain evidence that this had in fact occurred in
situations where an individual trained his or her counterpart on an opposite shift.
It was also apparent that much of this individual tacit knowledge could, in
principle, be transformed to explicit knowledge with enough effort. Such a
transformation would be interesting because it would represent a conversion
from a personal form of knowledge to a form appropriable by the organization.

This did not actually occur in any of the cases studied.

The situation was less clear with respect to the organizational form of tacit
knowledge, especially given the above argument that organizational knowledge
can potentially consist of both tacit and explicit components. In fact, it was
difﬁcult to deﬁne what would constitute a transformation in that context. In any
case, if we consider organizational tacit knowledge to be the dominant logic
about “how things should be done around here,” this was static during the course

of all of the case studies, so no transformations were observed.

In principle, the tacit portion of organizational knowledge could be made explicit
through means such as performance metrics, but it is not clear how a
transformation in the reverse direction might work. It is equally unclear how a
tacit-to-tacit transformation might take place with such knowledge, except
perhaps the case where newcomers are indoctrinated with the dominant logic.
Any change to the dominant logic of an organization would seem to be a

replacement phenomenon rather than a transformation, but no such change was

98

observed within the scope and context of any of these cases. With respect to
tacit organizational knowledge, the concept of knowledge creation through

transformations was not helpful or observable.

Since a large number of explicit outcomes of learning was observed, the
remaining transformation proposed by Nonaka (1994), explicit-to-explicit, should
have been widely observable. Conceptually, this would take place when a
particular instance of knowledge is observed in one setting and subsequently
applied in another. This sequence proved not to be observable in practice. To
the extent that it is possible to infer mental processes from observation, the
sequence went more as follows: An individual encounters an instance of explicit
knowledge. From this, the individual abstracts a set of principles and
associations, which are then combined with other such sets obtained from other
observations to form a mental model. The resulting (abstract) model is then
applied to the individual’s particular circumstances to create a new form of

explicit knowledge.

It was extremely difﬁcult to describe these events in terms of explicit-to-explicit
knowledge transformations. For one thing, of the thirty-eight noted examples of
explicit knowledge creation, only ten had any observable antecedents. Even in
those cases, it was clear that the apparent antecedent was not the only input.
The interesting point is that explicit antecedents were not imported in purely
explicit form. Rather, they passed through some state of abstraction in personal
memory before being applied in a given situation. This is consistent with the

observations of El Sawy, Eriksson, Raven and Carlsson (2001).

99

Some examples: In Case #1, when a sloped parts table with return chute for
empty boxes was proposed, the proposer did not have a clear design in mind.
Instead, he had a mental picture of what such solution looked like, from which he
could sketch up a variation that would ﬁt the particular application. In Case #3,
when the new work table was being designed, the designer employed some
ergonomic principles. This did not involve any actual measurements or reference
to written standards, but he clearly had formed a mental image of what an

ergonomically “good” table looked like.

The conclusion from the ﬁrst section above that organizational knowledge
contains elements of both tacit and explicit knowledge is consistent with these
observations. However, the learning process is more easily and more
meaningfully described in terms of information processing (from psychology) than
from a transformation perspective (epistemology). This will be discussed in more
detail below. For this reason, the concept of transformations between tacit and

explicit knowledge forms was not discussed with the expert panel.

Knowledge Classiﬁed By Location

Rather than classifying knowledge according to its type, it is also possible to do
so based on its location. This requires adopting a reiﬁed view of knowledge as
an identiﬁable thing. This view seemed promising at ﬁrst, but proved ultimately to
be uninteresting. It was not too difficult to catalog the sources that people
consulted in the organizational learning process (discussed in the next section);
what was difﬁcult was determining exactly what it was that they got from those

sources. It was never possible to say that someone acquired a speciﬁc piece of

100

knowledge from a speciﬁc place and applied it in anything like its original form.
Mostly they retrieved certain elements of knowledge that had to be adapted to

the situation at hand (see the discussion above about transformations).

These attempts to track knowledge by its location suggest that what constitutes
knowledge in one context may not be knowledge in another. This is consistent
with practice theory and the deﬁnition of organizational knowledge as situated
performance. In this view, it is more useful to think of knowledge as something
that is created at the point of use from some combination of local facts and
imported organizing principles (derived from knowledge in other contexts). This
view of knowledge can be seen in Hayek (1945) and Greeno, Moore and Smith

(1993).

In the cases under study, the creation of new knowledge always required some
combination of these factors - no examples can be cited where knowledge was
retrieved and applied unchanged. Perhaps the closest example is in Case #1
where the idea for cooling water was copied from another line. Even in that
instance it was only the concept that was imported; there was still a degree of

adaptation required.

Just as philosophers tend to gravitate to the position that knowledge is in some
way speciﬁc to the knower (Polanyi, 1958), it makes sense to say that
organizational knowledge is unique and speciﬁc to the organization and its
circumstances. Taking that view, it is not meaningful to say that organizational

knowledge exists in speciﬁed “bins” or even to suggest that it can be stored.

101

This concept of organizational knowledge as something unique to its setting (and
being deﬁned by situated performance) was discussed with the expert panel as a
precursor to the preliminary propositions. There was some ambivalence about
this deﬁnition that pervaded the discussion. It was clear that those on the panel
did not make a sharp distinction between individual and organizational
knowledge, and tended to apply a fairly broad interpretation of the term
knowledge in general. This conﬂicted somewhat with the research needs for
precise deﬁnitions, so this difference of interpretation had to be kept in mind

during discussions with the panel.
Actions By Individuals To Retrieve Or Store Knowledge By Location:

A consistent pattern of behavior was observed across all cases. When
confronted with the need to create knowledge (solve a problem), individuals
looked ﬁrst to their own experiences, second to the experiences of others
(through direct questioning), third to observable examples of similar situations
and last to archival or documentary information. It is tempting to attribute this to
normal human economy of effort, but there are important qualitative differences
between these sources that we can describe as a hierarchy of knowledge

density.

This idea will be developed further in the next section, but the essence is that the
story is in many ways more important than the bare facts. It can be seen that
these sources are in declining order of richness in terms of their context and
associations. When documentary evidence was introduced into these cases, it

tended to be devoid of context and was extremely hard to interpret. It had high

102

information density, but low knowledge density. In most of the cases where
documentary evidence was consulted (such as production schedules and
forecasts or process ﬂow diagrams) it was necessary to ﬁnd someone who could

interpret the data and provide the “rest of the story.”

The importance of context has been noted by many researchers. Alavi and
Leidner (2001) note that if the context in which knowledge was created is lost;
the essence of knowledge itself is lost. Similarly, Bartel and Garud (2003)
suggest that the value of narratives in the transmission of knowledge lies in their
ability to integrate context. Seen in this light, the distinction between knowledge
“sources” is more a matter of the amount of context they contain than one of
location. The role of narratives in transmitting context was discussed with the

expert panel, and will be covered in more detail in the next section.

Processing Of K_nowledge Ami/Or Information By Individuals (Narrative Versps

Paradigmatic):

The psychology literature gives us the “two modes of thought” described by
Bruner (1986). He suggested that humans process ideas in two distinct ways -
paradigmatic and narrative — that are not ultimately reconcilable with each other.
As he describes it, the difference between these modes is the “difference
between a logical argument and a good story.” Clear examples of both kinds of
processing were observed in all of the cases, lending credibility to the usefulness

of this distinction.

What was interesting was that knowledge-creating situations seemed to require

some intersection of the two modes. Some examples of attempts at pure

103

bottom-up, data-driven problem solving were observed, but except for the very
simplest situations this was far too laborious. What was needed was some sort
of organizing model or heuristic that could be used to efﬁciently manipulate the
local facts to arrive at a solution. Similarly, a model could not be employed
without taking into account the speciﬁc facts at hand. This closely parallels the
above discussion about the need for apparently explicit knowledge to pass
through a tacit phase in the context of what was, on the surface, an explicit-to-
explicit transformation. It is this combination of modes that creates the difﬁculty

with the transformation view of learning.

The vectors for these models or frameworks correspond to the sources of
knowledge listed in the preceding section and typically consisted of some sort of
story or narrative as described by Bartel and Garud (2003). These narratives
were typically resident in the memories of individuals rather than embedded in
archival material. The key to a good story was its contextual information or
associations that it contained which allowed assessment of its credibility and
relevance to the current situation. This was the point of Pentland (1999), who

suggested that narratives require an “evaluational reference.”

Interestingly, as with fables or parables, there was no particular requirement that
narratives be factually accurate, as long as they captured the essence of a
situation. A good story wouldn’t necessarily tell someone how to do something,
but would illustrate possibilities, suggest areas of relevance and give clues about
where to get more information. In other words, narratives could have low

information density, but high knowledge density.

104

The “two modes of thought” construct appears to be a useful tool in
understanding the creation of organizational knowledge. The only elaboration
needed is the observation that in any learning situation (where something new
was implemented) both modes were always observed to be present to some
degree. Generally the narrative mode tended to dominate when it was available.
When there was a shortage of narratives (as in many of the activities in Case
#2), paradigmatic or bottom-up processing took precedence, but this was

relatively unproductive.

The role of narratives was discussed extensively with the expert panel, and there
was substantial agreement that narratives are an important mechanism for
creating knowledge. A key point raised by the panel was the need for narratives
to have some built-in credibility check, similar to the “evaluational reference”

mentioned above.

Individual Cognition:

Cognition, as a property of an individual, is a broad concept, deﬁned by DeFillippi
and Ornstein (2003) as “thinking, reasoning and memory.” In understanding the
organizational learning process from the point of view of the individual actor, the
role of cognition was selected as important since it is cited in a great deal of work
on organizational learning. It was found in these cases that cognition operated in
two ways, both of which were important, but neither of which was sufﬁcient to
fully understand the process of organizational learning. The two modes that are
of interest to us here are memory, as in existing personal store of knowledge and

comprehension as in the ability to process and absorb new information.

105

First, organizational learning took place in two modes that can be called active
and passive. Most of the activity described in these cases is active learning,
where the teams were charged with achieving a goal for which they had to ﬁrst
seek out a solution and then apply it appropriately. In the passive mode, where
something is being taught to the subject, the issue is comprehension or ability to

absorb the material.

Passive learning in these cases consisted of the tools and principles from the
Kaizen training manual that were presented by the facilitators. For the most part,
cognition was not a limiting factor in this type of Ieaming; the material presented
was not complex and the individuals had sufﬁcient resources to comprehend
what was going on. In Case #2 there was an instance where this was not true,
but the consequences to organizational Ieaming were insigniﬁcant since it was
easy to work around the issue, either by putting extra effort into educating the

individual(s) or by simply bypassing them.

Where cognition did play a signiﬁcant role was in the active mode. This was
observed when trying to catalog where people turned to for knowledge or
information. It is perhaps not surprising that people only consulted sources or
pursued solutions that they already knew about. In this way, cognition appeared
to place a bound on the effectiveness of their search processes. These limits
though, are not based strictly on domain knowledge, but rather on knowledge
about knowledge (as in knowledge of the existence of knowledge or knowledge
about the location of requisite information). This still ﬁts with the deﬁnition of

cognition as it is normally understood.

106

It should be noted that in the discussion with the validation panel, there was
some ambivalence on this point. The majority of the panel agreed with the
proposition that cognitive limits place a bound on search capabilities, but there
were dissenting opinions that people could search beyond those limits if the
event were structured in a way that encouraged them to do so. This suggests
that while this proposition may true for these particular cases, the generalizability

should be considered carefully.

Finally, it was discussed in Chapter Two that there are differing views about
whether cognition is purely a trait of the individual or whether it is situated and
therefore at least partly a property of the environment. What we have considered
in this section is the effect that an individual’s cognition has on the organizational
learning process. The reverse question, which would derive from practice theory,
is the effect that the organizational environment would have on the individual’s
cognition. This could not be observed directly, but some indications can be
inferred in the next two sections.

Power and/or Knowledge As Moderators Of Grodp Activitlﬁ

Many small group characteristics are known to moderate the effectiveness of
group problem solving, and some of them were observed in these cases. In
Case #1, an example was seen where conﬂict avoidance may have led to
incomplete implementation of a solution, although this may have had more to do
with the dominant logic of action present (see below) than simply conﬂict

avoidance.

107

In all of the cases, there were examples where both position and expert power
led to some reduction in the level and breadth of discussion. More speciﬁcally, a
greater differential in positional power seemed to correlate with a reduced usage
of narratives and a corresponding increase in the use of paradigmatic
processing. One possible explanation is that the presence of higher-level team
members suppresses the motivation to introduce any sort of speculative or non-

conforming ideas.

While these inﬂuences may have affected the quality of the ultimate results, the
real issue is whether they altered the how-and-why mechanisms that this
research was designed to discover. It was not apparent that they did, in the
sense that variance between the cases did not seem to result in qualitatively

different patterns of behavior.

Shared Logics Of Act_ion @(istence Of And Changes To):

 

The term “shared logics of action” derives from Scott (2001) and the
cultural/cognitive view of an institution. As previously noted, the use of the word
“shared” introduces certain semantic difﬁculties, and it may be preferable to
speak of “dominant logics” (Prahalad & Bettis, 1986; Bettis & Prahalad, 1995). In
this view, the basis for acceptance or compliance with the logic is a certain
“taken-for-grantedness.” It was noted in Chapter Two that culture provides the
generally accepted “why” rationale for otherwise unanswerable questions (Martin,

2002), which is a substantially similar statement.

On the basis of these observations, the research protocol sought to identify the

existence of these dominant logics and their importance as a factor in the

108

learning process. As the case examples show, this proved to be an important
insight that opens a window into an area that has traditionally been of little
interest to Operations Management researchers. This research suggests that

this is a fruitful area for future work.

A common theme to all of the cases was an attempt to introduce lean production
principles into the companies’ operations. This was done both through the goal
selection for the events and the training materials that were used in the
introductory sessions. This played out differently in the three cases, but all of

them illustrated the key role of the “dominant logic.”

In Case #1, lean production was not the dominant logic. This was evident from
the general discussions, and one of the participants stated it explicitly. Some of
the team members recognized the weakness of the prevailing logic, but felt
powerless to change it. This recalls the institutional theory of Scott (2001), where
compliance to a dominant logic can be enforced in multiple ways. The
cultural/cognitive driver for compliance is the taken-for-grantedness mentioned
above. There are also regulatory drivers (in this case the performance
measurement system, which emphasized productivity measures) and moral or
normative drivers (in a lean production environment, it is appropriate that certain
resources sit idle from time to time, but this is an anathema in a productivity
driven culture). The combination of these mechanisms resulted in strong barriers

to any change to the dominant logic.

The effect of this logic in Case #1 could be seen when proposals were made for

setup time reduction. The reduction of man-hours spent on a setup was fully

109

consistent with a productivity-driven logic, and such proposals were accepted
readily. On the other hand, the reduction of setup duration by having additional
workers perform tasks in parallel was counter to the logic and such ideas were
not adopted. In this way it can be seen that there was indeed a dominant logic of
action and, more importantly, that it deﬁned the limits within which learning could

take place.

Case #2 provided another example. The key recommendation of the team was
to change the paint line schedule to reduce batch size and work in process
(WIP). This was counter to the prevailing logic of operation as articulated by the
Plant Foreman (and as expressed by one of the team members), which involved
large batches and substantial buffers. No part of the recommendation was
consistent with this prevailing logic, so it was left as an open question whether
the change would be implemented effectively or at all. This was another
example of the dominant logic acting to limit the range within which learning

could occur.

Case #3 was a different situation that conﬁrmed the observation. Here, the
dominant logic of the operation embraced the idea of small lot sizes, even to
single piece ﬂow. There were dissenting voices, but it was clear that small
batches were “how things are done around here.” As a result, there was no
resistance to implementing changes that facilitated fast changeovers and small

lot sizes.

These examples highlight the importance of the underlying rationale for behavior.

Several authors have noted the distinction between the “why” and the “what” in

110

an organization. The “why” rationale corresponds closely to the “theory in use”
described by Argyris and Schon (1978), which is the basis for the second loop of
double loop learning. In their deﬁnition of learning (the detection and correction
of error), errors must be corrected in both the what and the why aspects of
behavior. Similarly, Pentland and Feldman (2005) hold that organizational
routines consist of performative and ostensive layers, corresponding to the what
and why respectively and suggesting that routines cannot be properly understood

without understanding both layers.

The point of this relatively lengthy discussion is that the underlying rationale for
behavior (shared logic of action or theory in use or ostensive component) is
important. It is also true that organizations, as institutions, have many ways of
enforcing and perpetuating that rationale. This may explain why no changes to a
dominant logic were observed in the examples cited above — there is a major
cultural component involved, and one Kaizen event with six to eight participants

is insufﬁcient to bring about such a change.

Many researchers have commented on the difﬁculty of effecting changes to the
dominant logic of an organization. Much of this has dealt with the
unchallengeable nature of orthodoxy, characterized as myths (March & Olsen,
1975; Duncan & Weiss, 1979; Hedberg, 1981). Others have focused on the

need to unlearn or discredit what previously passed for knowledge (Weick, 1979;

Penﬂand,1995b)

Since this research did not observe any clear changes to a dominant logic, it is

not possible to describe how such changes can be made, but the sources of

111

difﬁculty can be seen. In particular, a logic is difﬁcult to challenge when it is tacit
- that is to say, when many members of the organization cannot articulate why a
particular logic prevails. Such was the case in Cases #1 and #2. An additional
difﬁculty, noted particularly in Case #1, is that the dominant logic is embedded in
the organization in many ways that must be challenged simultaneously if it is to

be changed.

The nature of this problem was agreed upon by the expert panel. In particular, it
was noted that when a major element of an organization’s logic required change,
that change would have to originate at the top of the organization. A Kaizen
event of the type studied here would not have sufﬁcient reach to make the

change.
Changes To Performative Aspects Of Routines

In all of the cases, changes were made to the outward or performative aspects of
production processes or routines. This was the expected outcome of the events,
so this is not in any way remarkable. The interesting aspects of the changes are
how they came about and their congruence (or lack of it) with the prevailing

“shared logic of action.” Both of these issues have been discussed above.
Changes In Situated Performance

A central premise of this research was that organizational learning would be
veriﬁed by an improvement in situated performance. Kaizen events are attractive
for study in this respect because of their emphasis on data and veriﬁcation of

improvement. A result of this emphasis is a focus on projecting the performance

112

beneﬁts of the changes that are implemented. These beneﬁts tend to be

deterministic (i.e.: easily calculated) and near-term in their effects.

Even so, the cases studied were all still too short in duration to collect true
operating data for full veriﬁcation of a performance improvement. As a result,
reliance is placed on the projections made by the event teams. Because of the
relatively deterministic and transparent nature of the linkage, these projections

give a level of assurance that an actual improvement had been made.

The issue of using performance improvement as a reﬂective indicator of Ieaming
becomes more complex when the learning involves the “logic of action."
Although it was not observed to happen, suppose that the Kaizen event in Case
#1 had created a revelation and caused the company to embrace lean production
philosophy, altering production schedules and changing performance
measurements throughout the organization. How long would it take to see an
improvement in performance, and would it then be possible to retrospectively
attribute any improvement to the speciﬁc instance of learning? Issues of time
lags and causal ambiguity have long plagued strategy research, and may explain
why this aspect of organizational knowledge has been much overlooked in

Operations Management research.

An instance of the type of learning described in the preceding paragraph would
take place at a higher level, a broader scope and over a longer time period than
could be observed within the unit of analysis of the case (the Kaizen event). This
highlights a limitation of the research, one that will not be easily overcome

through experimental design. The point here is that the focus on performance

113

may cause us to overlook certain instances of learning - not because the
deﬁnition is wrong, but because of measurement difﬁculties resulting from time

lags and causal ambiguity.

Autonomous Learning:

Although introduced and deﬁned in Chapter Two, this term was not included in
the original research framework or protocol because it was not expected that this
particular classiﬁcation of the process would be observable in these case
situations. Nonetheless, all of the cases did provide opportunities to observe
examples of this type of learning at work, resulting in the following observations.
The ﬁrst, somewhat obviously, is that autonomous learning requires autonomy.
This means not only the autonomy to experiment with or manipulate the work
process, but also the autonomy to make occasional mistakes. In practical terms,
in a manufacturing setting, this limits autonomous learning to the development of
individual skills at a work task. The second observation is that due to the nature
of manufacturing tasks this type of improvement plateaus very quickly — typically

after a few dozen repetitions.

The signiﬁcance with respect to this research is that autonomous Ieaming can be
used to explain short term adjustment phenomena, but does not appear to be a

signiﬁcant factor in the long term evolution of the performance curve.
Reactive Learning;

This term is used here to mean learning as a response to unexpected events.

This is also called quality based Ieaming, as in learning from mistakes. As with

114

autonomous learning, this was not included in the research protocol because it
could not be observed in process as a consequence of the experimental design.
However, the case studies produced several examples that were evidence of
responses to past events. The reason that these examples were visible is that
they were apparently non-value added activity. As such, they are evidence that
the apparent learning that results from unexpected events may be suboptimal,

even to the point of reducing overall performance.

It is important to note however, that the visible examples are a truncated sample.
More effective solutions, such as root cause elimination, would be largely
invisible in these case studies. As a result, while these observations are
interesting and worthy of additional study, they cannot be generalized at this

point.
Management lnﬂdence:

Within each case, there was a prioritization process whereby it was decided what
problems would be addressed and which ones would be ignored. This process
was fairly tacit and did not involve any signiﬁcant debate about the relative
merits. While the prioritization followed the stated goals of the Kaizen event in a
general way, there was substantial divergence between cases. The interesting
observation is that, in two of the cases at least, this prioritization seemed to
accurately anticipate the focus of management questioning at the end of the
event. (The third case had less management presence and questioning, so was

not necessarily representative.)

115

The question this presents is: does this constitute evidence of a dominant logic?
And, if it does, was it the result of managerial priorities? These questions could
not be answered within the scope of a Kaizen event, but stand as opportunities

for future research.
ﬂmmaw of Findings:

For clarity and convenience, the ﬁndings of the research are summarized in
Table 4. This table follows the content of the research protocol, with the addition

of the additional factors that were extracted from the case notes.

 

Table 4: Summagy of Findings re: Research Protocol

 

 

PROTOCOL KEY OBSERVATIONS
ELEMENT
Knowledge Distinction is clear at the individual level, but less so at the
classiﬁed as organizational. Both elements can be observed at

being explicit or organizational level, but seem to occur together - not
tacit distinct or independent.

 

Transformations It was very difﬁcult to describe any learning activities in
of knowledge these terms. In particular, what should have been the
within and easiest to observe (explicit-to-explicit) proved to be the
between explicit most difﬁcult. In an organizational context, this does not

 

and tacit appear to be a useful terminology.

Knowledge Again, not a particularly useful distinction. At the
classiﬁed by individual level it makes some sense, although all
location knowledge seems to pass through an abstract state in

personal memory. At the organizational level, the key
ﬁnding is that components of knowledge have no location:

 

they are diffused throughout the organization.

 

 

 

116

 

Iaple 4: continued

 

Actions by
individuals to
retrieve or store
knowledge by
location

Knowledge retrieval (to the extent that the term makes
sense) involved generation of a mental model. Preference
was for sources where this had already been done
(consultations with others, narratives). Knowledge was
“stored” incompletely in the form of physical artifacts,
procedures or records, and also in personal memories in

some abstract form.

 

Processing of
knowledge and/or
information by
individuals
(narrative versus
paradigmatic)

Both modes appeared to be necessary simultaneously for
creation of knowledge. In particular, the need for narrative
processing drove the prioritization of the sources
consulted (as in descriptive explanations as opposed to
hard data).

 

Individual
cognition (as in
pre-existing state

Individual cognition was observed to limit the range of
search activities - people only searched for things they
expected to ﬁnd. Individual cognition in the form of

 

 

of knowledge) processing capability was not observed to be a signiﬁcant
barrier to the absorption or implementation of new ideas
once introduced.

Power and/or Examples of hierarchical power and of expert status were

knowledge as
moderators of
group activity

 

observed. The effect appeared to be a reduction in the
use of narratives, but did not change the basic nature of
the activities.

 

117

 

 

Table 4: continued

 

“Shared” or
dominant logics of
action (existence
of and changes
to)

Logics of action were observed to lie on a continuum from
simple and instrumental (easy to test and change) to a
state of orthodoxy (based more on communal belief than
any empirical evidence). No changes to the latter type
were observed in this research; they served to bound the
Ieaming that was possible in these cases.

 

 

 

Changes to All cases had concrete examples of changes made to the

performative way things were done. The level of acceptance and

aspects of implementation appeared to reﬂect the dominant logic.

routines

Changes in All cases had credible projections that the changes made

situated would result in improved performance, but only in the near

performance term. More sweeping strategic changes (not observed)
would have been more difﬁcult to evaluate.

Autonomous Several examples were observed. This appears to be a

Learning short-term adjustment issue rather than a long-term

performance driver.

 

Reactive Learning

Some examples were noted, but this is an incomplete
sample (only suboptimal outcomes were visible). These
observations suggest that the Ieaming that occurs in
response to unexpected events may not be uniformly
beneﬁcial.

 

 

Management
Inﬂuence

 

It was observed that the project teams’ prioritization of
activities seemed to anticipate the direction of
management questioning. The existence of a causal
mechanism (one way or the other) would be interesting,

but could not be derived from these cases.

 

118

 

Part II: Development Of Propositions:

In this section, we consider the patterns noted in the preceding section and
express these as statements about the relationships between the elements of the
research framework. These statements are in the form of propositions that

constitute elements of a theory of organizational learning.

Because this section is following inductive logic, the statements about the
relationships should be ones that would, a priori, have predicted the patterns of
behavior that were observed. Since a theory must not only predict, but explain
as well, these statements must contain some explanation of why the observed
behaviors occurred. The implications of these propositions and their potential for

prediction in other settings will be discussed further in Chapter Five.

The propositions that follow are certainly not the only ones that could be inferred
from the data. Following the advice that a theory should be both useful
(Bacharach, 1989) and parsimonious (Whetten, 1989), attention is focused on
the issues most central to the research question and that have important or
interesting implications. A full discussion of those implications is reserved until
Chapter Five, which ultimately provides the justiﬁcation for focusing on these
particular propositions. The purpose of the present chapter is to show that these
propositions are consistent with observed behavior as repeated in multiple

settings.

While the following propositions were developed within the context of the original
research protocol (Figure 3), certain modiﬁcations were made on the basis of the

research ﬁndings. In particular, any concept of knowledge as a thing to be

119

stored, moved, used, transformed, etc., proved to be problematic in the
organizational context. The relationships summarized in these propositions form
a revised theoretical framework that is illustrated schematically in Figure 6. The
nature of these relationships will be described below, while a discussion of their

signiﬁcance will be given in Chapter Five.

 

~

 

 

 

 

 

 

 

 

 

 

 

 

." KNOWLEDGE "- MODERATORS
I. SOURCES , CognItIon,
\'\,~. _’_,-’ Narrative Processing
Paradigmatic _ .
processing Bases of Compliance:
- Expedience
‘ . lnstrumentality
_______ - Orthodoxy
e'ORGANlZATIONAE‘e I
‘ ‘x KNOWLEDGE ’,' .........
ROUTINES ‘ ----------- ' :e’ LOGIC
(Performative) s, OF ACTION /'
Economic Cultural
Performance Elements

 

 

ORGANIZATION

Figure 6: Schematic of Relationships in Propositions

Tentative Propositions:

1. Organizational Knowledge is aﬁPart Construct:

Starting in the center of the framework diagram, we consider the nature of

organizational knowledge itself. To be consistent with the deﬁnition of

120

organizational knowledge as situated performance, the knowledge must consist
of some performative elements that actually create the performance in question.
It was found, however, that these performative elements always existed in the
context of a theory about how it was that they were expected to result in the
intended performance. Organizational knowledge, then, must consist of both

performative and ostensive components.

This of course exactly mirrors the description of routines put forward by Pentland
and Feldman (2005), which is not a coincidence. There is a very close
association between organizational knowledge and organizational routines, to the
point where they can be considered synonymous. An alternative typology would
be to separate the components and refer to the perfOrmative elements as the
routine and the ostensive part as organizational knowledge. This would appear
to simplify the terminology, but the point of this proposition is that the two
components must be considered jointly, so this report will continue to refer to

organizational knowledge as a composite construct comprising both parts.

The implications of the two-part view of organizational knowledge will be
discussed in more detail in Chapter Five, but there are a number of possible
corollaries. One is that the two components have separate loci in the
organization. The performative aspects can be seen in the physical arrangement
of the work space, in the procedure manuals and other documentary or physical
form. The ostensive aspects are often unspoken and exist primarily in individual
memories. They are diffused through the organization in the form of narratives

and other abstract means. For example, while speciﬁc procedures or

121

performance measurements may be reﬂective of a particular logic, it is the
inference that members of the organization draw from these elements that
constitutes the ostensive component of organizational knowledge (for a related
discussion about inferring the logic behind performance measurements see:

Melnyk, Zsidisin, Luft, Burns, Calantone, Stewart and Hanson (2005).

2. The Ostensive Component ofK_nowledge is an Element of the Culture of the

Organization:

The reason for making this sweeping statement is that the ostensive components
of knowledge span a range from simple and explicit to complex and tacit. This
aligns very closely with the range that Scott (2001) calls the pillars of an
institution. Following his terminology, the prevailing logics span the range from
instnrmental to appropriateness to orthodoxy. Although only the state of
orthodoxy falls under the Cultural-Cognitive pillar in his typology, it is apparent
that if one digs below the surface of even the simplest of logics and asks “why?”
enough times, one will ultimately uncover a state of “taken-for-grantedness” that

signals a cultural element.

Instrumental logics of action tend to be simple, explicit, deterministic and
testable. These are the most visible and the easiest to change, based on
empirical evidence and appeals to lnstrumentality. Theories that constitute
organizational orthodoxy tend to be complex, tacit, non-deterministic and not
testable in the short term. The theories driving larger strategic issues tend to fall

in this category and tend to be very difﬁcult to change because their basis for

122

compliance is the state of “taken-for-grantedness" (Scott, 2001). Different levels

of organizational knowledge form a continuum between these extremes.

A corollary to this view is that beyond some threshold level, organizational
learning increasingly becomes synonymous with culture change. Cultural
change is well known to be a difﬁcult and complex process (Martin, 2002, p.
346), and is not normally considered in the course of Operations Management
research. This point was validated by the expert panel through the comment that
changes to the dominant logic had to be implemented from the top down — it

cannot be done through teaching and persuasion.

3. Organizational Learning Takes Place lndependentlv in tha Two Components:

In a sense, this follows from Proposition 1; recognizing that organizational
knowledge has two components and that those components have different loci
within the organization, it is improbable that change would occur simultaneously
in both elements, and certainly not through the same mechanism. In fact, in
almost all of the examples observed, no change took place to the ostensive
components of knowledge, so learning (in the sense of performative
improvements) was bounded by the existing theory of action. The implications of
this observation will be discussed in more detail in Chapter Five, but this
proposition has signiﬁcant consequences for the imitability of organizational

knowledge and for absorptive capacity.

123

4. Organizational Knowledge is Created by Individuals Using Narrative and

Paradigmatic lnformatﬁn Processing Simultaneodslv:

This addresses the central research question of how knowledge is created. The
term created is used advisedly, considering the content of Proposition 1. Since
organizational knowledge is seen here as the intersection of two independent
components, it is something that is unique in a particular context. This point is
made by Hayek (1945) and by Greeno, Moore et al. (1993). It is therefore more
appropriate to speak of creating knowledge than of applying or reusing other

knowledge.

In all of the examples observed, knowledge creation was the work of individuals
or groups who invariably combined two types of inputs. The ﬁrst was the set of
quantiﬁable facts representing the particular situation at hand. This represents
the paradigmatic processing of information per Bruner (1986). The other
component was some sort of mental image of how things should be rearranged
for maximum beneﬁt. These mental images were personal and abstract. When
necessary they were shared with team members through descriptions and

examples. This is the narrative mode of processing per Bartel and Garud (2003).

This duality mirrors the two-component view of organizational knowledge. The
quantiﬁable performative aspects of the situation can be processed by individuals
in a paradigmatic way, while the ostensive component was processed in
narrative form. The key point is that both components were always seen to be
present: there were always speciﬁc local facts to be dealt with and there was

always some mental model that shaped the approach.

124

A consequence of this proposition is that the creation of organizational
knowledge is bounded, or at least moderated in some way by individual cognition
and by the organizational culture. Both of these acted as ﬁlters on the narrative

mode of importing and processing information.

The role of cognition was seen when participants in the Kaizen events engaged
in search efforts in support of knowledge creation. It was observed that this
search was limited by their existing state of knowledge in the sense that they did
not look for knowledge or information unless they already knew that it existed. It
is not clear that this is a ﬁxed limit: there were differences of opinion among the
expert panelists on this point. Several members felt that individuals could be
prompted to search beyond their cognitive limits if they were suitably encouraged
to do so. Nonetheless, in the context of these Kaizen events, this limiting effect

seemed to be present.

The effect of culture could be seen in the use of narratives to inform team
members about what was “in-bounds” and what was not. Narratives also served
as trial balloons to see what the group would respond to and what would be
quietly ignored. A great deal of ﬁltering took place in this way as the group
simply failed to respond to ideas that did not ﬁt with the dominant logic. This can
be seen as an example of how a culture deﬁnes knowledge by inﬂuencing what

is considered at the level of the individual.

5. Learning is a Function of Expected Utility:

The above propositions have addressed the nature of organizational knowledge

and its creation, but to some extent, these are only moderating factors in the

125

learning process. None of them speak directly to the issue of why organizations
learn in the ﬁrst place. The idea that organizations learn because they expect
some sort of return from the effort is not novel (Levy, 1965; Oi, 1967), but it does
call into question the conventional wisdom of “learning from experience” (Huber,
1991). This proposition was prompted by the observation that none of the
instances of learning in these cases had any identiﬁable direct relationship to

past production experience.

This deals with the issue of causality. There is no argument that experience and
Ieaming are correlated; that has been amply established empirically. What was
noted in this research was a distinct absence of any observable link between
experience in the past and learning in ”the present. If we accept an absence of
causality between experience and learning, the remaining plausible explanation
in the literature is that learning is in anticipation of future experience and its

attendant payback.

The expert panel was generally supportive of this proposition (see Appendix VI),
and there was a comment that learning should be a function of expected utility.
There was, however, a reluctance to let go of the idea that organizations can
learn from experience. This may be the role of reactive learning, which was

brieﬂy discussed above, but is beyond the scope of this research.

Summagy of Chapter Four:

This chapter has presented the results of the case studies and their reduction to
a set of propositions. In the interests of clarity, the actual ﬁeld notes and case

reports are relegated to appendices. The bulk of the chapter is devoted to a

126

review of the research protocol in light of the actual ﬁeld observations. Some
additional considerations are introduced, based on additional ﬁndings from the

cases.

From the observations, a set of tentative propositions was developed. In the
context of this chapter, the key elements of these propositions are that they are
consistent with what was observed, and relevant to the research framework. A
preliminary version of these propositions was presented to an expert panel for
feedback based on extensive industrial experience. The comments and
feedback from this panel were incorporated into the results presented here. The
justiﬁcation for this particular set of propositions will be more fully developed in

Chapter Five, where the implications and conclusions are discussed.

127

CHAPTER FIVE: DISCUSSION AND CONCLUSIONS
Introduction:

The purpose of this chapter is to present a summary view of the ﬁndings, to
discuss their implications and to suggest some potential extensions of this
research that are critical to a fuller understanding of organizational knowledge
and learning. We will begin with an overview of the ﬁndings to highlight what is
novel and interesting about them. Note that “novel” here is not meant to imply
that these observations have never been made before, but rather that they are in

some way inconsistent with accepted wisdom.

These novel ﬁndings will then be used to formulate an answer to the research
question that was posed at the outset: “how do organizations create knowledge
in their operations?” This will be followed by a discussion of the limitations of this
research and suggestions about how the generalizability of the ﬁndings may be
enhanced by additional research. The chapter will close with what is potentially
the greatest contribution: the major areas of extension that are suggested by this

research as necessary components of a more comprehensive theory.

Overview of the Findings:

In this section, we make a number of statements that summarize what is novel or
even surprising about the ﬁndings. While similar observations have been made
individually by other researchers, it is their juxtaposition that suggests the
foundation of a new theory. The outline of such a theory is expressed graphically

in Figure 6 (presented in Chapter Four and reproduced below). The fact that

128

these observations present a challenge to conventional wisdom suggests a
degree of paradox. Poole and Van de Ven (1989) note that a paradox offers an
opportunity for the development of theory and suggest a number of way that this
can be approached. Of the methods they list, the approach here corresponds to
the introduction of a new variable; speciﬁcally the existence of organizational
values — the cultural element of organization knowledge. With that in mind, the

following is an overview of the more interesting ﬁndings:

 

 

 

 

 

 

 

 

 

 

 

 

 

I. SOURCES _ CognItIon,
"\y. I," Narraﬁve Processing
Paradigmatic . _
processing BaSlS of Compliance:
. Expedience
‘ - lnstrumentality
_______ - Orthodoxy
r'ORGANlZATIONAE‘e I
l X‘ KNOWLEDGE ," ..........
ROUTINES ‘ ----------- ' LOGIC
(Performative) 'g OF ACTION ."
Economic Cultural
Performance Elements

 

 

ORGANIZATION

Figure 6: Schematic of Relationships in Propositions

1. To discuss organizational knowledge, the operating routine and its underlying
logic must be considered simultaneously. The foundational work on routines

(based on Cyert & March, 1963) argued that the knowledge of an

129

organization was embedded in the operating routines. Subsequently, others
have suggested that it is the underlying “theme” or “culture” that captures the
knowledge of the organization (Barney, 1986; Miller, 1996). We ﬁnd that it is
not possible to discuss organizational knowledge and learning in terms of one
or the other; they are both necessary and interdependent elements. This is of
course very much the point of Pentland and Feldman (2005), but it does not

appear that the full signiﬁcance of this insight is yet widely appreciated.

There are a number of important implications of this ﬁnding, some of which
are elaborated in the additional propositions, but the essence can be
summarized as follows: Learning, unless it involves changes to the
underlying logic of the organization, will be constrained by that logic. This
provides a solid basis for explaining why organizations differ in their ability to
implement “improved” or “best” practices. Furthermore, the underlying logic
has characteristics that make it a part of the culture of an organization - it is
self-perpetuating and is extremely resistant to change. It is proposed that the
existence of differing logics represents a missing variable that has the

potential to explain much of the observed variance in this process.

It is fairly obvious that attempts to improve performance by focusing on tools
and practices will have little effect on the prevailing logic. This was conﬁrmed
in the research; proposed improvements that were inconsistent with the
prevailing logic were simply ignored. The implication is that any attempts to

achieve large increases in performance must be prepared to address both

130

aspects of organizational knowledge simultaneously, and by very different

means.

The research showed that changes which were seen to be improvements in
the context of the prevailing logic were implemented easily and without
resistance. From this we can speculate that if some learning event were to
result in changes to the prevailing logic, the necessary changes to the
operating routines would follow with minimal difﬁculty. This suggests that the
traditional approach to process improvement should be reversed to de-
emphasize the importance of speciﬁc tools and practices. The organization’s
logic of operation must be addressed ﬁrst. Unfortunately, this research does
not provide any good examples of how this might be done, but it does
suggest the need to pay attention to the cultural nature of organizational

knowledge.

. The process-based view that has been taken by this research has given us
insight into the mechanisms of organizational learning and a basis for
explaining why organizations increasingly become prisoners of past success.
This is the phenomenon noted by Leonard-Barton (1992), Benner and
Tushman (2003) and others. The two-part view of organizational knowledge
gives us an explanation of the reasons for this by showing that, beyond

incremental improvements, Ieaming requires changes to the prevailing logic.

Changes to logic are rarely incremental, so knowledge is not an absolute
state and that one cannot speak simply of greater or lesser knowledge along

a continuum. Instead, there are different states of organizational knowledge,

131

deﬁned primarily by their underlying logic, some of which will ultimately be
proven to be better than others by the test of situated performance. There
are two important consequences of this view, both of which lead directly to
the conclusion that organizational knowledge has a fundamental cultural

element.

The ﬁrst of these is that moving from one state of organizational knowledge to
a (hopefully) better one is not an evolutionary, incremental process, but
requires the absorption of a new logic and practices. More importantly, since
two different theories about how to conduct business cannot coexist, it also
requires the displacement of the previously prevailing logic. This research
suggests that it is ultimately the latter issue that determines an organization’s
ability to learn. The need for unlearning has been noted in the literature
(Weick, 1979; Schein, 1993; Pentland, 1995b), but it is relatively unstudied

and little has been said about how it might take place.

The issue of displacement or unlearning has important consequences for the
concept of Absorptive Capacity. The literature in this area (reviewed in
Chapter Two) focuses almost exclusively on the acquisition and application of
new knowledge, yet this research suggests that the limiting factor is actually
the ability to displace old knowledge. This theme can be seen very clearly in
The lnnovator’s Dilemma (Christensen, 1997) where ﬁrms’ responses to
changing technology'are limited much more by unwillingness to abandon
previously successful approaches than by any difﬁculty in the adoption of new

ones.

132

The second critical consequence of the two-part view of organizational
knowledge stems from the time component of the underlying logic. This logic
typically takes the form of a theory that says, “If we do this today, we will see
the desired result in the future.” The problem is that, except for trivial cases,
the theory is not testable because of the time lags. In other words, the
organization must be persuaded to believe something that cannot be proven.
Returning to an early deﬁnition of knowledge (attributed to Plato),
organizational knowledge can be considered to be a “justiﬁed belief” that
certain courses of action will be fruitful. The justiﬁcation in this setting is
largely a matter of faith. This is precisely the role of culture; to provide the

“why” rationale for othenrvise unanswerable questions (Martin, 2002).

Combining these two consequences, the displacement problem and the
cultural nature of the underlying logic, we can see the true magnitude of the
problem of organizational learning. Having previously asked an organization
to believe and act upon a theory that can neither be proven nor disproven, we
must now ask it to abandon that belief and adopt a new theory that cannot be
proven either. Experience suggests that this only takes place under
conditions of great dissatisfaction with the present state (Schein, 1993) which
may be caused by environmental conditions or precipitated deliberately as a

management tool (Eisenhardt & Martin, 2000).

. The role of the individual must be considered since all of the performance
improvements observed in this research were attributable to the actions of

identiﬁable individuals. The signiﬁcant ﬁnding here is the manner in which

133

these individuals processed information in order to make improvements that
would constitute organizational learning. Each potential learning situation
involved certain established facts that had to be dealt with. The range of
potential solutions springing from the facts at hand was always too great to
permit any sort of exhaustive evaluation, so the approach was guided by

some sort of mental model of how things ought to be done.

These mental models are important because they are the means by which
external sources of information and knowledge (in the individual, personal
sense of knowledge) are imported for application. This means of
transmission corresponds to what is often referred to as narrative processing
(Bruner, 1986; Bartel & Garud, 2003), and determines the preferred sources
of information. Information embedded in stories has high knowledge density
and is readily usable because it is presented in context and allows the
formation of a rich mental model. By contrast, codiﬁed information, as in
tabulated reports, may have high information density but has low knowledge
density because it is relatively devoid of context. All of the subjects of this
research found such sources to be very difﬁcult to use and demonstrated a
clear preference for the richer, narrative forms. This has implications for the
ﬁeld of knowledge management, suggesting that attempting to store
knowledge in codiﬁed form is ineffective, because this loses the context of the
narrative and as a result is rarely consulted. The point about context was
made by Alavi and Leidner (2001) and Bartel and Garud (2003). “Primitive”

societies have always transmitted their knowledge and culture through sagas,

134

legends and fables. The challenge for knowledge management as a ﬁeld is

to ﬁnd modern analogs of these narrative methods.

The role of narrative processing is important to the formulation of the theory
model shown in Figure 6 because it suggests the mechanism by which
organizational culture acts to ﬁlter or moderate the incorporation of external
information and knowledge. Just as multiple external sources are distilled
into a mental model of how things ought to be done, so too are the internal
sources - the multiple examples and stories of “how things are done around
here.” This serves to explain why non-conforming ideas are not so much

rejected as they are quietly ignored.

The reliance on narrative processing of information also suggests how
individual cognition may play a moderating role in the creation of
organizational knowledge. The sources with the highest knowledge density
are those closest to an individual’s own experience. The preference for these
sources and their relative ease of use tends to ﬁlter incoming information and
to restrict the range of search. Speciﬁcally, people tend not to look for things

they do not already know something about.

. The role of experience in organizational Ieaming is equivocal at best. This
statement is based on two points. One is the observation in this research that
none of the proposed process improvements drew in any identiﬁable way on
the company’s prior experience with that process. While many researchers
are careful to avoid the issue, others state fairly explicitly that experience

causes learning (Huber, 1991; Eisenhardt & Martin, 2000). The alternative

135

view, proposed here, is that learning is an act involving conscious effort which

is motivated by expected future utility.

The second point is that processes were observed that appeared to represent
sub-optimal responses to past problems. This would suggest that the
“Ieaming” that takes place in response to (negative) experiences may be
subject to certain biases (such as an ovenrveighting of the probability of
recurrence) and may fail the test of an overall improvement in situated

performance.

There is an established body of empirical research that appears to support
the position that experience causes learning; see speciﬁcally Lieberman
(1984), Huber (1991), Darr, Argote et al. (1995) and Hatch and Mowery ‘
(1998). However, as demonstrated in Appendix VIII, a model of Ieaming
based on expected future utility is also consistent with these empirical results,
and may even explain them better. This is an illustration of the difﬁculty of
inferring a process from the outcomes and serves as validation of the

importance of observing processes directly.

The implications of this conclusion are already well-appreciated in industry
where it is commonplace to demand additional price reductions in exchange
for Iong-terrn contracts. An expected future utility model as proposed here
would predict that costs would be lower at the end of one ﬁve-year contract
than they would be at the end of ﬁve successive One-year contracts, whereas
a model based on cumulative experience would not be able to distinguish

between these cases.

136

The Research Question:

This research began with the question; “How do organizations create knowledge
in their operations?” That question has served as the vehicle for the
development of a tentative theory of organizational learning expressed by the
framework in Figure 6 and the propositions articulated at the end of Chapter
Four. The preceding discussion has focused on the aspects of that theory that
are seen to be interesting, important and in need of further study. Before
progressing to a discussion of the limitations of these results and suggestions for
additional research, it is appropriate to summarize what has been learned and

express it as a direct answer to the research question.

Based on this research and the framework presented in Figure 6, the research

question can be answered in this way:

1. Organizations learn as a conscious act, in the form of an investment of time

or effort that is justiﬁed on the basis of its expected returns.

2. The evaluation of those returns is strongly inﬂuenced by the prevailing logic of
action of the organization. This logic may be simple and transparent, or it
may be complex and tacit. The logic may be considered to be a dimension of

the culture of an organization and reﬂects its beliefs and values.

3. The creation of organizational knowledge is carried out by individuals (alone
or in groups) who combine analysis of the facts at hand with mental models of
how things might be improved. The results are embedded within the

organization in the form of revised work practices or routines.

137

4. These mental models are constructed from a wide range of inputs, both from
within the organization and without. The preferred inputs are in narrative form
and are not readily reduced to codiﬁed form. The range of mental models
considered is constrained by the cognitive bounds of the individuals, but more

importantly, by the culture of the organization.

What remains unexplained in this prototypical theory is the issue of how
organizational culture is changed. The research provided no opportunities to
observe this in action, which may mean that it occurs in major jumps or that it
occurs so slowly and subtly that it is invisible in any short-term study. The former
possibility corresponds to the punctuated equilibrium view of organizational
culture change (Romanelli & Tushman, 1994; Loch & Huberman, 1999; Taylor,
1999; Beugelsdijk, Slangen & Herpen, 2002). While the concept of punctuated
equilibrium is consistent with the above observation that displacement of an
existing theory is only likely to occur under conditions of extreme dissatisfaction
(as created by extraordinary events), we simply do not yet know how

organizations “change their minds” about things, particularly of their own volition.

What we have learned is that there is a knowledge-based explanation for a
punctuated equilibrium model, where major learning involves discontinuous
changes to the prevailing logic of an organization. We have also learned that
Kaizen events are too small and too short to effect such changes in an
organization. At the shop ﬂoor level it is essential but difﬁcult to create an
awareness that the existing approaches are not working well enough, which

appears to be a necessary condition for change.

138

Applicabiligy and Limitations:

Having summarized the ﬁndings, it is appropriate to explicitly acknowledge the
inherent limitations of the research and the range of applicability before
proceeding to recommendations for future research. A primary source of
recommendations for future research will be to extend the generalizability of the
results by addressing their limitations. From the outset, external validity in the
sense of universal applicability was not the objective. Rather, the intent was to
ﬁll a gap in the research literature by observing carefully what real people did in
real situations in order to form the most basic building blocks of a theory of
organizational learning. The issue of internal validity of this research and the
steps taken to ensure it has been discussed in Chapter Three. The purpose of
this section is to review the constraints of the study that might be relaxed in order

to increase the range of applicability of the propositions.

o Constrained Format (Kaizen Events): Kaizen events of the type studied here
are only a subset of the types of events implemented within the subject
corporation. They tend to follow a more or less ﬁxed format, and make use of
a common training manual. This prompted the comment: “don’t the teams
just follow the book?” To an extent they did, for a portion of the event,
although the “book” was modiﬁed for local conditions by the facilitators. Even
within the more tightly deﬁned steps, there was still substantial variety in how
people approached the task. In a discussion of this issue, one member of the

expert panel asked somewhat rhetorically; “Have you ever been to the same

139

event twice?” This was true in these cases as well; the formal aspects of the

events did not appear to impose much constraint on basic behavior.

Nonetheless, to claim generalizability of the ﬁndings, it would be appropriate
to observe comparable events organized according to different formats. ,
There was one particular aspect of the event format that may have been
important and that was the time constraint. There was pressure to have all
the solutions in place by the end of the event. This may have been partly
responsible for the observation that people’s search efforts were tightly
bounded to domains they knew. The availability of more time might
encourage more exploratory search and lead to different conclusions about
the importance of cognitive limits in search. ObserVing the same behavior in

a less constrained environment would result in a much stronger conclusion.

Single Functional Area: Consistent with the Operations Management focus of
this research, the selected cases involved manufacturing operations in
relatively high volume production environments. Two consequences of this
choice are that the processes being addressed are: a) relatively simple in
nature and, b) involve a large number of repetitions. As a result, autonomous
learning effects (introduced in Chapter Two as part of different typologies of
learning) play out very quickly and are insigniﬁcant in the larger picture. This
environment also places a high premium on attributes such as efﬁciency and
reliability — things that may not be as central in other ﬁelds. It is not

suggested that these conclusions should automatically be applied to ﬁelds

140

such as New Product Development where the cycle times are much longer

and innovation may be more highly valued than it is in production operations.

0 deusterEnvironment: The industrial market segment occupied by the subject
companies is characterized by slow clockspeed as deﬁned by Fine (1998),
where changes to the dominant process and product technology are
infrequent. None of the studied companies had experienced any signiﬁcant
product or process technology changes in at least thirty years. Eisenhardt
and Martin (2000) suggest that this may be an important dimension for
understanding the adaptability of the companies’ culture, which was shown
here to be a signiﬁcant obstacle to discontinuous learning. It is entirely
possible that an organization accustomed to more rapid change might behave

differently with respect to cultural adaptations.

A related issue is the economic environment of the business. While all of the
companies studied were experiencing competition and price pressures, their
economic climate was relatively benign. There was no real sense that they
were ﬁghting for their lives. Given the above comments about dissatisfaction
with the present being necessary for certain types of learning, we might
expect to observe different patterns of behavior in settings where the

dissatisfaction is endemic.

0 Group Size: All cases involved small groups of about eight people. Some of
the work was subdivided into teams of two or three, and some was done by
individuals. There were no remarkable differences between these levels, but

it was not possible to observe the functioning of groups larger than this.

141

A key ﬁnding of this research is that the culture of an organization inﬂuences
the mechanisms by which knowledge creation takes place. What was not
observed was any example of how learning might act to change the culture.
Group size might be important here, since a small group would be expected

to exert less inﬂuence on culture than a larger one.

Type of Learning Studied: In the typology of learning (autonomous, induced,
and potentially reactive as proposed above), the cases selected addressed
only induced learning in the sense that they represented a conscious,
planned effort to learn, backed with an investment of time and money. Small
examples of autonomous Ieaming were observed within this setting, but they
were not seen to be instrumental in the larger picture. As a result, no
particular attention was paid to exactly how this took place. In a different
setting, such as a much longer cycle activity where autonomous learning
cannot be dismissed as a short-term transient phenomenon, this could
represent a gap in our understanding of organizational learning. Similarly,
learning resulting from unanticipated results could not be observed — a
necessary consequence of the experimental design. There was evidence
that such learning had taken place, but there was no opportunity to observe
the mechanisms at work. Again, in a highly dynamic and uncertain

environment, this may be a signiﬁcant limitation.

The Deﬁnition of Organizational Knowledge: Throughout this research, a
great deal of attention was paid to developing a working deﬁnition of

organizational knowledge (and, by extension, learning) that was both

142

grounded in the existing literature and at the same time usable for ﬁeld
research. Through the literature review and through consultations with
researchers and industry experts it became clear that knowledge and learning
are highly abstract concepts and that a wide range of meanings is attached to
them. The conclusions reported here are, of course, products of the
deﬁnitions used. To the extent that these deﬁnitions are well supported by
the literature, this is not seen as a limitation per se, but rather as a caution.
Any attempt to compare works in this area must carefully consider the

operating deﬁnitions used before comparing conclusions.

Suggestions for Future Research:

The discussions of the ﬁndings of this research and the limitations that apply to
them suggest a number of avenues for future research. Three particular
suggestions are offered here in the hopes of extending the usefulness and

explanatory power of these ﬁndings.

1. Attempt to replicate the ﬁndings in different settings: There are ﬁve particular
dimensions of the cases that werepotentially limiting. This work may be
usefully extended by selecting new cases that vary on these dimensions. It is
acknowledged that all of them create methodological difﬁculties that must be

overcome, which is perhaps why they represent unstudied areas.

a. Process cycle time: For the most part, the cases studied were
examples of processes with cycle times that could be expressed in
seconds or minutes. Many business processes have cycle times that

can be expressed in weeks, months or even years. Being able to

143

observe the differences between successive iterations of a repeated
task may add an additional dimension to our understanding of

organizational learning.

. Industrial Environment: This research has identiﬁed cultural factors

that inhibit change and learning. It may be possible that there are
other factors not observed here that stimulate and promote Ieaming.
Studying organizations that thrive in situations of high urgency (high

clockspeed or economic uncertainty) might uncover such factors.

. Group size: The project groups studied were all relatively small. What
changes as the groups get larger? This question is thought to be
particularly relevant with respect to the limits imposed by individual
cognition and the ability of the group to inﬂuence the culture of the

larger organization.

. Functional Setting: The research was, by design, restricted to studying
improvements in high volume manufacturing operations. While
innovation was valued in these improvement attempts, it was not
pursued with the same intensity as in, say, a new product design
environment. Having noted the important role of culture in learning, it
is conceivable that different functional areas would have their own
distinct cultures as consequences of the type of work and the
personnel involved. This may in turn have an effect on the

mechanisms of knowledge creation.

144

e. Time span: The events studied were bounded to one week in duration,
which was desirable for the research design. It is easy to suggest
conducting longer-term studies, but this is very difﬁcult to do in
practical situations. Two things that might be expected to differ in
longer term activities would be search behavior and culture change.
More time might permit a wider range of search activities and more use
of codiﬁed information. It is also possible that a longer-term study

might start to see evidence of culture change at work.

2. Observe learning from mistakes and other unexpected events: This research
provided some opportunities to observe what appeared to be learning from
past surprises. Quality-based learning, as this is often called, is generally
held to have a positive effect on performance (Fine, 1988), yet some of the
examples observed seemed to suggest overreaction and equivocal results.
Prospect theory (Kahneman & Tversky, 1979) would suggest that recent
memory of an unwanted event would cause an overestimation of the
probability of recurrence and result in less than optimal responses. The
research question would be, “is there a perceptual bias in the learning that
results from unexpected events?” The methodological difﬁculties of observing
unexpected events are acknowledged, but postmortem studies may be

possible.

3. Values as a missing element in performance models: Throughout the
discussion of the results of this research, we have emphasized the role of

culture in deﬁning how and what an organization learns. However, as a

145

subject for additional research, the term is problematic. As Smircich (1983)
notes: “The culture concept has been bon'owed from anthropology where
there is no consensus on its meaning. ” The term values was used above to
capture the aggregate effect of the shared logics of action that drive an
organization and we propose that this term can be deﬁned as a measurable
construct. It is proposed here that this represents a missing variable in
many attempts to model and explain organizational performance and drawing
attention to this may ultimately be the most important contribution of this

research.

The term values is emotionally laden, but the intent here is to use it in a non-
judgmental way. An organization’s values are not good or bad, nor are they
high or low. They are simply the set of beliefs that drives behavior. Two key
dimensions of values that have been identiﬁed in this research are path
preference (means to an end) and future orientation (degree of discounting of

future returns).

Klassen (2001) showed that a simple proxy for managerial values based on
future orientation could improve the ability to predict performance. Hanson,
Melnyk and Calantone (2003) showed that when a construct was created
from reﬂective indicators of values as path preferences, it was the dominant
predictor of performance in an empirical study. Both of these studies were
restricted to a single area of performance (environmentally responsible
manufacturing) and the challenge is to further deﬁne the construct and extend

these ﬁndings to a more general context. Similarly, Nahm, Vonderembse and

146

Koufteros (2004) developed a construct of customer orientation (a potential
dimension of a broader construct of values) and showed it to be a useful

predictor of performance.

Arguably, values can be measured. Using tools such as conjoint analysis, it
should be possible to develop a preference proﬁle of an organization that
maps the beliefs about the best path for the ﬁrm to pursue (choosing among
characteristics such as: price, quality, service, ﬂexibility, customer orientation,
environmental responsiveness, etc.). For examples and discussions of the
conjoint analysis method, see: Priem (1992), Carroll and Green (1995),
Farber and Griner (2000), Reutterer and Kotzab (2000) and Green, Krieger
and Wind (2001). To obtain a proﬁle of an organization, such questions can
be asked of an informant, but about the organization to minimize social
desirability bias. Such a line of inquiry could identify the dominant logic of an
organization, even if many of the informants personally do not understand it

or agree with it.

The purpose of developing such aconstruct is to improve the explanatory
power of models that predict organizational performance. The results cited

above suggest that this improvement may be signiﬁcant.

. Organizations in transition: We still doinot know how organizations “change
their minds” about the appropriate logic for running their business. This
suggests that companies deliberately seeking to change their strategic
direction would be ideal candidates for study. The value and richness of this

approach is signaled by the importance and need for organizational

147

unlearning that must accompany such a change, as noted by Weick (1979),
Schein (1993) and Pentland (1995b). The difﬁculty associated with this

unlearning process has been observed empirically by Christensen (1997).

More recently, Benner and Tushman (2003) have shown that a focus on
process control and improvement is inconsistent with attempts to pursue
innovation. This work is interesting because it focuses on a problem which
appears to be readily explainable by the ﬁndings of this research. We found
that the knowledge of an organization consists, in part, of a logic about how
certain actions will translate into results. An important consequence of this
ﬁnding is that a ﬁrm cannot have a blending of two differing logics during a

period of transition. For a new one to prevail, the Old one must be displaced.

The key difference between the approach suggested here and that of Benner
and Tushman (2003) (and that of Christensen, 1997, and of the related work
on punctuated equilibrium: Romanelli & Tushman, 1994; Loch & Huberman,
1999; Taylor, 1999; Beugelsdijk, Slangen et al., 2002) lies in the
process/outcome distinction. The work cited in this section is outcome-based,
meaning that explanation must be inferred from an examination of results. In
contrast, the present research has focused on the process of organizational
knowledge creation. By observing employees engaged in the process of
knowledge creation, the sources of difﬁculty became much clearer. Learning
could only take place within the framework of an established logic of action.
When the desired change requires a change to the established logic (as it did

in the cases studied by Benner and Tushman (2003) there must be some

148

means whereby the previous logic comes to be seen as being no longer

tenable.

No such rejection was observed in this research, which is why it is suggested
here to study organizations in transition. One possible framework for
transition would be intended moves from one of the three strategic categories
of Porter (1980) (differentiation, cost leadership or focus) to another, but more
subtle transitions might also be of interest. It is hypothesized that
organizations making a successful transition will exhibit some mechanism
whereby individual decision-makers are induced to abandon a previously-held

belief about how to conduct business in favor of a new one.

Such a mechanism may originate outside the organization (poor performance
in the marketplace, for example) or may be created within it. Eisenhardt and
Martin (2000) allude to a practice of deliberately creating crises in order to
facilitate such a change. Other mechanisms of change may be discovered,
as may various modes and mechanisms of resistance. The critical point is
that such mechanisms cannot be reliably inferred from the observation of
outcomes. To develop a better understanding of the issues of organizational
change and learning, it will be necessary to adopt a process perspective,
which is the central issue in this recommendation for additional research. It is
not sufﬁcient to survey top management about their strategic intent and
direction; it is necessary to understand just what it is that causes behavior to

change at the (ﬁgurative) shop ﬂoor level.

149

Concluding Remarks:

This research set out to answer the question of how organizations create
knowledge in their production operations. The motivation for the study came
from the observation of a pair of problems that have troubled researchers and
managers in Operations Management. Speciﬁcally, researchers have not yet
adequately explained heterogeneity among organizations in their ability to
implement speciﬁc management initiatives. Managers, on the other hand, are
tasked with managing something that they cannot measure - an oxymoron of
sorts. It was felt that direct observation of the learning process would shed

useful light on both of these issues.

A suitable “laboratory” was sought within which to observe organizational
learning in operation, and this was found in the setting of a series of Kaizen
events. While admittedly constrained in scope, these events served to isolate
the process being studied and to control for a number of potentially confounding
factors. A research protocol was constructed from a broad review of the
literature to ensure that the process was observed through a wide range of
theoretical “lenses.” Research followed an inductive approach, in which speciﬁc
observations are aggregated and generalized to a set of propositions that can be
extended to (and tested in) other settings. The logic of explanation that underlies

this generalization is that of pattern matching (Kaplan, 1964; Hunt, 1991).

A set of propositions was put fonrvard in Chapter Four. One of the consequences
of these propositions is a workable answer to the original research question, but

the usefulness of the question extends beyond a simple answer. Speciﬁcally, the

150

research motivated careful attention to the natures of organizational knowledge
and learning. As a result, the propositions as stated in the context of Figure 6 (in
Chapter Four) constitute a prototype theory of organizational learning. Of course
much work remains to be done to establish the validity and usefulness of this

theory and some suggestions are presented in this chapter.

We close this discussion by returning to the original problems that motivated the
research. The researcher’s problem, as originally stated at an operational level,
is why ﬁrms differ in their ability to effectively implement standardized initiatives
such as Total Quality Management (T QM). Recalling the comments of Poole
and Van de Ven (1989), such an inability to explain often suggests a missing
variable. The streams of literature focused on competitive priorities and
cumulative capabilities (see Chapter Two) are attempts to provide the missing

variable.

This research also suggests that there is indeed a missing variable, but that this
variable is a more complex construct reﬂecting the prevailing logic of the
organization about the appropriate way in which to conduct its affairs. This is
similar to the concept of competitive priorities, but is more process focused (how
as opposed to what) and hence more deeply imbedded in the organization.
Nonetheless, it is argued that such a logic (perhaps better described as the

values of the organization) can be measured.

The second, manager’s, problem relates directly to the issue of measurement. It
follows from the belief that “if you can’t measure it, you can’t manage it.”

Knowledge and learning are known to be important and must be managed, yet

151

they cannot be measured, hence the problem. This research does not provide a
complete answer to this problem, but by better understanding the mechanism
involved in organizational learning, it is possible to identify some observable
indicators of the process at work. Particular attention is drawn to the role of
narrative processing as a means by which knowledge density is maintained in an
organizational setting and a contributor to the creation of new organizational

knowledge.

In a more pragmatic view, this research suggests that learning in a small group
setting, such as a manager might control, takes place within the conﬁnes of the
prevailing logic or values of the organization. Again, it is argued that these

values may be measurable, but the more immediate cOntribution may be simply

to draw attention to these constraints and to make them discussable.

Having posited that the values of an organization are an essential component of
organizational knowledge, we must then conclude that there is a level of learning
that involves changes to these values. If we may close with a single normative
statement, we could further argue that the ultimate goal of the manager is not to
shape the organization’s actions to its values, but rather to adapt the values of
the organization to its environment so that the most appropriate actions will
follow. The present research identiﬁed many of the barriers to such change, but
did not allow us to prescribe a solution, so we must consider it to be but one

more step in the larger quest to deﬁne good management.

152

REFERENCES

Adler, P. S. and K. B. Clark (1991). "Behind the Learning Curve: A Sketch of the
Learning Process." Management Science 37(3): 267.

Alavi, M. and D. E. Leidner (2001). "Review: Knowledge management and
knowledge management systems: Conceptual foundations and research issues."

MIS Quarterly 25(1): 107.

Amin, A. and P. Cohendet (2000). "Organisational Learning and Governance
Through Embedded Practices." Journal of Management & Governance 4(1-2):
93.

Amundson, S. D. (1998). "Relationships between theory-driven empirical
research in operations management and other disciplines." Journal of Operations
Management 16(4): 341.

Argote, L. (1999). Organizational Learning: Creating, Retaining and Transferring
Knowledge. Boston/DordrechtlLondon, Kluwer Academic Publishers.

Argote, L., S. L. Beckman and D. Epple (1990). "The Persistence and Transfer of
Learning in Industrial Settings." Management Science 36(2): 140.

Argote, L. and D. Epple (1990). "Learning Curves in Manufacturing." Science
247(4945): 920.

Argyris, C. and D. A. SchOn (1978). Organizational LearninLA Theory of Action
Perspective. Reading, MA, Addison-Wesley Publishing Company.

Austin, R. B. (1996). Measuring and Managing Performance in Organizations.
New York, NY, Dorset House Publishing.

Bacharach, S. B. (1989). "Organizational Theories: Some Criteria for Evaluation."
Acadean oLf Management Review 14(4): 496-515.

Barnard, C. l. (1938). The Functions of the Executive. Cambridge, MA, Harvard
University Press.

 

Barney, J. (1991). "Firm Resources and Sustained Competitive Advantage."
Journal of Management 17(1): 99.

Barney, J. B. (1986). "Organizational Culture: Can it be a Source of Sustained
Competitive Advantage?" Academy of Management Review 11(3): 656-665.

Bartel, C. A. and R. Garud (2003). Narrative Knowledge in Action: Adaptive
Abduction as a Mechanism for Knowledge Creation and Exchange in

153

Organizations. The Blackwell Handpook of Organizational Learning and
Knowledge Management. M. Easterby-Smith and M. A. Lyles. Oxford, Blackwell
Publishing.

Bates, K. A. and E. J. Flynn (1995). "Innovation history and competitive
advantage: A resource-based view analysis of manufacturing technology
innovations." Academy of Management Jodrnal: 235.

Benner, M. J. and M. Tushman (2003). "Exploitation, Exploration and Process
Management: The Productivity Dilemma Revisited." Academy of Management
Review 28(2): 238-256.

Berger, P. L. and T. Luckmann (1966). The social construyction of reality; a
treatise in the sociology of knowledge. Garden City, NY, Doubleday.

Bettis, R. A. and M. A. Hitt (1995). "The new competitive landscape." Strategic
Management Journal 16: 7.

Bettis, R. A. and C. K. Prahalad (1995). "The Dominant Logic: Retrospective and
extension." imtegic Management Jodmalt1986-1998) 16(1): 5.

Bettis, R. A. and S. S. Wong (2003). Dominant Logic, Knowledge Creation, and
Managerial Choice. The Blackwell Haneﬁmk of Organizational Learning and
Knowledge Management. M. Easterby-Smith and M. A. Lyles. Oxford, Blackwell
Publishing.

Beugelsdijk, S., A. Slangen and M. v. Herpen (2002). "Shapes of organizational
change: The case of Heineken Inc." Journal of Organizational Change

Management 1 5(3): 31 1.

Boyer, K. K. and M. W. Lewis (2002). "Competitive priorities: Investigating the
need for trade-offs in operations strategy." Pmction and Operations

Management 11(1): 9.

Brainard, W. and J. Tobin (1968). "Pitfalls in Financial Model-Building." American
Economic Review 58(2): 99-122.

Brown, J. S., A. Collins and P. Duguid (1989). "Situated Cognition and the
Culture of Learning." Educational Researcher 18: 32-42.

Brown, J. S. and P. Duguid (1991). "Organizational Learning and Communities-
of-Practice: Toward a Uniﬁed View of Working, Learning, and Innovation."

Organization Science 2(1): 40.

Bruner, J. S. (1986). Actpal Minds, Possjﬂe World_s. Cambridge, MA, Harvard
University Press.

154

Buch, K. and D. Rivers (2001). "TQM: The role of leadership and culture."
Leadership & Organization Development Journal 22(7/8): 365.

Bullington, S. F., J. Y. Easley, A. G. Greenwood and K. E. Bullington (2002).
"Success Factors in Initiating Versus Maintaining a Quality Improvement

Process." Engineering Management Journal 14(3): 8-14.

Carroll, J. D. and P. E. Green (1995). "Psychometric methods in marketing
research: Part I, conjoint analysis." JMR Journal of Marketing Reseam 32(4):
385.

 

Choi, T. Y. and O. C. Behling (1997). "Top managers and TQM success: One
more look after all these years." The Academy of Managementgxecutive 1 1(1):
37.

Chrisman, J. J., C. W. Hofer and W. R. Boulton (1988). "Toward a System for
Classifying Business Strategies." Academy of Management Review 13(3): 413-
428.

Christensen, C. M. (1997). The Innovator's DilemmLWhen Naw Technolpgies
Cause Great Firms to FaI_il. Boston, MA, Harvard Business School Press.

Clark, K. B. (1996). "Competing Through Manufacturing and the New
Manufacturing Paradigm: Is Manufacturing Strategy Passe?" Production and

Operations Management 5(1): 42.
Coase, R. H. (1937). "The Nature of the F inn." Economica 4: 386-405.

Cohen, M. D. (1991). "Individual Learning and Organizational Routine: Emerging
Connections." Organization Sﬁcience 2(1, Special Issue: Organizational Learning:
Papers in Honor of (and by) James G. March): 135-139.

Cohen, M. D. and P. Bacdayan (1994). "Organizational Routines Are Stored As
Procedural Memory: Evidence from a Laboratory Study." Organization Science
5(4): 554-568.

Cohen, W. M. and D. A. Levinthal (1990). "Absorptive Capacity: A New
Perspective on Learning and Innovation." Administrative Science Quarterly(35):
128-152.

Collis, D. J. and C. A. Montgomery (1998). Corporate Strategy: A Resource-
Based Approach. Boston, MA, lrwinlMcGraw-Hill.

Conner, K. R. and C. K. Prahalad (1996). "A resource-based theory of the ﬁrm:
Knowledge versus opportunism." Organization Science 7(5): 477.

155

Cook, 8. D. N. and D. Yanow (1993). "Culture and Organizational Learning."
Journal of Management Inguig 2(4): 373-90.

Cummings, L. L., G. P. Huber and E. Arendt (1974). "Effects of Size and Spatial
Arrangements on Group Decision-Making." Academy of Managment Journal
17(3): 460.

Cyert, R. M., P. Kumar and J. R. Williams (1993). "Information, Market
Imperfections and Strategy." Strategic Management Journal (1986-1998)
14(SPECIAL ISSUE): 47.

Cyert, R. M. and J. G. March (1963). A Behavioral ThJeorv of the Firm.
Englewood Cliffs NJ, Prentice-Hall.

Dale, B. G., R. J. Boaden, M. Wilcox and R. E. McQuater (1997). "Sustaining
total quality management: what are the key issues?" The TQM Magazine 9(5):
372.

Darr, E. D., L. Argote and D. Epple (1995). "The acquisition, transfer, and
depreciation of knowledge in service organizations: Productivity in franchises."
Management Science 41(11): 1750.

Davenport, T. H. and L. Prusak (1998). Working Knowledge: How Organizations
Manage What They Know. Boston, MA, Harvard Business School Press.

DeCarolis, D. M. and D. L. Deeds (1999). "The impact of stocks and ﬂows of
organizational knowledge on ﬁrm performance: An empirical investigation of the
biotechnology industry." Strategic Management Journal 20(10): 953.

DeFillippi, R. and S. Ornstein (2003). Psychological Perspectives Underlying
Theories of Organizational Learning. The Bl_ackwell Handbook of Organizational
Learning and Knowledge Management. M. Easterby-Smith and M. A. Lyles.
Oxford, Blackwell Publishing.

Deming, W. E. (1986). th of Lhe Crisis. Cambridge MA, Massachusetts Institute
of Technology. Center for Advanced Engineering Study.

Denison, D. R. and A. K. Mishra (1995). "Toward a theory of organizational
culture and effectiveness." Organization Science 6(2): 204.

Dess, G. G. and R. B. J. Robinson (1984). "Measuring Organizational
Performance in the Absence of Objective Measures: The Case of the Privately-
held Firm and Conglomerate Business Unit." Strategic Management Journal (pre-
1986) 5(3): 265.

156

Dewey, J. (1916). Democracy and Education: Ap Introduction to the Philosophy
of Education. London, Collier-MacMillan.

Dewey, J. (1933). How We Think. Chicago, Henry Regnery.

Dewey, J. andA. F. Bentley (1949). Knowing and the Known. Boston, MA,
Beacon Press.

Drucker, P. F. (1993). Post Capitalist Society. New Yourk, Harper Business.

Duncan, R. and A. Weiss (1979). "Organizational Learning: Implications for
Organizational Design." Research in Organizational Behavior 1: 75-123.

Dutton, J. M. and A. Thomas (1984). "Treating Progress Functions as a
Managerial Opportunity." Academy of Management. Tha Academy of
Management Review 9(2): 235.

Easterby-Smith, M. and M. A. Lyles (2003). The Blackwell Hancﬂook of

Organizational Learning and Knowledge Management. Oxford, Blackwell
Publishing.

Eisenhardt, K. M. (1989). "Building Theories from Case Study Research."
Academy of Managemant Review 14(4): 532-550.

Eisenhardt, K. M. and J. A. Martin (2000). "Dynamic Capabilities: What Are
They?" Strategic Management Journal 21 (1 0/1 1): 1105-1 121.

El Sawy, O. A., l. Eriksson, A. Raven and S. Carlsson (2001). "Understanding
shared knowledge creation spaces around business processes: Precursors to
process innovation implementation." lnternaﬁpnal Jod_r_r_Iai of Technology
Management 22(1,2,3): 149.

Ellram, L. M. (1996). "The Use of the Case Study Method in Logistics Research."
Journal of Business Logistics 17(2): p. 93.

Epple, D., L. Argote and R. Devadas (1991). "Organizational Learning Curves: A
Method for Investigating lntra-Plant Transfer of Knowledge Acquired Through
Learning by Doing." Organization Scienca 2(1): 58.

Fahey, L. and L. Prusak (1998). "The eleven deadliest sins of knowledge
management." California Management Reviaw 40(3): 265.

Farber, S. and B. Griner (2000). "Using conjoint analysis to value ecosystem
change." Environmental Science 8 Lechnology 34(8): 1407.

157

Feldman, M. S. (2000). "Organizational Routines as a Source of Continuous
Change." Organization Science 11(6): 611-629.

Feldman, M. S. and B. T. Pentland (2003). "Reconceptualizing organizational
routines as a source of ﬂexibility and change." Adminisiative Science Quarterly
48(1): 94.

Feldman, M. S. and B. T. Pentland (2005). Issues in Empirical Field Studies of
Organizational Routines. Working Paper.

Ferdows, K. and A. De Meyer (1990). "Lasting Improvements in Manufacturing
Performance: In Search of a New Theory." Journal of Operations Management
9(2): 168-184.

Ferdows, K., A. De Meyer, J. Nakane, J. Miller and T. Vollman (1985).
"Manufacturers in US, Europe, Japan Disagree Over What Makes a Winner."
International Management 40(9): 82.

Fine, C. H. (1988). "A Quality Control Model with Learning Effects." Operations
Research 36(3): 437.

Fine, C. H. (1998). Clockspeed. Reading, MA, Perseus Books.

Flynn, B. B. and E. J. Flynn (2004). "An exploratory study of the nature of
cumulative capabilities." Journal of Operations Management 22(5): 439.

Flynn, B. B., S. Sakakibara and R. G. Schroeder (1995). "Relationship Between
JIT and TQM: Practices and Performance." Academy of Management Journal
38(5): 1325-1360.

Flynn, B. B., R. G. Schroeder and S. Sakakibara (1995). "The impact of quality
management practices on performance and competitive advantage." Decision
Sciences 26(5): 659.

Franz, T. M. and J. R. J. Larson (2002). "The impact of experts on information
sharing during group discussion." Small Grodp Research 33(4): 383.

Garud, R. and P. R. Nayyar (1994). "Transformative Capacity: Continual
Structuring By Intertemporal Technology Transfer." Strategic Management

Journal (1986-1998) 15(5): 367.

Glaser, B. and A. Strauss (1967). The Discovery of Grounded Theom: Strategies
of Qualitative Research. London, Wiedenfeld and Nicholson.

Grant, R. M. (1996). "Toward A Knowledge-Based Theory Of The Firm."
Strategic Management Journal (1986-1998) 17(Winter Special Issue): 109.

158

Green, P. E., A. M. Krieger and Y. Wind (2001). "Thirty years of conjoint analysis:
Reﬂections and prospects." Interfaces 31(3): 856.

Greeno, J. G. (1989). Situations, Mental Models and Generative Knowledge.
Complex Information Processingihe lrppact of Herbert A. gmon. D. Klahr and
K. Kotovsky. Hillsdale, NJ, Lawrence Erlbaum Associates: 285-318.

Greeno, J. G., J. L. Moore and D. R. Smith (1993). Transfer of Situated Learning.

Transfer on Trial: Intelligence, Cognition and lnstrdction. D. K. Dettennan and R.
J. Sternberg. Norwood, NJ, Ablex Publishing.

Gruenfeld, D. H., E. A. Mannix, K. Y. Williams and M. A. Neale (1996). "Group
composition and decision making: How member familiarity and information
distribution affect process and performance." Organizational Behavior and
Human Decision Processes 67(1): 1.

Guimaraes, T., M. lgbaria and M.-T. Lu (1992). "The Determinants of DSS
Success: An Integrated Model." BecisieLSCIences 23(2): 409.

Gurvitch, G. (1971). The social frameworks of knowledge. Oxford, Blackwell.

Hambrick, D. C. (1984). "Taxonomic Approaches to Studying Strategy." Journal
of Management 10(1): 27-41.

Hanson, J. D., S. A. Melnyk and R. J. Calantone (2003). "Core Values and
Environmental Management: A Strong Inference Approach." Greener
Managemeknt International In Press.

Harvey, J. B. (1988). The Abilene Paradox and Other Meditations on
Management. Lexington, MA, Lexington Books.

Hatch, N. W. and D. C. Mowery (1998). "Process innovation and learning by
doing in semiconductor manufacturing." Management Scienca 44(11): 1461.

Hayek, F. A. (1945). ’"The Use of Knowledge in Society." American Economic
Review 35: 519-30.

Hayes, R. H. and K. B. Clark (1986). 'Why Some Factories Are More Productive
than Others." Harvard Business Review 64(5): 66.

Hedberg, B. (1981). How Organizations Learn and Unlearn. Handbook of

Organizational Design. P. C. Nystrom and W. H. Starbuck. Oxford, Oxford
University Press: 3-27.

159

Henderson, R. and I. Cockbum (1994). "Measuring Competence? Exploring Firm
Effects in Pharmaceutical Research." Strategic Management Journal (1986-
1998) 15(SPECIAL ISSUE): 63.

Hitt, M. A. and D. R. Ireland (1985). "Corporate Distinctive Competence,
Strategy, Industry and Performance." Strategic Management Journal re-1986
6(3): 273.

Holzner, B. and J. Marx (1979). Knowledge Application: The Knowledge System
in Society. Boston, Allyn-Bacon.

Huber, G. P. (1991). "Organizational Learning: The Contributing Processes and
the Literatures." Organization Science 2(1): 88-115.

Huberman, B. A. (2001). "The Dynamics of Organizational Learning."
Computational and Mathematical Organization Theogr 7(2): 145.

Huff, A. S. and R. K. Reger (1987). "A Review of Strategic Process Research."
Journal of Management 13(2): 211-236.

Hunt, S. D. (1991). Mgern Marketi_ng Theory: Qritical Issues in tha Philosophy of
Marketing Science. Cincinnati, OH, South-Westem Publishing Co.

Hutchins, E. (1995). Cognition iathe Wild. Cambridge, MA, MIT Press.

Hwang, H.-G. and J. Guynes (1994). "The effect of group size on group
performance in computer-supported decision making." Information &
Management 26(4): 189.

lmai, M. (1997). Gemba Kaizen: a commonsense low-cost approach to
management. New York, NY, McGraw-Hill.

Jensen, A. (1992). "Stockout Costs in Distribution Systems for Spare Parts."
lntematjonal Jodrpal of Physical Bistritytion & Logistics Management 22(1): 15.

Kabanoff, B. and G. E. O'Brien (1979). "The Effects of Task Type and
Cooperation Upon Group Products and Performance." Organizational Behavior
and Human Performance 23(2): 163.

Kahneman, D. and A. Tversky (1979). "Prospect Theory: An Analysis of Decision
Under Risk." Econometrica 47(2): 263-292.

Kaplan, A. (1964). The Condpcd of lngdiry, Methodology for Behavioral Science.
San Francisco, CA, Chandler Publishing.

160

Kim, Y. and J. Lee (1993). "Manufacturing strategy and production systems: An
integrated framework." Mmal of Operations Mamement 11(1): 3.

Klassen, R. D. (2001). "Plant-level environmental management orientation: The
inﬂuence of management views and plant characteristics." Production and

Operations Management 10(3): 257.

Kogut, B. and U. Zander (1996). 'What ﬁrms do? Coordination, identity, and
learning." Organization Science 7(5): 502.

Kosslyn, S. M. (1978). Imagery and Internal Representation. Cognition and
Categorization. E. Rosch and B. B. Lloyd. Hillsdale, NJ, Lawrence Erlbaum
Associates: 217-61 .

Krumwiede, D. W. and J. P. Lavelle (2000). "The effect of top-manager
personality on a total quality management environment." Engineering
Management Journal 12(2): 9.

Lave, J. (1988). qunition in Practice: Mind. Mathematics and Culture in
Eveprday Life. Cambridge, Cambridge University Press.

Lave, J. and E. Wenger (1991). Situated Learning: Legitimata Peripheral
Participation. Cambridge, Cambridge University Press.

Leonard-Barton, D. (1992). "Core Capabilities and Core Rigidities: A Paradox in
Managing New Product Development." Stratagic Management Journal
13(SPECIAL ISSUE): 111.

Levitt, B. and J. G. March (1988). "Organizational Learning." Annual Review of
Sociology 14: 319-340.

 

Levy, F. K. (1965). "Adaptation in the Production Process." Management Science
11(6, Series B): B136-B154.

Li, G. and S. Rajagopalan (1998). "Process Improvement, Quality, and Learning
Effects." Management Science 44(1 1): 1517-1532.

Lieberman, M. B. (1984). "The Learning Curve and Pricing in the Chemical
Processing Industries." The Rand Journal of Bconomig 15(2): 213.

Littlepage, G. E. (1991). "Effects of Group Size and Task Characteristics on
Group Performance: A Test of Steiner's Model." Personali’w and Social

Psychology Bulletin 17(4): 449.

Loch, C. H. and B. A. Huberman (1999). "A punctuated-equilibrium model of
technology diffusion." Management Science 45(2): 160.

161

Lyles, M. A. (1988). "Learning Among Joint Venture Sophisticated F irms."
Management International Review 28: 71-87.

Lyles, M. A. and C. R. Schwenk (1992). "Top Management, Strategy and

Organizational Knowledge Structures." Journal of Management Studies 29: 155-
174.

Macdonald, J. (1996). "TQM - does it always work? Some reasons for
disappointment." Managing Service Quality 6(5): 5.

MacDufﬁe, J. P. (1997). "The road to "root cause": Shop-ﬂoor problem-solving at
three auto assembly plants." Management Science 43(4): 479.

Machlup, F. (1980). Knowledge and Knowledge Production. Princeton, NJ,
Princeton University Press.

Mahoney, J. T. and J. R. Pandian (1992). "The Resource-Based View Within the
Conversation of Strategic Management." Strategic Management Jodmal 13(5):
363.

Marble, R. P. (2003). "A system implementation study: Management commitment
to project management." lnfonnatjon & Management 41(1): 111.

March, J. G. (1991). "Exploration and Exploitation in Organizational Learning."
Organization Science 2(1): 71-87.

March, J. G. and J. P. Olsen (1975). "The Uncertainty of the Past: Organizational
Learning Under Ambiguity." European Journal of Political Research 3: 147-171.

March, J. G. and H. A. Simon (1958). Organizations, John Wiley & Sons, Inc.

Martin, J. (2002). Organizational Culture: Mapping the Terrain. Thousand Oaks,
CA, Sage.

McCabe, V. and G. J. Balzano, Eds. (1986). Elem Cognition: An Ecological
Perspective. Hillsdale, NJ, Lawrence Erlbaum Associates.

Melnyk, S. A., R. J. Calantone, F. L. Montabon and R. T. Smith (1998). "Short-
Tenn Action in Pursuit of Long-Term Improvements: Introducing Kaizen Events."
Production and Inventory Management Journal 39(4): 69-76.

Melnyk, S. A. and D. M. Stewart (2000). "Effective Process Improvement:

Developing Poka-Yoke Processes." Production and Inventory Management
Journal 41(4): p. 48.

162

Melnyk, S. A., G. A. Zsidisin, J. Luft, L. A. Burns, R. J. Calantone, D. M. Stewart
and J. D. Hanson (2005). Deﬁning Alignment in Performance Measurement
Systems: A Case Study in Metrics Deployment. Working Paper.

Meredith, J. (1998). "Building operations management theory through case and
ﬁeld research." Journal onperations Management 16(4): 441.

Miles, M. B. and A. M. Huberman (1994). Qualitative Data Analysis: Ap
Expanded Sourcebook. Thousand Oaks, CA, Sage.

Miller, D. (1996). "Conﬁgurations Revisited." Strategic Management Journal
(1986;1998) 17(7): 505.

Mukherjee, A. S., M. A. Lapre and L. N. V. Wassenhove (1998). "Knowledge
driven quality improvement." Management Science 44(11): 835.

Muth, J. F. (1986). "Search Theory and the Manufacturing Progress Function."
Management Science 32(8): 948.

Nahapiet, J. and S. Ghoshal (1998). "Social capital, intellectual capital, and the
organizational advantage." Academy of Management. The Aca_demy of
Management Review 23(2): 242.

Nahm, A. Y., M. A. Vonderembse and X. A. Koufteros (2004). "The Impact of
Organizational Culture on Time-Based Manufacturing and Performance."
Becision Sciences 35(4): 579.

Nelson, R. R. (1991). 'Why do Firms Differ, and How Does it Matter?" Strategic
Management Jodrnal (1986-1998) 12(SPECIAL ISSUE): 61.

Nelson, R. R. and S. G. Winter(1982). An Evolutionary Theory of _E__conomic
Change. Cambridge, MA, Belknap Press.

Nicolini, D. and M. B. Meznar (1995). "The social construction of organizational
learning: Conceptual and practical issues in the ﬁeld." Human Relations 48(7):
727.

Nonaka, I. (1994). "A Dynamic Theory of Organizational Knowledge Creation."
Organization Science 5(1): 14.

Nonaka, l. and H. Takeuchi (1995). The Knowledge-Creating Company. New
York, Oxford University Press.

Nunnally, J. (1994). ﬁychometric Methods. New York, McGraw-Hill.

163

Nwabueze, U. (2001). "An industry betrayed: The case of total quality
management in manufacturing." The TQM Magazine 13(6): 400.

Oi, W. Y. (1967). "The Neoclassical Foundations of Progress Functions." Journal
of Economics 77(307): 579-594.

Okhuysen, G. A. and K. M. Eisenhardt (2002). "Integrating knowledge in groups:
How formal interventions enable ﬂexibility." Organization Science 13(4): 370.

Orlikowski, W. J. (2002). "Knowing in practice: Enacting a collective capability in
distributed organizing." Organization Science 13(3): 249.

Orr, J. (1990). Sharing Knowledge, Celebrating Identity: War Stories and
Community Memory in a Service Culture. Qollective Remembering: Memory in
Sociey. D. S. Middleton and D. Edwards. Beverly Hills, CA, Sage.

Penrose, E. T. (1959). The Theory of the_Growt_h of t_h_e Firm. Oxford, Oxford
University Press.

Pentland, B. T. (1992). "Organizing Moves in Software Support Hot Lines."
Administrative Science Quarterly 37(1992): 527-548.

Pentland, B. T. (1995a). "Grammatical models of organizational processes."
Organization Science 6(5): 541.

Pentland, B. T. (1995b). "Information Systems and Organizational Learning: The
Social Epistemology of Organizational Knowledge Systems." Accting., Mgmt. &
Infoﬁch. 5(1): 1-21.

Pentland, B. T. (1999). "Building process theory with narrative: From description

to explanation." Academy of Management. The Academy of Management
Review 24(4): 711.

Pentland, B. T. and M. S. Feldman (2005). "Organizational Routines as a Unit of
Analysis." ndustrial and Cogporate Change In Press.

Pentland, B. T. and H. H. Rueter (1994). "Organizational routines as grammars of

action." Administrative Science Quarterly 39(3): 484.

Petroni, A. (2002). "Critical factors of MRP implementation in small and medium-

sized ﬁrms." ntemational Journal of Ogrations & Production Management
22(3): 329.

Polanyi, M. (1958). Personal Knowledge: Towards a Post-Critical Philosophy.
Chicago, University of Chicago Press.

164

Poole, M. S. and A. H. Van de Ven (1989). "Using Paradox to Build Management
and Organization Theories." Academy of Management Review 14(4): 562-578.

Porter, M. (1980). Competitive Strategy. New York, NY, Free Press.

Powell, T. C. (1995). "Total Quality Management as Competitive Advantage: A
Review and Empirical Study." Strategic Management Jodrnal 16(1): 15-37.

Prahalad, C. K. and R. A. Bettis (1986). "The Dominant Logic; a New Linkage

Between Diversity and Performance." Strategic Management Journal (1986-
1998) 7(6): 485.

Prahalad, C. K. and G. Hamel (1990). "The Core Competence of the
Corporation." Harvard Business Review 68(3): 79.

Priem, R. L. (1992). "An Application of Metric Conjoint Analysis for the Evaluation
of Top Managers' Individual Strategic Decision Making Processes: A Research
Note." Strategic Management Journal 11986-1 998) 13(SPECIAL ISSUE): 143.

Pylyshyn, Z. W. (1984). Computation and Cognition: Towards a Foundation for
Cpgnitive Science. Cambridge, MA, MIT Press. ,

Reutterer, T. and H. W. Kotzab (2000). "The use of conjoint-analysis for
measuring preferences in supply chain design." Industrial Marketing

Management 29(1): 27.

Rogoff, B. (1990). Apprenticeship in Thinki_ng. New York, Oxford University
Press.

Romanelli, E. and M. L. Tushman (1994). "Organizational transformation as
punctuated equilibrium: An empirical test." Academy of Management Journal
37(5): 1141. -

Rosch, E. (1978). Principles of Categorization. Cognition and Categorization. E.
Rosch and B. B. Lloyd. Hillsdale, NJ, Lawrence Erlbaum Associates.

Rosch, E. and B. B. Lloyd, Eds. (1978). Cpgnition and Categorization. Hillsdale,
NJ, Lawrence Erlbaum Associates.

Rugman, A. M. and A. Verbeke (2002). "Edith Penrose's contribution to the
resource-based view of strategic management." Strategic Management Joprnal
23(8): 769.

Sabherwal, R. and l. Becerra-Fernandez (2003). "An Empirical Study of the

Effect of Knowledge Management Processes at Individual, Group, and
Organizational Levels*." Decision Sciences 34(2): 225.

165

Sakakibara, S., B. B. Flynn, R. G. Schroeder and W. T. Morris (1997). "The
impact of just-in-time manufacturing and its infrastructure on manufacturing
performance." Management Science 43(9): 1246.

Schein, E. H. (1993). "How Can Organizations Learn Faster? The Challenge of
Entering the Green Room." Sloan Management Review 34: 85-92.

Schmenner, R. W. and M. L. Swink (1998). "On Theory in Operations
Management." Journal of Operations Management 17: 97-113.

Scott, W. R. (2001). lnstitptions and Organizations. Thousand Oaks, CA, Sage.

Shaw, R. and T. R. Alley (1985). How To Draw Learning Curves: Their Use and
Justiﬁcation. Issues in the Ecological Study of Learning. T. D. Johnston and A. T.
Pietrewicz. Hillsdale, NJ, Lawrence Erlbaum Associates.

Simon, H. A. (1991). "Bounded Rationality and Organizational Learning."
Organization Science 2(1): 125-134.

Simon, H. A. and C. A. Kaplan (1989). Foundations of Cognitive Science.
Foundations of Cognitive Science. M. I. Posner. Cambridge MA, MIT Press.

Sinclair, G., S. Klepper and W. Cohen (2000). 'What’s experience got to do with
it? Sources of cost reduction in a large specialty chemicals producer."
Management Science 46(1): 28.

Skinner, W. (1966). "Production Under Pressure." Harvard Business Review
44(6): 139-145.

Smircich, L. (1983). "Concepts of Culture and Organizational Analysis."
Administrative Science Quarterly 28(3): 339-358.

Spear, S. and H. K. Bowen (1999). "Decoding the DNA of the Toyota Production
System." Harvard Business Review 77(5): 96.

Spender, J.-C. (1996a). "Making Knowledge the Basis of a Dynamic Theory of
the Firm." Strategic Management Journal (1986-1998) 17(Winter Special Issue):
45.

Spender, J.-C. (1996b). "Organizational knowledge, Ieaming and memory: three
concepts in search of a theory." Journal of Organizational Change Management
9(1): 63.

Spender, J.-C. and R. M. Grant (1996). "Knowledge And The Firm: Overview."
Strategic Management Journal (1 986-1 998) 17(Winter Special Issue): 5.

166

Stake, R. E. (1994). Case Studies. Handbook of Qualitative Research. N. K.
Denzin and Y. S. Lincoln. Thousand Oaks, Sage: 236.

Starbuck, W. H. (1992). "Learning by Knowledge-Intensive F irms." The Journal of
Management Studies 29(6): 713.

Stasser, G. and W. Titus (1987). "Effects of lnforrnation Load and Percentage of
Shared Information on the Dissemination of Unshared Information during Group

Discussion." Journal of Personalig and Social Psychology 53(1): 81.

Stewart, D. M. and S. A. Melnyk (2000). "Effective process improvement
developing poka-yoke processes." Production and Inventory Management
Journal 41 (4): 48.

Straus, S. G. and J. E. McGrath (1994). "Does the medium matter? The
interaction of task type and technology on group performance and member

reactions." Journal of Applied Psychology 79(1): 87.

Strauss, A. and J. Corbin (1990). Basics of Qualitative Research. Newbury Park,
CA, Sage Publications.

Suchman, L. (1987). Plans and Situated Actions: The Proplem of Human -
mchine Interaction. Cambridge, Cambridge University Press.

Sum, C.-C., J. S. K. Ang and L.-N. Yeo (1997). "Contextual elements of critical
success factors in MRP implementation." Proﬁlction aﬂl Inventory Management
Journal 38(3): 77.

Szulanski, G. (1996). "Exploring lntemal Stickiness: lmpediments to the Transfer
of Best Practice Within the Firm." Strategic Management Journal 17(Special
Issue: Knowledge and the Firm): 27-43.

Taylor, S. S. (1999). "Making sense of revolutionary change: differences in
members' stories." Journal of Organizational Change Management 12(6): 524.

Teece, D. J. (1980). "Economies of Scope and the Scope of the Enterprise."
Journal of Econgmic BehaLiW & Organization 1: 223-245.

Teece, D. J., G. Pisano and A. Shuen (1997). "Dynamic Capabilities And
Strategic Management." Strategic Management Jodrnal (1986-1998) 18(7): 509.

Thomas-Hunt, M. C., T. Y. Ogden and M. A. Neale (2003). 'Who's really

sharing? Effects of social and expert status on knowledge exchange within
groups." Management Science 49(4): 464.

167

Tobin, J. (1969). "A General Equilibrium Approach to Monetary Theory." Journal
of Money, Credit and Banking 1(1): 15-29.

Tsoukas, H. (2003). Do We Really Understand Tacit Knowledge? The Bﬁckwell

Handbook of Organizational Learning and Knowledge Management. M.
Easterby-Smith and M. A. Lyles. Oxford, Blackwell Publishing.

Venkatraman, N. (1989). "The Concept Of Fit In Strategy Research: Toward
Verbal And Statistical Correspondence." Academy of Management. The

Academy of Management Review 14(3): 423.

Wacker, J. G. (1998). "A deﬁnition of theory: Research guidelines for different
theory-building research methods in operations management." Journal of

Operations Management 16(4): 361.

Wacker, J. G. (2004). "A theory of formal conceptual deﬁnitions: developing
theory-building measurement instruments." _J_oprnal of Operations Management
22(6): 629.

Walsh, J. P. (1995). "Managerial and organizational cognition: Notes from a trip
down memory lane." Organization Science 6(3): 280.

Ward, A., J. K. Liker, J. J. Cristiano and D. K. I. Sobek (1995). "The Second
Toyota Paradox: How Delaying Decisions Can Make Better Cars Faster." Sloan
Management Review(Spring, 1995): 43-61.

Weick, K. E. (1979). The Social Psycholaqy of Organizing. New York, McGraw-
Hill.

Weldon, M. S. and K. D. Bellinger (1997). "Collective memory: Collaborative and
individual processes in remembering." Journal of Experimental Psychology 23(5):
1 160.

Wenger, E. (1998). _Commmities of chtice: Le_arning, Meaning and Identity.
Cambridge, Cambridge University Press.

Wemerfelt, B. (1984). "A Resource-Based View of the Firm." Strategic
Management Journal 5(2): 171.

Whetten, D. A. (1989). 'What Constitutes a Theoretical Contribution?" Academy
of Management Review 14(4): 490-495.

Williamson, O. E. (1981 ). "The Economics of Organization: The Transaction Cost
Approach." The American Journal of Sociology 87(3): 548-577.

168

Winter, S. G. (1987). Knowledge and Competence as Strategic Assets. EC:
Competitive Challenge: Strategies for Incarstrial Innovation and Renewal. D. J.
Teece. Cambridge, MA, Ballinger.

Womack, J., D. Jones and D. Roos (1991). The Machine That Changed the
World. New York, NY, Harper Perennial.

Wright, T. P. (1936). "Factors Affecting the Cost of Airplanes." Journal of the
Aeronautical Sciences 3: 122-128.

Yetton, P. and P. Bottger (1983). "The Relationships Among Group Size,
Member Ability, Social Decision Schemes, and Performance." Organizational
Behavior and Human Performance 32(2): 145.

Yin, R. K. (1994). Case Study Research Design and Methods. Thousand Oaks,
CA, Sage Publications.

Zahra, S. A. and G. George (2002). "Absorptive capacity: A review,
reconceptualization, and extension." Academy of Management. The Academy of
Management Review 27(2): 185.

Zahra, S. A. and A. P. Nielsen (2002). "Sources of capabilities, integration and
technology commercialization." Strategic Management Journal 23(5): 377.

Zangwill, W. l. and P. B. Kantor (1998). "Toward a theory of continuous
improvement and the learning curve." Management Science 44(7): 910.

169

APPENDIX l

Validation Questionnaire for Expert Panel

 

I have stated the following propositions. On the
basis of your experience, please tell me whether
you agree, disagree or think I am missing
something important. Feel free to use the space
available to tell me why; especially in the last
instance!

Strongly Disagree

Disagree

Neutral / Not Sure

Agree

Strongly Agree

You're Missing The
Point! (see notes)

 

1. a) They may go together, but experience does not cause
learning.
Comments:

 

1. b) In a production environment, learning is an investment and
is evaluated on a forward looking basis.
Comments:

 

2. “Stories" are very important because they capture a rich
range of context and associations that make principles
credible and useful.

Comments:

 

3. Cognition (what a person knows, or knows about) is a
strongly limiting factor in the search for new knowledge (self-
motivated learning).

Comments:

 

4. a) The limiting factor in organizational learning is n_o_t
cognition; it is displacement of what is already “known” (the
competing paradigm).

Comments:

 

4. b) It is important to understand the difference between what
the organization does and My it does it (the logic behind it).
Comments:

 

 

5. Knowledge is not imported intact. It is created by combining
local facts with some organizing framework (almost always
external).

Comments:

 

 

 

 

 

 

 

170

 

APPENDIX lI
Kaizen Event #1 - Dec 14-17. 2004
Introduction:

The company studied in this case is a North American manufacturer of plumbing
products. The company is part of a larger plumbing products group and has a
number of sister plants that manufacture similar products using similar
processes. In spite of being part of a corporate group, the company was
acquired as a standalone business and retains its own unique culture. The
reporting structure also provides a degree of isolation from the business policies

of the sister companies.

The business outlook for the company is generally positive, driven by strong
growth of the home building and home improvement markets. This growth has
been accompanied by consolidation in the customer base, resulting in increased
demands for cost reduction and ﬂexibility of delivery. The process improvement

event studied in this case is aimed directly at addressing these pressures.

The workforce is stable, unionized and well-educated. The work environment
appears to be positive and collaborative. The workforce draws on a pool of Old
World immigrants, contributing to a higher-than-average educational and skill

level, particularly in the skilled trade functions.

The process under review was a semi-automated brazing carousel that could be
conﬁgured to assemble multiple product variants. Because of the pressures

noted above, there was a desire to implement a more “lean” production regimen,

171

requiring more frequent product changeovers. This was not a priority when the
equipment was designed, and product variations have proliferated with the result
that a major changeover required several hours. Consequently, these
changeovers were made on third shift wherever possible, and runs of less than
two shifts were strongly resisted. A stated goal of the project was a 50%

reduction in setup time.

The project team consisted of nine members including an outside facilitator and
the researcher serving as a participant-observer. The balance of the team was a
combination of personnel from the affected area and from other areas in the
operation. The individual primarily responsible for changeovers was part of the
team, which provided expertise but also created certain tensions as changes

were proposed and discussed.

The event uncovered signiﬁcant opportunity to reduce the changeover time, but
the results can only be considered a partial success. The outward evidence for
this statement is that no system was put in place to close the loop by measuring
and tracking changeover times. Some of the reasons for this are discussed in
the analysis of the case transcript. The more subjective reason for this
conclusion is the observation that the full beneﬁts of reduced changeover time
were not recognized by the system. The only beneﬁt that was fully valued was

the direct reduction in labor associated with the changeover.

172

Review of Field Notes In the Context of the Research Protocol:

Knowledge classiﬁed as being explicit or tacit:

Two instances of tacit knowledge were observed in these cases that were part of
the organizational knowledge relevant to the process (as opposed to tacit
knowledge held by individuals). The ﬁrst was the body of knowledge needed to
set up the burners and soldering equipment. The second form of tacit knowledge
was in the form of the shared logic of individuals about the correct way to run a

plant.

Some (primarily the facilitators) subscribed to lean principles and saw setup
reduction as a key enabler. Others, (primarily those most closely associated with
the process), were in the efﬁciency mindset, where setup time reduction was less
important, since the real goal was to reduce the number of setups. Several
participants had no particular point of view either way. The justiﬁcation for calling
this tacit knowledge is that most of the participants were not in a position to fully
state and justify the logic behind it, but accepted it as the way the company

worked. These issues were discussed, but not to any resolution.

The bulk of the activities in this case involved what we would call explicit
knowledge. Seventeen instances were noted where the output of the activity
was an explicit procedure or artifact that represented an increase in
organizational knowledge as previously deﬁned. The antecedents for these
pieces of explicit knowledge were less clear. Most of the activity in the event
involved some sort of conversion of personal knowledge into an organizational

form. This will be discussed more in the next section.

173

Transformapons of knowledge within andpetween explicit and tacit

No transformations involving the above examples of tacit knowledge took place
during the case, although it was evident that such transformations had taken
place in the past. An example would be the transmission of the setup knowledge
to the night shift setup person, presumably through some sort of “over-the-
shoulder” coaching. Transformation of this knowledge to an explicit set of setup

rules would have been an interesting result, but this did not take place.

With respect to the tacit knowledge about “how to run a factory”, no
transformations took place in the sense that the state of knowledge was not
observably different at the end of the project. It was possible to see how, through
being deeply embedded in the organization, the prevailing logic was strongly
resistant to change, even though many in the group could appreciate the logic for
a change. In fact, there was a palpable feeling at a couple of points that setup
reduction efforts had reached the point of diminishing returns and that it was time
to back off. This sort of stress would be unlikely to persist in the long term -
eventually a single logic would prevail. The result in this case was a sort of

compromise - described elsewhere.

The bulk of the activities in this case resulted in what we would call explicit
knowledge, but it proved to be extremely difﬁcult to speak of transmissions or
transformations of explicit knowledge. While the outcomes could be observed in
the form of equipment or process changes, it was not generally possible to
identify a single source or precursor. Of the seventeen instances, only two had

any clear origin. Even those examples (the use of a perforated bucket for cooling

174

water or the use of a quarter-turn nut) seemed to have more than a single
example as an antecedent. At a minimum, there was an issue of adaptation to
the speciﬁc local conditions. What this meant was that the team member
engaged in the transformation was not simply copying a practice, but rather
applying a set of principles that he or she had abstracted from the speciﬁc
example. Events that would appear to be explicit to explicit transformations
seemed actually to be a transformation of multiple explicit sources into some sort
of generalized tacit mental model that was subsequently transformed into a
speciﬁc, explicit form. To usefully describe these events, it seems more useful to
consider the mental processing involved (see below). The concept of
transformations, speciﬁcally explicit to explicit transformations, did not prove to
be a useful way to describe these events.

Knowledge Classiﬁed By Location

From a knowledge management point of view, it is convenient to think about
knowledge as having a place, where it can be stored, retrieved, etc. From that
perspective, the form of knowledge that dominated the event was that which
existed in personal memories. The other forms that came into play during this
case were documented knowledge and knowledge embedded in existing routines

or setups.

Documentary materials observed during this case were: the initial training
materials (manual and videos), the production forecasts by part number, and the
Kaizen event report from a sister plant. Some documentary materials were

created as part of the process, notably the video of the setup process and the

175

data collection to assess the need for repacking after bright dip. For the most
part, these documentary materials consisted of data for temporary use rather
than knowledge. The only real attempts to document learning for future use were
in the training materials (and it was not clear that they had any direct inﬂuence on
the results), and the Kaizen event report. This last item served as a thought
starter, but it was not used as a source of deﬁnitive knowledge. The reason
seemed to be that reports of this kind are lacking in contextual richness, making
it hard for individuals to make use of the learning without substantially retracing

the steps.

Knowledge was embedded in the production process (including the setup
process) that we studied. By observing closely what was done, or by asking
“why?” of a knowledgeable informant, the team members could fairly efﬁciently
reconstruct learning that had gone before. Questions about the cleaning routines
were a good example. Other processes (routines) within the plant were also
available for study, but only one such example was noted. This was the adoption
of the perforated bucket sprinkler for part cooling. Interestingly, to judge from the
story-telling portion of the event, when individuals reached into their memories for
relevant knowledge, what they were often recalling was a routine that had been
observed elsewhere. One the routine was recalled, it could then be dissected for

its relevant knowledge content.

This suggests a ﬁnal comment on the role of memory as a knowledge bin. In
many cases, what was in personal memory was not the ultimate knowledge itself,

but a form of meta-knowledge. If someone didn’t know the answer to a question,

176

chances were good that they had an idea of who would. Likewise, a piece of
knowledge may have been imperfectly remembered, but the individual knew
where to look it up. Unless knowledge had these meta-tags in someone’s head,
it was inaccessible to the group in practical terms. .

ActionaBy lneﬂiﬂdpals To Retrieve Or Store Knowledge By Location

It was clear that everyone’s ﬁrst choice of where to look for knowledge was their
own memories and experiences. All seventeen examples listed involved some
level of personal knowledge. The second favorite choice was to ask someone
else to make use of their personal knowledge. Four of the examples involved
speciﬁcally seeking out someone else’s knowledge (as when SS called a gas
contractor for an interpretation of regulations). There was a willingness to look at
existing routines in order to learn from them. Three examples of this were
apparent (the dual burner concept, the bucket for cooling and the sloping parts
rack), but this was not a preference. Undoubtedly some of what existed in
personal memory was the result of observations of other routines in the past (not
observable by the researcher). There was a strong resistance to using

documentary materials such as production data or written reports.

The reason for this hierarchy of preferences seems to be related to the ease of
extracting what is relevant to the current context. When drawing on memory
(one’s own or someone else’s) it is possible to draw on underlying principles to ﬁt
the knowledge to the need. In contrast, in viewing existing processes or
consulting documentary forms, one is faced with facts lacking in context. The

team found it extremely difﬁcult to use documentary sources without calling in

177

someone who knew the story behind the reports (this was true of the production

forecasts for example).

When it came to storing knowledge, the same hierarchy seemed to apply.
Although it was not possible to measure it, it seems safe to say that most
participants took away more in personal memory than in any other form. The
production processes were changed of course, since that was the purpose of the
event, but there was little or no attempt to document what was done, or more to

the point, why it was done a certain way.

While it was possible to identify “bins” of knowledge by location, it was not so
much the location that was interesting as the type of bin. As the above
discussion suggests, it is the contextual richness that makes knowledge sources

interesting; suggesting a need to consider the processing involved.

Processing Of Knowledge And/Or Information By Individuals (Narrative Versus
Paradigmatic)

 

According to Bruner, 1986, this is the difference between a logical argument and
a good story. There was evidence of both kinds of information processing going
on in this case, but it is not clear that they were as distinct as Bruner suggests
they might be. As far as could be observed, the principal way in which outside
information or knowledge was brought into the case was through stories, or
narratives as they resided in someone’s memory. In these instances, it was the
explanatory value of the stories that made them useful; it was not necessary that
the actual “facts” contained in the stories be strictly true (the Toyota story for

example).

178

There was also clear evidence of bottom-up thinking as in re-arranging the
pieces of a puzzle to get the best solution. This could be seen in the re-design of
various ﬁxtures. What is not so clear is the extent to which the imported
narratives served as triggers to allow more efﬁcient sorting of the alternative
solutions. For example, the idea of re-designing the <rb> ﬁxtures to make use of
burner positions that were common to other parts could be seen as a classic
bottom-up analysis of the situation, but it seemed to have been triggered by the
story about the sister plant using dual sets of burners. While it is possible to
recognize the two forms of processing involved, it seems that they may interact in

important ways.

Individual Cognition (As In Pre-Existing State Of Knowledge)

 

It has been suggested that learning can only take place within the cognitive limits
of individuals, and that this extends in some way to groups. There was no
evidence that this was a limiting factor in the adoption of knowledge this case,
with the possible exception of the fact that most team members were not fully
aware of the implications and potential beneﬁts of setup time reduction. This
may have tempered enthusiasm somewhat, but it certainly did not inhibit the
actual efforts. Of course, it is worth noting that the team members were all
selected on the basis of having some potential to contribute to the exercise, so it

is not surprising that cognitive limits would not be a major factor in this respect.

Where this factor can be seen is in the choice of where to look. As noted above,
team members looked ﬁrst to their own memories. This would certainly suggest

a cognitive limit on the group’s ability to seek out new knowledge. There is no

179

reason to believe that this is an absolute limit, but in the context of this case, the
universe of knowledge to be consulted was more or less what was in the heads
of the project team. An exception would be the cases where someone didn’t
know something, but knew who to ask (for example, the feasibility of the some
the control system changes to the machinery had to be checked with a
maintenance specialist).

Power And/Or Knowledge As Modepaprs Of Gm Activity

It is known in group dynamics that the power and/or inﬂuence of individuals can
inhibit the expression of divergent ideas and skew the decision process. This did
not appear to be a major factor in the actual development of ideas, although the
team members each had their own areas of expertise that were tacitly
recognized by the group. Ideas expressed by someone with this level of

credibility tended to be accepted quite readily.

There was one area of interpersonal relationships that affected the outcome, and
that was the fact that the setup person was part of the team. It was possible that
any suggestions that the team came up with could be construed as a criticism of
this person’s current performance. A desire to avoid conﬂict may have inhibited
the generation of ideas to some extent, and it certainly contributed to the

avoidance of a performance measurement to close the loop.

The existence of these factors may have affected the amount of Ieaming that
took place (the content view), but there was no clear evidence that they changed

the mechanism of learning (process view).

180

Shared Logics Of Action (Eastencegf And Changes To)

The idea of shared logic can be applied at multiple levels of analysis. At the
lowest, simplest levels, the logic tends to be transparent, self-contained and
readily understood. For example, in the case of the revision to the cooling
process, the logic is very basic: cool the part sufﬁciently and uniformly enough for
the operator to remove it from the ﬁxture. The correctness of the logic can be
checked very quickly. Cases like this are not very interesting except that they
provide a contrast to logics that are not simple, span interdependent functions

and cannot be readily tested.

Two examples of these higher-level logics were noted in this case. The ﬁrst was
observed in the prioritization process for the activities to be undertaken. Without
any explicit debate, the team gravitated towards a prioritization that weighted
safety and ergonomic improvements highly relative to the other cases. This
prioritization was mirrored by the questioning of the senior management team
after the presentation. This appears to indicate a shared logic or belief within this
particular company that safety and ergonomic issues are an important
component of success. The “correctness” of this logic is not easily checked, nor
is it very transparent. There was no evidence that the team was even aware that

they were making a choice in this matter.

The second instance was the “lean” versus “efﬁcient” debate. The comments of
many team members indicated that, while they understood (to a degree) the logic
of lean, the dominant belief of the company was that efﬁciency (few setups, long

runs) was the path to success. The team was relatively powerless in this respect

181

because the competing logics are not simple to understand and compare,
cannot be readily tested for correctness, and span multiple functions which must

all act in concert for the logic to be effective.

No changes to either of these logics were noted in this case, which should not be
a surprise. In the ﬁrst instance, there was no evidence that there was any need
for a change. In the second instance, an argument for change could be made,
but a change would require a massive, simultaneous realignment of all aspects
of the production function. This was beyond the scope of the process

improvement event being studied.
Changes To Performatie Aspects Of Rodtines

As described in the transcript of events, the team made numerous changes to
ﬁxtures and methods that represented the results of a learning process. In fact,
from the organization’s point of view, these changes stood as the only results of
the learning, since there was no other record of what was done, or the logic
behind it. Much of this information resided in the heads of team members, but

was not necessarily accessible to the organization.

The actual changes are not interesting from a theoretical perspective; what is
interesting is whether they are in accord or conﬂict with some shared logic.
Following the theme of the previous section, most of the changes were fully
consistent with existing logic, especially at the simpler levels. Even at a higher
level, improvements to safety and ergonomic factors were consistent with the
company’s logic. These improvements were readily implemented and could be

expected to work as planned.

182

The more interesting case was the effort to reduce setup time. Spending less
man-hours doing setups is consistent with an efﬁciency logic, and these changes
were readily implemented. Reducing setup duration, so as to be able to do more
of them, is not consistent with efﬁciency logic and some changes were not
implemented for this reason. For example, ideas were proposed that would have
taken some of the changeover activities off-line so as to reduce the production
downtime. Doing things this way was often less efﬁcient, and may have required
an increase in indirect labor. Such changes did not make it onto the action list.
From a lean production perspective, some of the necessary tools had been put in
place, but it was obvious that they would not be used for their intended purpose,

which was to do more setups.

The key point of this observation is that changes to the performative aspects of
routines can be made, but that is not sufﬁcient. If the changes are to persist and
be used, they must be consistent with the “shared logic of action”. Where this
agreement is absent, it is not fair to claim that organizational knowledge has

been increased in a sustainable way.
Changes In Situated Performance

Related to the above discussion, the test of organizational learning is whether
performance has been improved. In the short duration of a Kaizen event it is
often not possible to conﬁrm this with actual data, so learning must be inferred

from projections.

In the case of the setup time reduction effort, it is reasonable to accept the

team’s projections that fewer man-hours will be required for future setups, so we

183

can say that performance has improved and some learning took place. Realizing
the full beneﬁts of setup time reduction however, requires that the number of
setups be increased to allow a reduction in inventory. There was no evidence
that this would actually happen, so the improvement in performance was less

than it might have been under a different logic.

There were also a number of improvements claimed in the area of safety and
ergonomics. For these, no quantitative measures are available. It may be fair to
claim that organizational knowledge has increased on the basis of face validity -
that is, the resulting state is obviously better than what went before, but in the
context of this case, this claim is not necessary to identify the case as a learning

event
Observations From the Field Notes In Addition to the Research Protocol:

Some observations were made that were not speciﬁcally anticipated by the
research protocol, but were considered interesting in the context of the research

question. These are noted for comparison to other cases.

Stone-Telling:

As the project team became more familiar with each other and with the task at
hand, a story telling phase emerged. This could be seen as an example of the
narrative mode of thought, but there seemed to be more to it than that. Some

possible explanations are:

0 A good faith effort to bring information to the table that might be relevant to

the problem at hand (this would be the narrative mode).

184

o Staking out an area of personal expertise and building credibility for that
position (this would indicate a presence of expert power in the group

dynamic).

0 Testing the bounds of the acceptable solution space by noting the reaction
of others to the stories (this would be a check on the dominant logic of

action).
Evidence of Aptonomous Learning:

Taking note of the distinction between induced and autonomous Ieaming (Fine,
1988; Li & Rajagopalan, 1998), it was not anticipated that it would be possible to
observe autonomous learning taking place. After all, the unit of analysis was
chosen to be an induced learning event. However, such learning was observed
following the changes to the operators’ work station. The operator who was part
of the team wanted to try out the new arrangement to confirm that it was an
improvement. The carousel was set up and approximately 30 parts were
produced before the operator determined that “learning” was complete and that
the result was an improvement. A similar learning process apparently took place

on second shift as well, but this was not observed.

This observation showed that autonomous learning, based on experience, does
take place. It also suggests that the mechanism is an informed trial and error
process, which would be expected to follow the classical learning curve power
function. It was clear that this type of learning could only take place on the part
of an individual, or a work group small enough and independent enough to

autonomously vary their work practices in search of an optimum. It was also

185

clear that in the context of typical high-volume production tasks that this is a
short-lived effect. This is consistent with the literature (Epple, Argote & Devadas,

1991)
_E_vi_dence of Reactive Learning:

In observing a setup procedure, it was noted that the person doing the setup
spent a great deal of time cleaning the mounting areas for the ﬁxtures and their
attachment holes. Upon questioning, it was learned that this was because even
small amounts of contamination in these areas could result in misalignment of
the ﬁxtures sufﬁcient to affect the production process. It became clear that the
operator had implemented this procedure in response to problems that had
occurred in the past. It is somewhat difﬁcult to categorize this type of learning. It
seems to reﬂect quality-based learning (Fine, 1988), but it falls somewhere
between induced and autonomous depending on how problems are resolved.

We will use the term reactive learning to identify this situation.

From the point of view of the project team, these situations were interesting
because they were often non-value-added steps. As such, they were
opportunities for cost reduction if the root cause could be eliminated or
addressed in a more efﬁcient way. From a research perspective, they were
interesting because the result of this type of learning was often the acceptance of
a known added cost in exchange for the avoidance of an uncertain, but probably
larger one. Whether or not this is a beneﬁcial tradeoff depends on an evaluation

of the expected cost of the negative event. There are reasons to believe that a

186

risk-averse agent will over estimate those costs. This raises the intriguing

possibility that quality learning may actually result in a decrease in performance.
What Was Not Bone:

The absence of certain behaviors may be as signiﬁcant as the presence of
others, but is harder to observe. The focus on cognition made it obvious that
many potential knowledge sources were not being consulted. The boundary
tended to align with cognition in the sense that people did not look for solutions
that they did not know existed. It is not clear that this is truly a cognitive limit, or
just a result of the “budget” of time and energy that the team could allocate to

search efforts. It is probably a little of both.

As a corollary to this observation, it became clear after a while that the
company’s cumulative experience in manufacturing these products was not
contributing in any positive way to the search for improvements. This is contrary
to the idea of Ieaming from experience. Where experience did come into play
was in the knowledge of things that would not work - thus making the search

efforts more efﬁcient.

Management Inﬂuence:

A ﬁnal, interesting observation that was not anticipated by the research protocol
was the line of questioning by senior management during the presentation of the
project results. The priority of the questions seemed to mirror the priority that the
team attached to various potential action items. For example, the team seemed

to give a high priority to safety and ergonomic improvements, and the

187

questioning suggested a similar priority on the part of management. This raises
a question of whether there is a causal relationship between the two, and if so, in

what direction.

188

Transcript of Events (Marginal notes are italicized and indented):

Day 1: 7:00 - 3:30

Participants: Facilitator (also mfg eng.) (“FAC”)

Outside facilitator (OS)

Researcher

Tooling expert (“TE”)

Assembly worker (in affected area) (“AW”)
Setup specialist (in affected area) (“SS”)
Warehouse supervisor (“WS”)

Tool and Die specialist (“DM”)

Quality rep. (“QR”)

FAC: intro to Kaizen, key points:

Kaizen events run in parallel with VANE and other Cl initiatives

“Creativity before capital”

“Quick and crude better than slow and elegant”

Intended that participants should learn from events and apply learning to other
areas.

Typical goals: ﬂoor space reduction, labor reduction, reduction in distance
traveled, innovation, set up time reduction, safety improvements

Point from training material: Kaizen is intended to be data-driven

Speciﬁc goals for this event:

50% reduction in setup time

189

- 4 safety/ergo improvements
- 1 poke-yoke implementation
- Plus: investigate elimination of repacking of parts after bright-dip for retesting

(external to the actual brazing operation).

0 In retrospect, the choice of goals was a fairly critical step. Ofﬁcially,
this was billed as a setup reduction event, with the other objectives
seemingly thrown in. It is not clear that the team felt that way; in many
instances they seemed content to pursue any process improvement
that suggested itself. The term “setup reduction” tended to be used
among the group to identify the event in a general way, but there was
no good read on the relative priorities in the minds of the individual
participants. This emerged as an interesting issue as the event
progressed and prioritizations were made. Suggests that in future
studies there should be some questioning about the source of the
objectives, and perhaps a straw poll of the participants’ priorities at the

beginning.

0 Showed video of die change reduction efforts at another company (not part of
the parent corporation) that used a strong analogy to racing pit crews. Ended

with a sense that die change reduction had become an end in itself.

0 This last point made the video a little counterproductive — it
seemed that the participants (in the video) had lost sight of the

real reason for their efforts. Ultimately, this was an issue for our

190

group — how to make improvements in a way that was coherent

with the operating philosophy of the plant.

0 Also showed video of process ﬂow analysis at Mercury Marine.

O

This video had no obvious relevance to setup reduction - major
takeaway was the prevalence of wasted transportation in the
process. Not clear whether the video shaped the mental
models of the participants one way or another, but it does raise
the general issue of framing of the problem: to what extent
does the facilitator inﬂuence the outcome of the event by the

choice of introductory material ?

Illustrative stories told (during training phase, team members used these to

amplify on the points being raised:

0 There was a distinct story-telling mode that emerged here, and re-

appeared throughout the event as members tried to make sense of

certain issues. There were two phases that appeared in sequence,

although they overlapped. The ﬁrst phase was external stories where

participants seemed to be trying to do three things:

0

O

Show themselves to be knowledgeable and credible resources
Import ideas that are ﬁxed in their minds, with the intent that
these might be useful and relevant to the group

Test their own (and the group ’8) understanding of the concepts

involved using examples that are not too close to home.

191

o The second phase of story telling was stories about “how we do things
around here”. The purpose of this phase was less clear, but seemed
to involve the following:

0 Highlight the organizational values and constraints that would
impact the project, but do so in a non-confrontational way (also
a way of educating the inexperienced team members).

0 Test the commitment of the team members to these values and
constraints to develop a shared sense of where the team will
and will not go to generate solutions

TE: example of Toyota downsizing machines to match production volumes,
contrast to own process (keep running excess capacity).
FAC: story of how synchronizing a sub-assembly production to line speed, to
reduce WIP, exposed other ﬂaws in the system.
WS suggested that the idea of using more setups, smaller runs could be
extended to incoming materials.
WS noted problems with the use of efﬁciency as a key metric: leads to
ignoring forecasts and over-producing. This opened discussion on cultural
issues; many agreed that there was a problem of performance measurement,
but there did not seem to be anyone with the will or ability to challenge the
culture.

. This appears to be evidence of a dominant logic or belief within the

organization that is not necessarily universally shared, but inﬂuences

behavior in a strong way

192

WS also noted that this led to push processing

WS, during discussion on 58, noted need to standardize packaging of
incoming material to be able to organize work spaces.

OS told story of die changes at Ford: preheating of injection molds to speed
ramp up.

OS noted the use of the built-in airjacks on the race cars in the video, led to
short discussion of racing practices to speed pit stops (vent lines, brake pad

retractors, # of workers involved).

Assessment: group generally accepted the proposition that reduced setup
time was a good thing, but not everyone believed that the company would be
able to use the beneﬁts appropriately. SS acknowledged upfront that he was
instilled with a culture of high “efﬁciency”, long runs and maximum uptime.
There was a delicate balance, since anything the group came up with could
be construed as a criticism of his work practices. The group trod carefully

around this issue for the whole event.

This last point is fairly key. The story telling process brought out the
picture of a company that didn’t really embrace the concept of “lean”
and one that was not going to be able to make full use of the beneﬁts
of setup time reduction. This set up a (very muted) conﬂict that
persisted throughout the event. On the one hand, there was the
mission orientation that focused on the goal of reducing setups and
wanted to leave no stone unturned. On the other hand, there was a

more pragmatic view wherein “hot swaps” were never going to be part

193

of the company’s operations, with most setups being done on off shifts.
In this view, reducing the amount of work in a setup is still a good
thing, but solutions that would reduce setup duration by moving tasks
“Ofﬂine” without actually reducing the amount of work are not
interesting. At various points in the process it was possible to feel the
tension between these views. In the end, the official view prevailed in
terms of the results presented, but there were a number of the
proposals that the team knew were probably not going to get
implemented, and no mechanism was put in place to ensure that they

would. See the overall comments about the event.

 

Review of Existing Process:

. Reviewing assembled data prior to visiting plant ﬂoor revealed some
uncertainty about what part numbers were active and what the volumes

really were.

0 Even though the group had supposedly complete data on part numbers,
forecasts and historical production totals, the data were somewhat
equivocal and were devoid of some contextual information. To get the
answers we wanted, we needed someone familiar with production to tell
us: “there are three main families of parts needing major setups; we cycle
through the minor setups for the individual part numbers. The volume
splits are about 40/40/20 - you don’t see that from the data because this
family was introduced late this year and those other part numbers are

obsolete now. . .. ” There was more in this vein, but the point is that the

194

group needed an interpretation of the data, not the data itself. The most

efficient way to get that was to ask someone who knew.

0 Team spent about a half hour observing the production process in operation and

poking around the equipment.

0 On returning to the conference room, the team created the following list of issues

with the process (in no particular order):

0

0

Mounting of ﬁxtures could be made easier to change

Front guard too high for operator

Operator was boxed in by parts

Need for some sort of remote control to operate machine from rear
Layer separators in baskets - heavy and in poor condition

Different parts baskets at different heights

Guards had to be open to allow airﬂow, didn’t act as guards

Restricted access around equipment

Cooling water ﬂows back onto table, carrying ﬂux residue

Loose tools scattered around inside equipment

Brazing table too high for operator

Parts did not always release easily - some discussion about this, some
thought this was a problem with uneven cooling, while others thought that
some of the ﬁxtures were damaged.

Steam blows back at operator

Need key switch for gas mixture switch

Adjustments on ﬂuxers too complicated (many fasteners)

195

0 Could move ﬂuxers one station closer to operator (reduce heat problems
and need for keeping guard door open)

0 Solder feed mechanism could be turned vertical to improve routing of lines.

0 General housekeeping issues, clutter

o Anti-freeze leakage from air cylinders with explosive results.

0 Drums of acid ﬂux in un-contained area

0 No clear identiﬁcation of rejects

0 Broken enclosure door

0 Soldering mechanism is “a mess”

0 AW and SS drew their issue list from past experience and made no new
observations during this exercise. The rest of the team had no accumulated

experience; comments were based strictly on observation.
Video of Changeover operation:

The team didn’t have the opportunity to watch a setup “live”, but watched a video
of SS doing a changeover. Detailed discussion of this video was deferred until

Day 2, but some issues emerged:

. Setups are clearly the personal domain of SS; attempts to reduce setup time
will need to be managed in that light. There was a certain resistance to
simplifying or standardizing the work. Apparently, the process had evolved to
its present state to deal with problems that could come up, but would not
necessarily be anticipated by an inexperienced group such as ours. ‘We tried

that before” was heard more than once.

196

All of the adjustments (burner positions, ﬂux position and amount, solder
position and amount, etc) have major components of tacit knowledge. The
specialist knows intuitively how to get them right or close to it, but cannot
easily explain how this is done. There are multiple combinations of settings
that will work acceptably. SS commented that his counterpart on the night
shift who also did setups would do things differently (but obviously in ways
that still worked). This presents a barrier to standardizing the process or to

speeding it up.

0 The setup recipe seems to be the only real example of tacit knowledge in
the process. Had the team been able to capture this knowledge in a setup
procedure or a ﬁxture, it would have represented a transformation (per
Nonaka) of tacit to explicit knowledge. In the end, this did not happen: the
team spent a good deal of time thinking about it, but in the end, there was
some combination of inability and unwillingness on the part of SS to
support this. Also, the team couldn’t think of any way in which doing so
would speed up the changeover process, relative to the “by eye”
procedure currently being used. Allowing the continued use of tacit
knowledge has both advantages and disadvantages. As already noted, it
is fast and cheap. On the other hand, it introduces variance, as between
shifts. There is also a suspicion that if the knowledge were captured
completely, it would be possible to optimize somewhat on use of ﬂux,
solder and gas. There was no enthusiasm on the part of the team for

pursuing this line of investigation.

197

pay 2 - 7:00 - 7:00

Finished watching setup video. SS stated that ﬁne tuning was _alw_aya required -
this was primarily with respect to burner location, but apparently applied to ﬂuxer
and solderer settings too. In fact, watching the operation suggested that the
basic setup procedure was to get the settings close on a ﬁrst pass, knowing that
they could be ﬁne-tuned fairly easily and quickly when the heated parts gave a

visual indication of the correctness of the settings.

Brainstorming on material handling issues - highlighted distinction between direct
and indirect labor costs. Comment by WS re: an additional indirect person
reinforced issue. Clear that the accounting for different types of labor does not

give a clear picture of the cost of setups and may lead to bad decisions.

SS - discussion of burner adjustments, described it as a balancing act between
heat-up and ﬂuxing times. He functions as a walking “expert system” and was

unable to be too speciﬁc about how he arrives at a correct setup.

Brainstorming of quick-attach ﬁxture design. Issue came more from observation
of time spent cleaning mounting positions than from the speed of ﬁxture
attachment itself. SS explained that the extensive cleaning was needed because
otherwise ﬂux residue (or other debris) could cause the ﬁxtures to not seat
properly, resulting in uneven heating or misapplication of ﬂux or solder. The goal
was to make the mounting features less sensitive to contamination from ﬂux
residue, hence proposal to use studs instead of holes. Idea of TE (stud and 1/4

turn nut) was adopted with little debate. Experience applied was generic in

198

nature, not company speciﬁc. Much discussion on details of fasteners, back and
forth between DM and rest of team. TE went to get industrial equipment catalog

to illustrate idea.

0 Note here that the speciﬁc facts (knowledge) needed to design the revised
mounting system were contained in the catalog, however, TE didn’t search
randomly through catalogs to ﬁnd something that would work - clearly he
had a partial image in his mind; enough to go ﬁnd the rest of the details.
This is a recunent theme: the remembered image that somehow “ﬁts” to
the problem and contains enough detail to allow reconstruction of the key

elements.

FAC referred to report of a similar Kaizen at a sister plant to see what they had
done (but did not share this report with the group). Found that they used dual
sets of burners for 4” and 8” parts to avoid changeovers and raised this as a
suggestion. This stimulated discussion about the beneﬁts of the idea. SS was
knowledgeable about the speciﬁc burner components and felt it might be

feasible: action item to check.

0 This other Kaizen event report highlighted the difﬁculties in transfening
“best practices”. Even though the other plant produced nominally the
same parts on a very similar piece of equipment, the team found the
results in the report to be relatively hard to use. Part of the problem is
that the report contained the results, but not the Ieaming process, so the

team members had to “get inside” the process before they could assess

199

the results. Once they did so, they tended to have their own interpretation

of the situation, so it was still hard for them to accept outside ideas.

Brainstorming on improvements to ﬂuxer adjustment mechanisms. DM proposed

a revised mounting design - accepted with little debate.

o It is interesting to consider why this idea was accepted so readily. DM had
a certain degree of status as an opinion leader, but this doesn’t seem to
have silenced critics in other situations. Clearly, no one thought they had a
better idea. It seemed to boil down to this: DM was perceived to be
competent in the area he was discussing, (even if no one else quite
understood what he was proposing), and more importantly, he was offering
to assume responsibility for getting it done, making it a non-issue to other

participants.

SS suggested that ﬁxtures for <rb> assembly be redesigned to reposition part for
maximum commonality with other parts, particularly so the same burners could
be used without replacement. Idea thought to be good, investigation begun. In
retrospect, this was probably the most innovative suggestion for setup time
reduction; appears to have been stimulated by the discussion on dual sets of
burners. Clear that his speciﬁc experience allowed him to form a mental picture
of the operation (one that the team did not observe) in order to apply lessons

from other areas.

0 This is an example of knowledge creation that is very hard to classify, yet
proves to be critically important. It is not really meaningful to talk about this

as a transformation or movement Of knowledge. Certain pieces of

200

knowledge may have come together as triggers, but there is still the
element of synthesis involved. Likewise, it can be considered to be a form
of bottom-up processing of factual data on hand, but there is still the role of
these imported triggers. The best explanation for now seems to be that the
importation of certain knowledge in narrative form broke some sort of

paradigm and allowed for more effective bottom-up processing.

Ideas prioritized and assigned to champions for investigation/rmplementation.

Key activities:

1. <rb> assembly ﬁxture redesign - investigation showed that ﬁxtures could be
repositioned with adapter blocks, no downtime. Material ordered.

2. Fixture mounting to table - downtime required to implement, prototype to be
built, parts and material ordered.

3. Indexer button at rear of machine

4. Controls to cycle ﬂuxer and solderer. This and above: consulted with in-
house electrical expert, found to be easily feasible ($25 worth of key parts).

5. Use of ﬁxed, dual burners - of 20 burners on end bodies that needed to be
changed between 4” and 8” parts, 10 could be made dual, leaving 10 to be
adjusted. (These last 10 are the easiest to adjust.) Estimated cost about
$600 for parts, some of which will come out of backup stock. Estimated setup
savings about 2 minutes. SS thought that there might be a legal problem with
idle burners - consulted outside source (a contractor?) and determined it was

OK

201

o This last point is another example of needing not the facts, but the
interpretation of the facts. This is much more effectively handled by asking
someone who has the interpretation than by looking things up in the book.

6. Modify mounting of ﬂuxer for easier adjustment. After investigation, DM’s

idea was scrapped for a simpler alternative (by him).

7. Tool and ﬁxture organization - changes largely resisted by SS, didn’t see

much beneﬁt, but agreed to make efforts in this direction

8. Improve cleaning water hoses - underway

9. Rotate solder feeder by 90 deg. Deemed to be feasible, sketching of

brackets begun.

10. Heat shield for ﬂux lines (to allow closing door)

11.Modify front guard for easier access - done

12. Parts presentation improvement - 3-person brainstorm on alternatives. OS

described a part presentation table observed elsewhere with sloping top and
chute underneath for disposal of empty cartons - this was accepted as a basic
operating principle. One alternative mocked up and tried for 31 parts by AW,
found to be acceptable. Set up left in place for evaluation by night shift
operators.

0 This was the only real example of a classical Ieaming curve that was

observed. A new conﬁguration was tried, _but it was obvious that it wasn’t
going to be better right away. In this case it took the operator about 30
repetitions to become adapted to the new routine. Consistent with the

literature that “autonomous” Ieaming plays out pretty quickly.

202

13. Investigate removing glass panels in doors to allow them to be closed while

allowing some airﬂow.

0 Running with doors open was a “routine” that had been adopted to forestall
problems with excessive heat build up in cabinet. While this routine did not
have any productivity penalty, it did bypass an intended safety feature.

The Ieaming in this case was to revise this routine through better attention
to the root cause of the problem.

14. Move guard rails and basket storage to make more room around machine for

working. - underway

15. Revise water cooling system - SS replicated a system used elsewhere in

plant (plastic bucket with perforations).

o This is the best example in this case of explicit to explicit knowledge
transformation. Even so, the concept was not imported unchanged.
Rather, a core idea was adapted on the basis of local realities and
seemed to involve an understanding of how the idea worked and under
what conditions.

16. Key switch for gas mixture - reviewed, is feasible, will be done

17. Relocation of ﬂux barrels (safety) - done

Day 3: 7:00 - 4:00

Day spent primarily in individual efforts as each mini-team attempted to complete
their tasks. Activities compared to checklist. The team realized that little

progress could be made on the re-packing issue since little was known about the

203

types and frequency of defects. QR undertook to create a defect tracking sheet
to capture one month's data. Encountered resistance from production supervisor
who didn’t want operators taking any extra time to record defects. Eventually
reached a compromise with supervisor and union reps to allow operators to use
a version of the tracking sheet as their time card on a temporary basis. No action

to be taken until data is collected.

A number of the ideas were found to be ﬂawed in some way:

0 New burners could only be used for half of the ones that needed to be
changed
0 Initial idea for ﬂuxer mounting was dropped and replaced with another

concept (this was done by DM who had proposed the Original ideal and taken
responsibility to make it work).

0 Heat shield idea for ﬂuxer lines was dropped, then reinstated

0 Parts presentation table was an improvement for operator, but created
problems with lift height for re-stocking - reﬁnement of idea ongoing

0 Windows in doors could not simply be removed (safety issue) - eventually
replaced with SS mesh (cut from obsolete parts baskets that were being
converted to ﬁxture storage).

0 Fixture storage ideas mostly abandoned in favor of cutting down some
existing baskets found on the ﬂoor (see above)

0 Dual control found to be somewhat more complicated than ﬁrst thought

but still feasible

204

The group re-convened to assess the impact of the changes on setup time. The
timed list of steps taken from the setup video was used as a baseline. The
validity check on the adjustment (elimination or time change) for each step was
this: did it seem reasonable, and did SS agree that it was valid? No actual
experiments were carried out. The idea of a tracking sheet to record changeover

times surfaced brieﬂy, but the group had no enthusiasm for pursuing it.

Final tally: 49% setup time reduction, 7 safety/ergo improvements, 1 poke-yoke

and data tracking set up to investigate the repack issue.

Day 4: 7:00 - 1:00

Morning spent putting ﬁnishing touches on some items and preparing the
presentation to management. This included a rehearsal of the presentation by all

hands.
Results presented to senior management team.
Key observations here were the questions/comments of the management team:

- Noted that 50% reduction of setup time was a BHAG, and complimented

team. No challenges were made to the results, method of calculation etc.
- Most comments (favorable) revolved around the safety/ergo improvements:

0 Key question: was this just a reaction to what the team presented (the
only objective that was exceeded), or is it reﬂective of management

priorities that inﬂuenced what the team actually worked on?

205

This is an area that needs more study. A good ﬁrst step would be to
better understand the origin of the project objectives and the basis for
prioritization (see introduction). A key question is whether
management values are transmitted in a tacit way to the team
members in such a way as to inﬂuence their choices about priority.
This can be thought of as another body of tacit knowledge inﬂuencing
the project outcome. Even if not all team members share this body of
knowledge about “how things are done around here”, through
prolonged interaction, they will probably arrive at a sort of consensus.

(a tacit to tacit transformation or socialization per Nonaka, 1994).

206

APPENDIX Ill
Kaizen Event #2 - March 14 - 18. 2005

Introduction:

The company being studied in this case is a manufacturer of mechanical
equipment that is sold to both contractors and home owners. The company is a
standalone and has no sister operations. The product line is technically
unsophisticated and margins are low. The lack of sophistication is indicated by
the fact that the company maintains no design staff as such. All design work is
contracted out as needed. The major customers are Lowe’s and Home Depot,
contributing to the pressure on margins. An opportunity existed for a major new
program with one of these customers, but it was known that it would require a
very competitive bid. The market for the products is generally stable, with no
major changes expected. All of these indicators suggest a continuing pressure

on cost efﬁciency.

The workforce is largely transient in the sense of being hired through contracting
agencies. Turnover is very high and many of the workers speak very little
English. Multiple languages are spoken on the shop ﬂoor, so it is not a given that
the workers can communicate with each other either. There is a clear distinction
between employees and contract workers, even though they may work side by
side. Other than readily transportable skills like welding, the company makes an
effort to retain employees with speciﬁc knowledge. It was interesting to note that

corporate policy on Kaizen events is to make a commitment that no one will be

207

let go if their job is eliminated as a result of the event. In this case, that
commitment was not made, ostensibly because of the ﬂuid nature of the
contracted workforce. This may have reduced the inputs from the assembly

workers.

The operation being studied was an assembly cell that produced frames from
steel stock through cutting, bending and welding. The cell had been the subject
of a Kaizen event within the last year, aimed at improving the workﬂow, so no
major improvements were expected in that respect. The issue of concern was
the lumpy production schedule, resulting in high levels of work in process (WIP)
and inefﬁcient use of labor. This was a classic theory of constraints situation

where the cell fed a bottleneck operation, but was pushing product.

The project team consisted of seven members including the outside facilitator
and the researcher. The Director of Operations was part of the team as were
three assembly workers (not from the area under study). This created a clear
power differential in the group and may have suppressed innovation to a degree.
Of the three assembly workers, two were contract employees and one was not.

None of the three spoke English to any useful degree.

The Kaizen event successfully drew attention to the root cause of the problem
and proposed a solution. Full implementation of the solution was beyond the
scope of the event and it is not known if this is being pursued. Some minor

improvements were also made to the production cell itself.

208

Review of Field Notes In the Context of the Research Protocol:

Knowledge Classiﬁe_d As Being Explicit gr Tacit:

Numerous small examples of tacit knowledge were observed in the production
process, the most important probably being in the tube bending area (how to
tweak the result to get it to ﬁt). These really did not impact the changes that the

team made, or vice versa.

As noted in the previous case, there is a second form of tacit knowledge that
resides in the organization and that is the knowledge of the “right” way to do
things. As with many organizations, the “right” way seemed to involve large
batches to maximize efﬁciency and generous use of buffer stock to avoid
shutdowns. This philosophy came out in the question and answer session at the
end of the project, but during the team’s work, one of the assembly workers
expressed concerns about having enough buffer inventory. This is a predictable

view from an hourly employee in a push system.

From the action lists, fourteen examples of explicit results of learning were noted,

with varying degrees of signiﬁcance in the overall picture.
Transforma_tions Of Knowledge Within Angﬁetween Explicit And Tacit

The only observed form of tacit knowledge that was relevant to the Kaizen event
was the knowledge or sense of “how to run the plant". No changes or
transformations were observed that involved this knowledge, but it is possible
that the event itself began a process of transformation. By forcing the Plant

Foreman to articulate some of this knowledge, it is possible that a debate could

209

be started that might result in some transformation, either to an explicit form or to

another form of tacit knowledge.

Of the fourteen noted instances that resulted in some form of explicit knowledge
being put in place, eight had some sort of clear antecedent. These were not
directly observable, but it could be inferred that the participants involved in them
were importing an idea that they had seen elsewhere, relatively unchanged. This
importing involved varying degrees of abstraction. Some ideas, like the use of
the wire tie gun, were simple repetitions of something seen before. Others, like
the pull schedule for welded frames were applications of principles that had been
internalized to some extent - in this case by the outside facilitator. The former
might be described as an example of explicit to explicit transformation, while the
latter could be seen as a tacit to explicit transformation. It is difﬁcult to say where
one mode ends and the other starts, and it is even more difﬁcult to see any value
in the distinction. It seems fairer to say that every result representing some form
of organizational knowledge involved both tacit and explicit antecedents, in
varying proportion. The six items for which clear antecedents were not visible
were all apparently examples of Ieaming by “rearranging pieces of the puzzle”,
which is hard to classify as a transformation. The processing model (below) is a

more descriptive tool in these cases.

Knowledge Classiﬁed By Location

From the point of view of getting or placing knowledge in some location, there
were three main “bins”. The most-consulted form of knowledge was what

individuals carried around in their heads. The ﬁrst reaction was almost always to

210

dig into one’s own memory or to ask somebody. Whether appropriately or not,
this method has the advantage of ﬁltering out unnecessary detail and adding an
interpretive layer. As in Case I, it wasn’t the speciﬁc knowledge itself that was so
important, but rather the meta-knowledge. The team member had to know that a
certain type of solution probably existed, and they had to know who to ask or

where to look it up.

The second source of knowledge was the process as it was currently
implemented. The main value-adding steps were generally clear enough, but
there were also a myriad of small time-saving details in the process, as well as
non-value adding steps that were there for the purpose of avoiding certain types
of trouble. These last often became obvious when the team proposed
elimination of apparent waste but was then infomned about the reason for the
practice. This embedded knowledge effectively served as an additional set of

constraints on the improvement process.

The ﬁnal source was the collected documentation about the process and the
production schedule. Most of the actual process documentation came from a
previous Kaizen event and a value stream mapping exercise. As noted in the
transcript, the earlier report was not shared with the team, only consulted by
FAC. As also noted, the documentary evidence was only useful to a point, and

usually had to be augmented by the interpretation of a knowledgeable person.

The hierarchy of preference was clear. People ﬁrst scan their own memories for
either a solution or a pointer to where a solution might be found. The latter

alternative often led to consulting others who might have relevant expertise.

211

Presumably this also led to actually looking something up (documentary
evidence). This was observed on a couple of occasions (looking something up in
a catalog or consulting a production schedule). This was the source of some of
the simple, explicit details, but anything that involved fundamental principles,

such as the re-scheduling of the paint oven, depended on personal memories.

Actions By Individuals To Retrieve Or Store Knowledge By Location

Retrieval of knowledge appeared to follow the previously-observed hierarchy:
search one’s own memory, ask someone, look at the process, and as a last
resort, look up archival information. The disadvantages of the last two sources
were apparent in this case as the team was repeatedly frustrated in trying to use
them. The problem was essentially that it was difﬁcult to get the whole story, and
difﬁcult to sort the relevant from the irrelevant. Consulting individual memory
(one’s own or someone else’s) provides a sense of context and completeness

that allows for more conﬁdence.

Storage of knowledge was typical. The processes were changed in certain ways,
and that stood as the record of whatever Ieaming took place. The participants
walked away from the event with whatever they remembered. No one took any

notes or documented any form of why things were done.

Processin Of KnowledgeAng/Or Information By Individuals (Narrative Versus
Paradigmatic)

The majority of the effort and activity observed in this particular case was the

bottom-up kind of processing as the team members struggled to rearrange the

212

pieces of the puzzle in a logical way. In this case there was very little evidence
of the importation of more abstract principles and concepts that would lead to

top—down analysis.

It is not always easy to tell which mode is operating. For example, one of the key
outcomes of the event was a different schedule for the paint line. On the face of
it, this looks like the outcome of bottom-up analysis, but RES made the comment
very early in the process that this problem looked like more of a paint line
scheduling problem than a weld booth rearrangement problem. Was this
observation accepted at face value, did it inﬂuence the search process for new
solutions, or was it just coincidental? The observations don’t give us a clear
answer to this, but there does appear to be a level of interaction between the

narrative and the analytical modes of processing.

When viewed through the lens of tacit and explicit knowledge and their
transformations, it seemed to be the case that there was no such thing as a pure
explicit to explicit transformation — there was always a tacit component. Looking
now through the lens of narrative versus paradigmatic processing, it seems that
any learning outcome involved elements of both, albeit in varying proportions. All
fourteen instances of knowledge implementation involved some synthesis of the
facts at hand (paradigmatic processing) combined with some organizing principle

imported in narrative (not strictly factual) form.
Individual Cognition JAs In Pre-ExistingState Of Knowledge)
In this case, some severe cognitive limits were observed on the part of individual

team members, because of language and potentially cultural barriers. This

213

hampered the working of the team, but was not fatal. As long as there was some
core set of members with the necessary skills, the work could progress. An

individual who lacked the cognitive base could be educated, or simply bypassed.

That said, there was still an issue with the adoption of lean manufacturing
principles. In the presentation, the Plant Foreman showed that he understood
the principles well enough; he just didn’t believe that they were the best way to
run a plant, given their particular circumstances. There were no obvious
cognitive barriers to implementation of lean production techniques. The problem
was that there was an existing model of how to best run the plant, and until that

model could be shown to be wrong, no alternative would be adopted.
Power And/Or Knowlegge As Moderators Of Group Activiy

Power relationships probably had a signiﬁcant effect on the behavior observed.
As noted in the transcript, the operators likely saw themselves as lower-level
players based on their employment status and language barriers. It is not
surprising that they would say as little as possible. In a similar vein, QC was
relatively new on the job and was also unlikely to challenge anything or anyone.
All of the inside team members tended to defer to DIR. This imbalance made a
much weaker team than might othenrvise have been the case, and likely reduced
the amount of learning that took place. As with the previous cases, there was no
reason to believe that this effect changed the basic mechanism by which learning

took place.

214

Shared Logics Of Actipn (Existence Of And Changes To)

As with the previous case, it is the higher-level logics that are interesting. There
was clearly a prevailing logic through this plant that emphasized batch mode
production and large buffer stocks. This appears to have been the result of each
area trying to optimize its performance locally rather than any sort of plant-wide
policy. This may also have been driven by the high priority placed by the GM on
productivity. While he apparently believed that lean principles would lead to

productivity gains; that belief had not permeated the organization.

Often, this is the result of a performance measurement system that measures
productivity at each activity level - forcing the local optimization. Questioning
revealed that performance measurement at the work area level was, in fact,
productivity-based. Given this, large batch “efﬁcient” production is in fact the
best way to run an area. No change was observed in the dominant logic during,

or as a result of this event.
Changes To Performative Aspects Of Rogtines

Changes to the outward or performative aspects of work routines were an
important part of this and all Kaizen events. As noted above, fourteen instances
of changes to routines were noted. These could be grouped into three major
areas. The ﬁrst was the layout of the equipment to streamline the work ﬂow.

The effect of this was not quantiﬁed, but should result in some productivity gains.

The second change was to introduce a standard work process into the machining

steps using small transfer batches as a way to approach one piece ﬂow. During

215

initial trials, this did not result in increased productivity, but a certain amount of
autonomous learning had to take place before that determination could be made.

It was unknown whether this change would be sustained.

The third change was to the scheduling of the paint line, where batch sizes were
to be reduced from a day’s demand to an hour’s. lf implemented, this will reduce
work in process substantially. Again, it was unknown as of the close of the event

whether this change would be fully implemented and sustained.
Changes In Situateg Performgp0e

As suggested by the previous section, no veriﬁcation of improvement was
obtained during the course of the event, so it is necessary to rely on the
projections of the team for improved productivity and reduced work-in-process.
This case is similar to the ﬁrst; the tools for improvement were implemented, but
unless there is a corresponding shift in the shared logic of action (to run the plant
more according to lean principles) it is unlikely that the full beneﬁts will be

realized or sustained.
Observations From the Field Notes In Addition to the Research Protocol:

Some observations were made that were not speciﬁcally anticipated by the
research protocol, but were considered interesting in the context of the research

question. These are noted for comparison to other cases.

StoLy-Telling:

It was noted in Case I that a story-telling phase occurred as the team started to

come to grips with the problem. In this case, attempts were made by a couple of

216

participants to start such session, but it was not picked up by the others and it
died out fairly quickly. This seemed to be in some way reﬂective of the problems

of team dynamics that persisted throughout the event.
Evidence of Autonomous Learning:

As with the ﬁrst case, there were opportunities to observe autonomous learning
when revised procedures were tried out. Three speciﬁc instances were
observed; one was the increased changeover frequency on the paint line and the
other two were in the machining area where small batch processing was
implemented. In the case of the paint line, four changeovers were performed in a
four hour period, which the operators felt was sufﬁcient to establish the routine.

In the case of the machining operations, approximately ﬁve batches were run of
each process. This was not enough to achieve the best possible results, but was

felt by the operators to be close enough for evaluation purposes.

It was clear that the operators would have little tolerance for routines that would
require more than one shift to optimize. This reinforces the ﬁnding that
autonomous learning by doing is a short term phenomenon, at least in a high
volume manufacturing context. This form of learning curve is best seen in the
reverse sense; as a period of inefﬁciency following the implementation of a

supposedly improved routine.

Evidence of Reactive Learning:

In the analysis of the process ﬂow for standard work determination, a number of

non value adding steps were observed. Actually, they were seen during informal

217

observations; when the operators knew they were being timed, they tended to
omit these steps. Speciﬁc examples were a slug clearing step in a hole drilling
process and two different deburring operations. None of these steps were strictly
necessary; rather they prevented potential problems in later operations. They
had apparently been instituted by the operators, which perhaps explains why

they tended to be omitted when the operators felt they were being watched.

The interesting question is whether these were cost-effective solutions to the
problems. It did not appear that the operators had the resources to do a root
cause analysis, or any cost-beneﬁt analysis. The pattern emerging here is that
production processes tend to accumulate an overhead of insurance steps to
protect against events that had caused problems in the past but that may, or may
not, occur again. This suggests that the role of reactive learning may be

indeterminate in its effect on performance.
Management Inﬂuence:

While the main stated goal of the project was reduction in work-in-process, the
team tended to place more emphasis on productivity issues. This was echoed in
comments by the Director of Operations during the event, when he stated that it
was really all about labor cost reduction. When the results were presented, the
greatest management enthusiasm was for the productivity improvements. There
was an equivocal response to the WIP reduction proposals; the General
Manager was in favor, but there was some dissension in the management ranks.

Some safety and ergonomic improvements were made, but these were not a

218

high priority for the team or for the management audience. Overall, the team

seemed to be able to accurately gauge management’s priorities.

219

Transcript of Events (Marginal notes are italicized and indented):
Day 1: 1:00 — 5:00
Deﬁned scope of event: frame assembly cell from cutoff saw to paint line

Project goals: 10% improvement in productivity, reduction of welded frame WIP

by 90%, reduce ﬂoor space by 10%, implement 3 safety/ergo improvements.

Before the event started, the origin of the goals was discussed with FAC
(see list of participants below). As far as local management was
concerned, the visible presence of WIP after the welding operations was
the indicator that improvements were needed, so this was the key
objective. The productivity goal was added because this is a key concem
at this plant, and even though the work cell had been re-ananged several
months eariier, this goal was kept to encourage some stretch. The
remaining goals seem to have been added to round out the process and
broaden the thinking. It was not speciﬁcally stated, but it appeared that
the motivation for including them came mostly from Corporate. As a
backdrop to the proceedings, it should be pointed out that normal
Corporate procedure was to insist that prior to conducting a Kaizen event,
the company should issue a statement that no one who was displaced as
a result of the event would lose their job. This company had elected not to
make such a statement, partly because of their heavy reliance on
temporary, contract labor. It is hard to identify exactly what impact this

had, but it would suggest a lack of incentive for shop-ﬂoor participation.

220

Participants:

Facilitator from Corporate (FAC)

Researcher (RES)

Director of Operations (DIR)

Quality Control representative (QC)

Operator from large machine assembly area (0P1)
Operator from cable assembly area (OP2)

Operator from paint line (0P3)

Notes on participant team: DIR and 0P1 had multiple years experience with the

company; the others were 1 yr or less. The three operators spoke extremely

limited English although DIR stated that they understood most of what was going

on. Their inability to communicate severely limited their contributions, and the

latter point was never convincingly demonstrated.

The severe limitation on communication immediately. suggests a very real
impact of individual cognitive limits. It appeared that the operators were
unable to make much of a contribution because they simply didn’t
understand what was going on, nor could they articulate any ideas they
may have had. There is also the question of whether they were
cognitively able to deal with the concepts of lean production, but being
unable to penetrate the ﬁrst layer, this was not relevant. It is important to
note that cognitive limits were not the only possible reasons for silence

and passive participation. The operators were sitting with their bosses’

221

bosses’ boss, who could dismiss them easily, especially since two of them
were contract employees (see note aboVe). Furthermore, given the
company’s refusal to adopt the no layoff policy, there was a good chance
that the outcome of the event could involve some job losses. While this
may not have affected them personally, there was certainly the possibility
of repercussions through their plant networks. Whatever the reasons, they
could not or would not participate in any meaningful way. This left a very

small “team” to accomplish the tasks.
General Manager attended opening session and wrap-up (GM)

GM’s opening comments: the stated goals are only guidelines; the team is free to
redeﬁne them as appropriate; made a point of emphasizing the competitive

environment and the need to stay competitive (i.e.: reduce costs).

In opening remarks, DIR suggested that the prevailing idea was to physically
relocate the assembly cell so that the weld booths were adjacent to the paint line
so that frames could pass directly from the welder to the line instead of being
stacked in batches and carried to the paint line. Clearly this solution anticipated
the ﬁndings, and he admitted this. He had also done some preliminary

investigation of the feasibility of the physical move.

At this point, RES suggested that the controlling factor was not the
physical anangement of the weld booths, but the scheduling of the paint
line, which was technically out of scope. It was not clear that any thinking
was changed at this point, but this ultimately proved to be the central

issue.

222

RES asked FAC ofﬂine about the source of the objectives. Answer was that they
came from ﬁndings of earlier Kaizen events, problems that were observed (but
not necessarily well understood) and from interaction with the corporate
facilitating group. It was not explicitly stated, but the impression was that the
WIP and productivity goals were mainly from the company, while the safety/ergo

and ﬂoor space goals probably came mostly from corporate.

Training portion proceeded following the standard manual, thru the Standard

Operations section.

Operators were totally silent throughout - not clear at all how much they were

getting.

QC would respond to occasional points with stories from his previous job —
echoes of the story-telling that occurred in the ﬁrst case study, but the group

dynamics did not support it very well.

In the previous event, the story-telling phase was more developed. It
appeared that the participants used stories to a) establish their credentials
in front of the group, b) test their understanding of the project scope
against the expectations of others and c) to introduce ideas that they felt
would be helpful. It seemed that QC was attempting to do these things,
but didn’t get much support. In keeping with their roles, FAC and RES
were mostly silent, as were the operators (as previously discussed). DIR

participated to a small degree, but the dynamic never developed.

223

Showed the Mercury Marine video as example of value stream mapping — led
into discussion of waste and non-value added operations. No response from

participants.

Showed HP video to introduce lean production concepts — again, no clear

response.

Ended day at 5:00

Day 2 — Tuesday, 8:00 — 5:00
Training continued with stop watch exercise for timing standard work.

Operators had some difﬁculties understanding the preCise procedure that was
being explained, but seemed to generally get the concept. For this and pretty
much all further activities, operators were teamed with a participant who spoke

English. Not clear whether this helped or inhibited their expression of ideas.
5-S section of training was presented by DIR.

Close of training was the section on managing Kaizen events — repeat of
philosophy of “quick and crude better than slow and elegant” and emphasis on

"learning by trying”.

Not clear at this point that the group was ready, willing or able to tackle the

job.

Event goals were re-stated and some data introduced.

224

The baseline for welded frame WIP was given as an average of 133 frames, or

slightly less than one day’s production.

A value stream map was copied and circulated to the team. This had been done
about six months earlier and contained information on cycle times for all
operations. FAC also had a report from an earlier Kaizen event on this same cell
that had more recent cycle time data. This report was not distributed to the team,
but FAC consulted it from time to time throughout the event, mostly as a check

on the results that were being obtained.

Additional data distributed to team: report of last 12 months production by part
number and a summary of stafﬁng and productivity of "shop and weld” area as it

was called.

The data were helpful to a point, but often failed to comespond to what
was observed. It was often necessary to ask for interpretations or
explanations from someone “in the know”. It tumed out that there was
much more going on than the basic production numbers would indicate
(certain re-work operations, jobs for other production cells, extra steps for
product variations that were captured by part numbers, but not noted in

the value stream map, etc. ). Not clear why FAC did not distribute results
of previous Kaizen event (this seems to be standard practice).

Presumably, the idea is not to constrain the thinking in the present by what
went before. If this is true, it is an interesting commentary on the idea of

Ieaming from experience.

225

After some discussion of this data, the team went to take a ﬁrst look at the

production area. The team spent about 45 minutes observing the operations as

individuals, then came back to the conference room to share what was found:

RES:

0

OP1:

OP2:

0P3:

QC:

ad hoc batching of parts through shop area
cramped work area around bender

makeshift stands for cooling fans using wire carriers
welding curtains torn or missing

too much equipment on ﬂoor (not in use)

scheduling of paint line not obvious — impacts our work

nothing

Bender looks hard to use (note: there was a used NC-controlled bender
sitting next to the area that was supposedly in the process of being
refurbished to replace the manual one in use — didn’t appear that any

progress was imminent.)

Nothing

Loose power cords strung around

Paint line loading area too spread out

226

o Frames were being wiped down prior to loading on paint line; why?

DIR:
o Unguarded foot switch on one machine
0 Mixed parts in bins
o Unsafe air hose (bulged) and non OSHA approved blow gun
0 Loose cords - trip hazards
0 Too many boxes of powder at paint line (surplus inventory)
0 Carts of defective parts - looked like they had been there a long time
0 Fan cords — trip hazard

0 Air line trailing across ﬂoor behind grinding booth

0 Parts coming off paint line are too hot to touch

0 High volume parts were using 2-pc hangers (two operations to load)

0 Cooling after welding

0 Layout of machines used for nosepiece production too spread out (forces
batching)

o No means for chip clearing or collection at milling machines

0 Bent parts need a lot of tweaking after they come off machine
Team deployed in groups to measure standard work operations.
RES paired with OP1 to measure nose piece production and tube bending.

During this process, all teams spent a considerable amount of time observing

and talking with the operators.

227

Of the ﬁve primary operators in the area, only one was a native English-
speaker. This individual was engaged in the process and had numerous
ideas. Because of the language barriers, the team probably relied too
much on this person to the exclusion of others who may have had better

insights or ideas. Certainly a cognitive barrier.

Returned to the team room to add up the cycle times and compare to existing

data.

Biggest discrepancy was in welding — seemed that 2 welders on 1 shift would be
adequate to handle production, but stafﬁng is 2/2. DIR said that they were
planning to take that down to 2 on ﬁrst shift, one on second. Called in area
supervisor who said that welders could do 45 frames/day, but our cycle time
measurements suggested that they could do 45 in 4 hours. This gap was never

satisfactorily explained.

There seems to be a recuning gap between what people “know", what is
recorded and what is observed. Recorded data is generally hard to
interpret, and often wrong or incomplete. Observing is too tedious, so

people seem to gravitate to and accept what others “know”.

Group discussion (note: “group” does not include the 3 operators who were

generally silent throughout) about the true issues and root causes.

Much discussion focused on the cutoff saw - the only operation with signiﬁcant

automatic machine time (about 30 sec. per cycle, each cycle produced 2 parts).

228

The waiting time during this operation was the only clear-cut example of waste in

the process.

Attempted a line rearrangement exercise to streamline motion and reduce space.

RES, OP1 and QC were assigned to this task.

RES suggested using paper cutouts as a language-free way to allow the
operators to express their ideas. It didn’t work very well; there was scant
evidence that they even understood what we were trying to do and we got

minimal input.

We did get some small efforts from OP1, but observed that he only seemed to

conceptualize moving one piece of equipment at a time.
Additional evidence of cognitive limits?

The group struggled with ﬁnding an improved layout — prompting QC to ask
“Don’t you sometimes reach the point where it is as good as it gets?” He also
expressed concern with the process — that we were treating people like robots
and trying to drive them to maximum production 100% of the time without any
allowance for human variability. (On a related note, DIR privately expressed
some cynicism to RES about the process; that it was all about head count

reduction and nothing else was important.)

These comments suggest a deeper cultural layer of understanding about
what we were doing - and disagreements about the rightness of the
model. It certainly appeared that QC and DIR were not committed to the

sort of productivity goals that GM had in mind.

229

Broke for the day with very little progress on rearrangement.

Day 3 — Wednesday; 8:00 — 6:00
Resumed team work on rearrangement

Called in one of the more articulate cell operators to get opinions on the layout

we were proposing - had no real objections or alternate suggestions.

Discussion moved to feasibility of actually moving the welding booths as

originally conceived, particularly with respect to the fume extractors.

Discussed these difﬁculties with DIR — he decided to call a contractor for an

opinion.
Sub-team made one more trip out to the cell for fact-ﬁnding, measurements.
Brainstorming session led by QC at easel on rearrangement alternatives.

OP1 started to express some suggestions and ideas, but was very difﬁcult to
understand. The English-speaking members of the group made an effort to
understand him, but this level of effort could not be sustained for any length of

time. (Note: OPl’s native language was not the same as that of 0P2 and OPS.)

Sense of frustration was building in the team: rearranging things really didn’t

seem to improve anything.

This was entirely bottom-up thinking - there was no evidence of new ideas

being introduced from the outside.

230

DIR also experiencing frustration — advised that moving weld booths was not
feasible on any short-term schedule, contrary to what he believed prior to event.

This put the team back to square 1 on rearrangement.

Parallel activity — RES consulted industrial supply catalogs to identify solutions to
issues identiﬁed on second day and observed during the timing exercise. Parts
placed on order with exception of an automated cooling system for one mill
(catalog did not have enough information, plus needed to consult with
Maintenance ﬁrst). Also, RES observed an operator on a different line using
clumsy, manual operations to tighten and cut off wire ties. Showed DIR the

range of guns available in the catalog and agreed to order one to try out.

Key to ordering in parts was knowing that solutions to the issues existed,

and where solutions would likely be found - example of meta-knowledge.

RES, QC and OP1 continued brainstorming on rearrangement without getting

very far - could not ﬁnd any examples to import from other areas.

Consulted line workers again for ideas - as ideas were discussed it became
clear that the shop area performed many more jobs than was ﬁrst assumed. The
“shop” title was appropriate, because in many ways this was a general purpose

machine shop.
This started to expose the limitations of documentary knowledge.

When it became clear that QC and OP1 were not going to come up with any
proposals, RES more or less dictated a rearrangement plan. It put the machines

in the correct production order and streamlined the ﬂow from one to the other.

231

There was also potential for a slight ﬂoor space reduction which couldn’t be fully
realized because of some clearance issues. Mostly it was a cosmetic

rearrangement.

Went out to plant to start implementing rearrangement (other sub groups were

doing the same with some of their improvements).

Saw plenty of evidence of learning by trying - people would rearrange things, try

it out, see how it looked, then adjust.

This could be called autonomous Ieaming — itjust didn’t apply to the

production process itself.

Observation: on the plant ﬂoor, the operators were much more in their element; if
given a clear task to do, they were quite effective and innovative in

implementation.

Additional observation: when the team entered and left the conference room, the
three operators always used the back door; even though this was not the most

direct route to the plant ﬂoor (everyone else used the front door!)

This could be evidence that power and cultural issues were more

important than cognitive ones.

Broke for the day with most of the physical rearrangement complete.

Day 4, Thursday 8:00 — 5:00

232

Day was spent primarily ﬁnalizing and trying out ideas proposed on previous day.

Most work was in small sub-groups.

Additional infomration was sought from accounting about what parts went
through the paint oven. It became clear that the original idea of synchronizing

the welders to the paint line could never have worked.

Based on this, a cyclical painting sequence was proposed where the bulk of the
day shift would be spent cycling back and forth between the two highest volume
parts on an hourly cycle (for WIP reduction). The paint line supervisor came in
and discussed the process and agreed to try it out for part of the shift. (This was

done and It seemed feasible.)

Another example of autonomous Ieaming - in this case roughly 4

changeovers were sufﬁcient to decide that the process was feasible.

A pair of carts was fabricated to serve as “supermarkets” for welded frames to

hold roughly an hour’s production.

RES consulted with maintenance person and used internet to research options
on automated coolant dispenser for mill — made proposal to be followed up on

30—day list.

The team proposed small batches of nose pieces and cross members as way to
approach one-piece ﬂow. Actual batch sizes were 4 for nose pieces and 5 for
cross members. Both modes were tested - initially took about twice as long as

larger batch processing, but some improvement doubtless possible with practice.

233

More evidence of autonomous Ieaming. Would guess that it might take a

day or two to reach a steady state (roughly: 50 - 100 cycles).

It seemed like every time the team went out into the plant, they saw operations
being performed that had not been taken into account. Many of these were small
production jobs for parts other than the primary ones we were studying; others
were additional steps or rework operations that were not documented. This
made it very difﬁcult to propose a standard work regimen. The team was
handicapped by not having any member who really knew what went on in this
production cell. Most of this sort of information came from one of the production
workers, but it came in bits and pieces and the team was never sure when or if

they had it all.

Also saw evidence of little details and maintenance type steps that had been
omitted during the timing exercise, perhaps because the operators wanted to
look better. These included periodic cleaning of chips from the mill, cleaning out

the ﬁxtures between each piece and a de-burring step.

The team was left with signiﬁcant discrepancies in time required. The welding
time standards were mentioned before — these were never resolved. The team
also felt that, based on what was observed and measured, that all work other
than welding could be covered with 2 people on one shift when the actual stafﬁng
was 3 on ﬁrst shift and 1 on second. Eventually proposed 2/1, believing that
implementing standard work for the primary parts would allow them to be
completed on one shift and one person on the second shift could take care of all

the other jobs.

234

The elements proposed were: cyclical painting schedule, supermarket for welded
frames, standard work in shop with small transfer batches. This put in place the
major elements of a lean production system where only the painting is scheduled
and everything else runs on Kanban signals. What is not clear is whether a true

understanding of the concept exists within the operation.
The overall concept was discussed in the full group.

OP3 expressed concern that too much time would be lost in paint changeovers
because of the different bake and line speed requirements of the different colors.
The supervisor was invited back in to discuss this issue — explained that it was
not a big deal since the difference in speeds was not great and didn’t hurt to have
parts in a little longer than necessary. The result would be a very slight loss in
painting capacity, but that is not currently a constraint. There is still the issue of

overspray and how much separation needs to be maintained between colors.

OP1 expressed concern that the cyclical painting sequence would cause
problems for the assembly area (his work area). He was not able to articulate
why this would be, but presumably he was afraid that without a large buffer, they

would be in danger of running out.

Day 5 -— Friday — 7:30 -1:00
First few hours spend preparing for the presentation.

To their credit, OP’s 1-3 agreed to participate although it was very difﬁcult for

them.

235

The presentation drew relatively few questions.

GM focused primarily on the claimed productivity improvements and head count

reductions.

There were no questions about the safety and ergonomic issues, although these

were looked at during the tour of the cell.

This seems to conﬁrm the company’s priorities as discussed at the
beginning. It is probable that everyone in the company is well aware of
the prioritization that goes on, and will reﬂect that in their areas. It is
probably no accident that the safety/ergo areas seemed to have been

largely overtooked in the past.

The general foreman expressed concern about operating with reduced levels of
WIP. He said that he didn’t feel comfortable with less than one day’s buffer
because of the unreliable labor situation. He was never sure who was going to

show up and felt that he would be in danger of shutting down production.

The GM responded with the rocks and stream analogy — if the rocks are the
problem, don’t cover them up by raising the water level; get rid of them. Clearly

this issue was far from being settled.

This invites a discussion of absorptive capacity; speciﬁcally, the ability of
the company to absorb and utilize lean production practices. Absorptive
capacity is traditionally framed as a cognitive issue, and there were
obvious cognitive issues here. It is not clear that cognition is a concept

that applies readily to an organization, but individual cognitive limits come

236

into play. On reﬂection, individual cognition did not seem to be a limiting
factor - certainly there were individuals who didn ’t understand the concept
or signiﬁcance of lean production, but those individuals could be educated,
replaced or bypassed. The issue seemed to be at a deeper, cultural level.
There was a prevailing “justiﬁed belief’ about how the operation should be
run, per the foreman ’s reaction. The ability of the company to “absorb”
lean production depended on its ability to displace an existing paradigm.
The more widely shared the paradigm, and the more tacit it is (as in not
being articulated), the more difﬁcult this will be. This suggests that

organizational cognition can be viewed as a cultural phenomenon.

Reﬂections on what is being left behind in this process:

Hard details (artifacts):

O

O

0

Revised machine layout

Supermarket tables for welded frames

New air vise and ﬁxture on a drill press

Prototype of revised frame hanger for paint line

Cooling fan for parts coming out of paint oven

“Sanitized” work area (loose cords tied up, new ﬂoor mats, new blow guns,
torn welding curtains replaced, etc.)

Self-ejecting chuck key

Quarter turn knob on tumbler

Wire tie gun in use

Soft details (documentation, but no implementation)

237

Proposal for smoothed production plan in painting
Proposal for pull system for welded frames
Proposal for small batch ﬂow in shop 1

Proposal for automated coolant dispenser

Revised stafﬁng schedule for shop, based on standard work

Abstract details (ideas, concepts)

0

0

General ideas about lean production, pull systems, one-piece ﬂow
Some person-to-person transfers of ideas (ex: in talking with one of the
workers, RES suggested use of an expanding collet to ﬁxture the nose

piece which was too big for a conventional collet)

238

APPENDIX IV

Kaizen Event #3 - June 27 - July 1l 2005

Introduction:

The plant in which this case is set is a North American manufacturer of plumbing
products. It is a sister plant to the one studied in event #1, but has signiﬁcant
differences in its corporate history, reporting structure, employee demographics
and in local culture. The business outlook for the company is generally positive,
driven by strong growth of the home building and home improvement markets.
This growth has been accompanied by consolidation in the customer base,
resulting in increased demands for cost reduction and ﬂexibility of delivery. In
response to these pressures, local management has embraced a ﬂexible
manufacturing strategy with lot sizes of ﬁnished goods SKUs as low as one.
'Continuous improvement is a high priority within the plant as evidenced by the

fact that this Kaizen event was #64 for the facility.

The workforce in the plant is drawn from a rural area and is characterized by high
average seniority. Industrial experience in other facilities is rare, as is post-
secondary education (at least within the event team, but apparently this is true of
the plant in general). The plant is not unionized, and the work environment

appears to be reasonably cooperative.

The work area that was the subject of this event was a packing line where end-
item SKUs are packed for shipment to the customer. Approximately 15 items go

into a box to form the ﬁnal SKU. The packing line is synchronized to the output

239

of a “racetrack” production line that produces a family of core valve components
around which the ﬁnal product variants are built. As a result of the diversity of
product, it is normal to change end-item SKU as often as 90 times per shift; each
change requiring a shufﬂing of components to and from the line. It was the

difﬁculties created by this situation that the event set out to address.

The event was undertaken without speciﬁc metrics or numerical targets; success
was measured by visual reduction in clutter, operator testimonials that things
were improved, and it was expected that longer term (month to month)
improvements would be seen in quality costs. In this respect, this event was
unlike the previous cases where speciﬁc improvement targets were identiﬁed.
This event was much more a case of purposive experimentation, an element of
the Toyota production system as described by Spear & Bowen, 1999. The
prevailing philosophy seems to be to use Kaizen events as a way to introduce a
controlled amount of variation into an otherwise rigid system so that the better
ideas can be retained. Failures to improve the process at hand are not
necessarily viewed as failures of the philosophy. The fact that this was event

#64 in this plant supports this view.

Review of Field Notes In the Context of the Research Protocol:

Knowledge Classiﬁed As Being Explicit Or Tacit:

 

Two types of tacit knowledge were visible in this case. The clearest example
was the process used by the team leaders to optimize their production

schedules. They used similar heuristics to do this, but appear to have arrived at

240

their present level of knowledge independently. In a simple test aimed at making

the process explicit, they did not come up with the same results.

The second form of tacit knowledge is more debatable. This was the prevailing
philosophy within the plant that ﬂexible production with lot sizes down to one was
the correct way to run the business. This rationale was accepted but not shared
by several members of the project team. These individuals felt that small lot

sizes created more work and trouble than they were worth.

All of the improvements put in place by the team could be classiﬁed as explicit
forms of organizational knowledge. Seven examples were noted where the

process left the work processes in a different (and hopefully improved) state.
Transformations Of Knowlglge Within And_Between Explicit And Tacit

Some attempt was made during the event to convert tacit knowledge to explicit
knowledge by making the decision rules for scheduling more explicit so that the
task could be automated. This was not very successful for two reasons: a) even
with the simplest decision rule (minimizing component part changes) the
production scheduling problem is a large scale example of the traveling
salesman problem and is very computationally intensive, and b) the actual
decision rules involve a tradeoff between minimizing component changes and
spreading them out over time. There is no simple substitution elasticity for these

criteria, making this a very complex, non-linear problem.

Another attempt to convert tacit knowledge to explicit was to rearrange the

supply shelves to group the component parts by the ﬁnal assembly “family”. On

241

the surface, this may appear to be a fairly simple-minded improvement, but it can
also be seen as a transformation of tacit knowledge (the ordering rules for

scheduling) to explicit knowledge (the efﬁcient ordering of parts on the shelves).

As with the prior cases, the large majority of the work that was done in this event
involved very explicit forms of knowledge, principally the physical arrangement of
the work space. None of the seven explicit results had any visible antecedent,
with the possible exception of the new packing bench which was generally similar
to other designs in use in the plant. The rest were examples of common sense
rearrangement of the particular elements present; perhaps guided by some
principles observed over time. Trying to analyze these events as transformations

is not particularly useful.

Knowledge Classiﬁed By Location

It is fairly easy to identify the new knowledge created within this event and to
categorize it. Ovenrvhelmingly, it is embedded in the physical layout of the work
area - the racks, shelving and assembly ﬁxtures that were modiﬁed. There was
no documentary record of lessons learned, except perhaps to the extent that the
researcher began to document the decision rules for production scheduling for
future use. It could be suggested that some lessons learned might reside in the

memories of the individuals involved, but there is a difﬁculty with this view.

The difﬁculty arises when we try to identify sources of knowledge. No documents
or experts were consulted; the source of everything that was done was the
collective memory of the participants. This collective memory consisted (as far

as could be observed) of what they had seen being used elsewhere, primarily in

242

the same plant. This being the case, it would not be expected that the
participants would take away any new knowledge, although the mixing of people
from different areas of the plant may have broadened the experience base of

some or all of them. This was not observable.

More likely, what was taken away was the experience of what details did not
work out quite as planned. This would be consistent with a view that the chief
value of experience is in knowing what not to do. Overall, attempting to identify
sources of knowledge is a frustrating exercise. This is leading to a view that
knowledge, at least in the narrow pragmatic deﬁnition, is not something that one
simply gets in one place and applies at another; rather it is something unique that

is created in situ each time it is needed.

As in previous cases, the preferred source of knowledge was the individuals’ own
memories. The hierarchy was more pronounced in this case, with very little

consultation of others and no observable use of any documentary form.

Actions By Individuals To Retrieve Or Store Knowledge By Location

As mentioned, the storage of the new knowledge is easy to observe - it resides
exclusively in the physical arrangement of parts and ﬁxtures. The only
documentation observed was a single, back-of-the-envelope sketch of how one

of the racks would look.

As with previous cases, the overwhelmingly preferred source of knowledge was
the team member’s own memory. Occasionally, they would go to look at another

line to see how something was done. No outside sources were consulted.

243

Processing Of K_nowlec_lge AnglOr InformatpLBv Individuals (Narratifve Versus

Paradigmatic)

Bottom up processing was very evident in this event, and appeared to be the
dominant mode of thought. Most of the racks, ﬁxtures, etc. were built with little
more than a sketch on the back of an envelope, and there was a great deal of
rearranging things until they ﬁtted. That said, this was not random trial and error.
There were clearly some organizing principles that were being applied. For
example, ergonomic considerations were important in determining the height and
angle of the ﬂow racks. In this example, these principles were not expressed in
any objective form such as number of inches off the ﬂoor, but rather in the form
of a mental model of what a “good” rack looked like. This is clearly a narrative or
top-down form of processing. As with the previous cases, any implementation of
an improved work practice had identiﬁable elements of both kinds of processing,
suggesting that the combination of the two may be a necessary condition for

organizational learning, as it is deﬁned in this research.

Individual Cognition (As In Pre-Existing State Of Knowledge)

 

In the previous cases it was noted that cognition played a role in limiting the
scope of search efforts, in the sense that people did not look for things that they
did not already know something about. On the surface, that seemed to be the
case in this event as well. As described in the overview, most of the participants
lacked any post-secondary education, or any industrial experience outside their
present company. In part this was a consequence of the plant’s small town

location, which also dictated that there was very little surrounding industrial

244

infrastructure that could have served as a resource. As a result, it was not
surprising that the universe of potential ideas was limited to what existed in the

plant.

On the other hand, it should also be pointed out that the preamble to the Kaizen
event and the deﬁnition of scope and objectives did little or nothing to encourage
more speculative search. While search efforts were deﬁnitely constrained, it is
perhaps not fair to draw the inference that cognitive limits are the causal factor.
In the previous event it was also remarked that the cognitive limits of individuals
had surprisingly little impact on the ability of the organization as a whole to
absorb new ideas that were presented. This situation did not really arise in this

case, so there was no additional light to be shed on this issue.

Power And/Or Knowledge As Mogerators Of Group Activity

 

Both organizational and expert power tended to come into play. In this event,
one of the facilitators was a production supervisor, and so was nominally higher
in rank than the participants. This did not appear to inﬂuence the progress of the
event in any way. The two shift team leaders from the assembly line were also
participants, and their knowledge and experience constituted a source of expert
power. Generally speaking there was a tendency to defer to their assessment of
the situation and to accept their ideas. This facilitated a certain degree of

learning from experience as will be discussed below.

245

Shared Logics Of Action (Existence Of And Chmes To)

A familiar conﬂict was apparent during the course of this event, the conﬂict
between a “lean” philosophy (characterized by one piece ﬂow) and the
“efﬁciency” model that emphasizes batching and large runs. In this case, the
situation was reversed from prior cases: plant management had embraced at

least the ﬂexibility aspect of lean and would schedule runs as small as one piece.

The performance metrics in place however, still reﬂected the efﬁciency paradigm
with work cells being judged on parts per labor hour. Those whose perspective
on the business was limited to an assembly line (such as the participants in this
event) tended to embrace the efﬁciency model and wanted to see longer runs, at
least partly because of the metrics. This caused no particular problems in this
event, but the conﬂict was visible nonetheless. Although there are
inconsistencies, and pockets of resistance, the dominant logic of the company
appears to be one-piece ﬂow. No changes to this logic were caused or observed

during the course of the event.

On a smaller scale, another example was observed. When the team leader for
the ﬁrst shift was presented with the mistake-prooﬁng ﬁxture, it required that he
change the direction in which he had his shift assemble the part. This procedure
had been implemented for a logical and valid reason. Because the revised
procedure also had a clear and valid logic, it was a simple matter for him to
weigh the beneﬁts and agree to the change. There was no conﬂict with the basic

beliefs about how parts should be assembled.

246

Changes To Perfonnatjve Aspects Of Routines

As with the other cases, the sole output of the event was a revised set of ﬁxtures
and processes. What is interesting is the way that this came about. Kaizen
events are examples of induced learning more or less by deﬁnition since they
represent an investment in improvement of the process. Autonomous learning
on the other hand has been observed to only occur within work units (typically an
individual) that are capable of autonomously generating variation in their work
processes to effect a process of artiﬁcial selection. This event was unique
among those studied so far in that it combined the two. By bringing the two shift
team leaders together and encouraging them to experiment as opposed to
making a speciﬁc improvement, the event allowed the packing line as a whole to
make one iteration on a classical trial-and-error learning curve. Furthermore, this
was not random trial and error. The experience of the team leaders allowed
them to bring their past thinking to bear on the situation. Much of what they did
in this event was enacting ideas for improvement that they had been

accumulating through observation and analysis.

As such, this is the ﬁrst example seen where the company has actually been able
to learn from experience. This connection is still somewhat tenuous since it
depends on the experience of the individuals involved rather than the experience
of the company. It also requires that those individuals be insightful observers
and devote some effort to thinking about how things might be improved. This still
does not support the argument that experience causes Ieaming, but it suggests a

way that experience might moderate Ieaming in a positive way.

247

Changes In Situated Performance

This case was somewhat different from the previous two in that there were no
speciﬁc improvement targets stated at the outset. As a result, there were no
metrics used that would deﬁne an improvement in performance. The most that
can be offered in this case was the visual reduction of clutter in the work area
and the testimonials from the operators that the new racks and ﬁxtures were an
improvement. This is a somewhat weak argument, but if these issues are
considered to be important to the company, then they are sufﬁcient to state that

some degree of learning took place.
Observations From the Field Notes In Addition to the Research Protocol:

Some observations were made that were not speciﬁcally anticipated by the
research protocol, but were considered interesting in the context of the research

question. These are noted for comparison to other cases.

Stogy Telling:

Story telling was present in this case as with the others, but it was much more
social in nature and less task oriented. The examples that were task oriented
were very focused. This can perhaps be explained by a couple of factors. One
is that there was not much sense of task ambiguity at the outset, so there was
less pressure to feel out the situation. Another possible explanation is that the
participants had a homogeneous industrial background, with little diversity of
experience to bring to the table. The task-focused stories tended to be a way

that lessons from the past were passed along to the newer team members.

248

Evidence of Autonomous Learning:

As discussed above, the entire event could be considered as an attempt to
promote autonomous Ieaming for a whole work area. This is interesting because
It is something that would not otherwise take place. It is the ﬁrst instance seen of

induced variation and selection to simulate autonomous learning.

The typical examples of autonomous learning were seen when the revised
procedures were turned over to the regular operators. When the mistake-proof
ﬁxture was introduced, one of the team leaders tried it out for a few repetitions
and pronounced it acceptable. Both shifts used the new ﬁxtures, and while the
learning effects were not observed directly, both shifts reported that they worked
satisfactorily. This would suggest that any learning effects were incorporated

very quickly.

Evidence of Reactive Learning:

No speciﬁc instances of reactive Ieaming were observed in this case. One
reason for this might be that the object of the project was a layout more than a
process, so less attention was paid to what the operators actually did. Another
reason might be that the company had a history of repeated Kaizen events,
which may have previously eliminated the more obvious band-aid ﬁxes on the
line. If this were actually the case, it would be an interesting commentary on the

results of reactive Ieaming.

249

Transcript of Events:

Day 1: Monday; 11:00 am start:

Present:

FAC1 (Co-facilitator from HR)

FAC2 (Co-facilitator; Production supervisor)
TL2 (Team Leader, second shift)

TL1 (Team Leader, ﬁrst shift)

CL3 (Chem lab tech, 3'd shift)

TST (Tester)

ASM (Assembly)

RES (Researcher)

Event began with opening remarks by the Plant Manager:

0 Emphasized the importance of Line 10 to the company (highest volume line)
0 Also, competitive situation, need to embrace “lean” production (interestingly,
the competition was not only other companies and China, but other plants

within the same company - apparently they can compete for jobs).
. Described the Kaizen process as being important in its own right — even if no

dramatic improvements are made, it sets the stage for the next event.

These remarks were delivered in a low-key, conversational style. This was not

something dramatically new to the company.

Review of Scope and Objectives:

 

Scope: “Improve material ﬂow at Line 10 pack-out”

250

Objectives:

Redesign pack-out station

Implement safety/ergonomic improvements

Implement mistake-prooﬁng (poke-yoke)

“Implementation of BOMs” (no explanation of this item provided at this stage)

Improve return system for carry-over parts.

FAC1 reviewed standard ground rules for Kaizen events, followed by self-

introductions.

Round table discussion of expectations for the event: ‘

“Fresh ideas”

“easier processes”

“More efﬁcient way to pack”

RES asked if there were any measures or standards for these objectives
(“how will we know if we’ve been successful?”) Group agreed that there was
no good measure, particularly for the ﬁrst objective; developing some kind of

measure needed to be one of the action items.
Success was not necessarily an expectation - the culture was such that an

intelligent failure would be OK - consider the event to be an experiment.

0 This is signiﬁcantly different from the previous events studied where there
were speciﬁc improvements that were expected. This was more of an

experimental mode to facilitate trial and enor Ieaming of the type normally

251

seen in autonomous Ieaming. This almost seems to be a new category:
“Induced autonomous Ieaming. ”
0 Reduction in rejects (customer returns or internal audits).

0 Reduction in supply stock at line.
FAC1 continued with standard training manual material: “what is Kaizen?”, etc.

Discussion of history of Kaizen at company - very slow acceptance at ﬁrst;
history of “fad of the month” type programs. Comparison to older Cl events —

current style is more democratic.

o This represents evidence of management commitment to the process that

is sustained over time — cleariy an important factor.

CL3 asked whether the company did or could learn from what others did and
from previous Kaizen events. FACZ answered that this was done to some extent
as innovations on one line were transferred to another, but also acknowledged
that it wasn’t always feasible. He spoke of working with three sister plants, but
not being able to transplant ideas from one to the other — in fact, plants have a
history of going in different directions. He attributed part of this to logistical

issues that inﬂuenced a plant’s ability to implement Kanban type systems.

This is consistent with the emerging ﬁnding that knowledge in one context

is rarely usable in another without signiﬁcant adaptation.

(Side note: on a pre-event tour of the plant with the local Cl manager; he
pointed out the parts “supermarkets” for pull replenishment. RES asked if

this was working in the sense of preventing stockouts without people

252

maintaining secret stashes of inventory. He replied that it was, and while
it took time to build trust, management now believed in it and wanted it

replicated throughout the plant.)

TL1 raised complaints about material storage in the plant, as it affected Line 10.
As background, he explained that the line might run as many as 100 part
numbers on a shift, with an average run size of about 40 parts, but many in the 2
or 3 unit range. This prompted remarks from CL3 and ASM about wanting to see
larger minimum run sizes and fewer changeovers. They felt that the company
was committed to following lean principles without recognition of the difﬁculty it

caused for them.

This is evidence of the continuing struggle between “lean ” and “efﬁciency”
paradigms. It was clear that local management were committed to the
ideal of single piece ﬂow and that this was not negotiable, but those
actually working on the line felt that efﬁciency was being compromised
and that the tradeoff was not necessarily a good one. Later it became
clear that efficiency type metrics were still very much in force; likely

cOntributing to this conﬂict.
Lunch Break

Considerable discussion over lunch, primarily between TL1, TL2 and ASM, about

ways to improve their jobs, organize the line and improve material handling.

253

Evidence of prior thinking about these issues. No mention of any outside

reference points or examples.

Strong reaction to the overall complexity of the job - wanted more support on
commonizing parts and packaging. Deﬁnitely wanted longer runs — a pack out

might consist of 10-15 parts, so the problem is to change over all these parts just

to run 3 units.

Historical stories emerged about designing and building racking, reworking the
mistakes, throwing away the non-ﬂexible designs. Mostly a distinction between

assembling “fast-tube” structure and welding up angle iron.

Further evidence of the lean/efﬁcient conﬂict and the tendency to pursue
local rather than global optima. Story telling starting to emerge at this
stage — in contrast to other events, the stories seemed to be tightly
focused on the status and history of the process at hand. This may be
because of the relative lack of outside experiences to bring to the table
(see overview). Another reason might be the way the problem was
framed initially: there was no particular encouragement to import ideas

from elsewhere, only to bring out the ideas that individuals already had.

Visit to work area:

Overall description: the production line portion is a conveyor loop ~30’ long that
produces the main valve body. There are a number of different variants, but they
are based on a common design and changeovers are not difﬁcult. The valve

body is the last part to go in the retail package, and it goes straight from the line

254

into the box, which is then sealed. There is a packaging line about 15’ long that
prepares the packages before they receive the main valve body. As noted
above, there are about 10 - 15 other items that go into the box, including internal
packaging. These other items are purchased parts or are manufactured
elsewhere in the plant. They are delivered to the area in boxes or tubs of 20-50
pieces. Furthermore, many of these parts must be handled carefully since they
are cosmetic in nature and cannot be simply dumped into bins for line
replenishment. This creates difﬁculties with the up-to-100 changeovers required
per day. TL1 explained that his ﬁrst task every day is to sort the production
schedule into the most feasible sequence for changeovers. TL2 does the same

thing, but with slightly different decision rules.

This would be an example of tacit knowledge, given the complexity of
choosing between ﬁnish commonality and component commonality. Each
team leader apparently had his or her own heuristics for this process,

which they were not able to articulate fully.

Behind the packing line is a large holding area for supply stock, which is in
addition to the main inventory holding area in the warehouse. The area by the

line handles a lot of partially used containers.

During the line tour, TL1, TL2 and ASM spent much of the time showing others
what their ideas would look like in practice. Other team members were observed

studying elements of the process and thinking out loud about improvements.

255

These were mainly local optimization type efforts, and didn’t consider the
larger picture (accepted it as given). No outside themes or ideas were

observed in this process.

Back in the meeting room, the following problem areas were identiﬁed:
. Parts ﬂow
0 Finding parts
0 Rack design
0 Need for more shelving (I)

. Need for more pre-staging of parts

Up to this point, information processing has been entirely bottom up or
paradigmatic — no evidence of any concepts or ideas imported from
outside sources. Also, this event has been unique in that there was no
real story-telling session - other than the “war stories” about shelving and
the discussion over lunch. Clear that the participants had some ideas
about how to organize things, but the source of those ideas was not so
clear. They appear to have been internally generated. The general lack
of storytelling may have been due to the fact that the participants already
had a high comfort level with the process and just wanted to “get on with
it”, or may have been due to a lack of use of examples by the facilitators

(who were insiders).

2:00 Continuation of the training materials: Takt time, etc.

256

Showed the Deep Dive video.
3:00 Reconvened to discuss action plans

Question about whether it would be feasible to rearrange parts in holding area by
ﬁnish instead of by type - to facilitate changeovers. Generally agreed to be a lot
of work, but doable in a week. TL1 started telling a few war stories from the line,

started to illuminate some of the issues that come up.

Discussion about what we might do about the mistake-prooﬁng goal — suggested

that plaster guard reversal might be a mistake that could be addressed.

3:30 - Break for the day.
Day 2: 7:30 am.

FAC2 reported that he had checked into getting BOM’s to use for our planning —
had been informed that they “weren’t too well maintained" and may not be useful

for our purposes.

This recalls the comment by one of the panel of experts in the validation
process that getting clean data was difﬁcult to the point that he had more
or less given up on it. Typically it seems that even a slight uncertainty
about archival data throws the whole set into question - there is not

sufﬁcient context to sort out what is useful.

Discussion about moving parts out of original shipping boxes and into more
standardized containers like baskets for bringing to line. Preferred method is to

bring boxes to line except for very short runs - less handling, damage, easier to

257

return. Problem is that incoming boxes are not standard sizes (and not

consistent run to run).

Discussion about rack design (ﬂat versus angled, right height). Seemed

predetermined that the existing racks would be replaced.

Team leaders wanted to see different arrangement for making up boxes and
inserting trim plate (ﬁrst part in). Current arrangement requires walking around
table (and into aisle) often. Discussion involved a lot of thinking out loud about

different alternatives.

Discussion revealed that those with experience on the line had a set of historical

production standards in their minds as reference points.

“Rack” team of TL1, CL3 and ASM went out to the line to brainstorm their new

design and to ask operators for input.

RES and FAC2 put heads together to design a mistake-prooﬁng device for the
assembly ﬁxtures. Worked up a preliminary design and took to model shop to

mock up.
9:00
Team met back in room to review rack designs and mistake-proof design.

TL1 pointed out a ﬂaw in mistake-proof ﬁxture - a minor modiﬁcation was needed

to address this.

Some discussion between TL1 and TL2 on this design. As background info,
three different bodies can be run on this ﬁxture. One can only be positioned one

way, while the other two can be run frontward or backward. One of these two is

258

already mistake proof by its design. The last one, which needs mistake-prooﬁng
and can be run either way, is the high volume unit. First shift runs the part
backwards to make it easier to install a screw, while second shift runs it frontward
to keep it the same as the other parts. Neither shift wanted to change, but it was
clear that one of them would have to. After the modiﬁcations to address TL1’s
issues, all parts would have to run frontward; not the way he liked to do it. He

saw the reason why it had to be that way and agreed to go along.

This illustrates the point that for a new routine to be adopted, the
ostensive part of the new routine must replace the conesponding part of
the old one. When the ostensive parts of the old and new routines are
based on clear and articulated logic, it is easy to evaluate the change

objectively and adopt it or not.

Discussion about rack designs - involved a lot of thinking out loud; “what-if” sorts

of discussions. No reference made to any external principles or examples.

Also discussion about how parts should be stored in racks (by ﬁnish, by family,
etc.). This mirrored the previous day’s discussion about how production should
be scheduled. This showed that the team leaders had certain decision rules
about this, but could not articulate them completely and liked to feel that they

exercised some creativity in this.

Some of the ideas proposed were certainly replications of things seen elsewhere,
but this was not stated explicitly. Given the long tenure of the participants, they
reacted mostly to things that had been introduced elsewhere in the plant. No

outside ideas were identiﬁable.

259

Given the demographics of the participants, this was not too surprising as

very few had ever worked anywhere else comparable.

Also seemed clear that beCause of their long experience, the team had a list of
accumulated ideas that they had thought about and wanted to try, but couldn’t
outside of the Kaizen framework. This seems to be a case of autonomous

learning, but requires a catalyst (inducement).
Hence the previous comment about induced autonomous Ieaming.

930am:

Out on ﬂoor in teams: rearranging existing racks and shelving and building new

racks.

Considerable evidence of trying things to see how they looked. Quite often, the
participants were not comfortable articulating their ideas in words or drawings;

they preferred to explain by showing.

At the same time, they were testing their ideas for feasibility - not all of them

looked so good when tried out.

Ideas would occasionally run into a wall, when there seemed to be a consensus

that some things were just too much work for the beneﬁts expected.

This is a tension that was observed in both the other events. There is a
sort of mental fatigue level that is reached where it seems easier to accept

the process as it is rather than deal with all the issues of changing it.

3:00 pm:

260

Review of progress:

Rearrangement of the work area provided more room for staging parts —

presumed to be an improvement, but no good metrics. Should save labor and

reduce mistakes.

4:00 pm: break for day.

Day 3, 7:30 am:

Team meeting: review of progress and assignments

RES to make sketch of ﬁxture modiﬁcations for model shop — went to see

machinist, laid out dimensions and set up a drilling ﬁxture to prepare for

changeover.
CL3 to tear down old fast-tube area
TL1 and TL2 to take pictures of line (“before”)

9:00 am: RES, TL1, CL3, ASM, FACZ worked in fast-tube area, tearing down old
racking and building new. Observed that there was a need for better sorting and
storage of components, both new and re-used. TL1 knew of some available

racks that would work. Group agreed to add a 5-S of the area to project scope.
TL2 and TST re-arranged parts in the pack supply area, grouping parts by ﬁnish.

Lunch Break — group discussion agreed that enough progress was made to add
the above-mentioned 58 to objectives, also to rearrange “shower-guard” area

(speciﬁcs not discussed).

261

Lively discussion over lunch, covering areas such as ﬂooded basements, mole

eradication, swimming pool maintenance, etc. (no task-related discussion).

The story telling impulse seems to be alive and well, just not directed at

the problem at hand.
1 :00 pm: First shift went home early to allow for changeover of ﬁxtures.

RES, TL1 and FAC2 stripped ﬁxtures offline and took to machine shop.
Returned with modiﬁed parts and re-installed. TL1 tested a few parts and agreed

that concept worked. Total time 1:45.

TL1 commented that second shift was a good place to introduce new techniques
because the workers weren’t old-timers who were set in their ways; less likely to
complain. Also noted that they could tell the ﬁrst shift that “the second shift had

no trouble with it.”

TL2 agreed to stay late to show the second shift the changes and explain how it

worked. TL1 agreed to come in early the next day to do the same with ﬁrst shift.

3:30: Break for day.

Day 4; 7:30

Reassemble in team room. Report back from second shift that the mistake-

prooﬁng ﬁxtures worked perfectly.

262

Morning to be spent putting ﬁnishing touches on the new racks, ready for

installation at 1:00 as per previous day.
5-S on fast tube area completed during morning.

RES and TL1 went to production control ofﬁce to print out sample BOMs for ﬁve
part numbers. RES wrote a small spreadsheet program to put part numbers in
best production order as test of heuristics used by TLs. TL2 selected the same
order that the program showed to be optimal. TL1 chose a sequence with about

15% more part changes.

Lunch Break

Report back from ﬁrst shift that there were no problems with the modiﬁed ﬁxtures.
1:00 Changeover time

Old tables and bins removed from line, new ones installed.

Numerous small problems discovered, several hours spent trying out the new
procedures, modifying aspects of it, suggesting changes. Very similar to design
process on Day 2, much trial and error, explaining ideas by showing. Many of

the ideas implemented, some held over for the morning.

Again, a fatigue limit seemed to be reached; some of the details seemed

too much to deal with at the moment - address later or work around.

3:30 RES and FAC2 met with plant manager to discuss build schedule
optimization idea. He seemed very receptive and noted that they already did
something similar with the changeover schedule for the automatic screw

machines. Agreed to devote some manpower to making it work. Was fully

263

aware of the problems associated with replicated data if the scheduling is done

external to the plant’s MRP system.

4:00 Break for day.

Day 5, Friday 7:30

Morning spent getting pictures, putting together the presentation and rehearsing

the parts.

The presentation was uneventful and there was minimal questioning. Part of this
may have been due to the absence of the General Manager. Some questions
from the Unit Manager about how we expected an improvement in quality

problems as a result of the changes.

It did not seem that he considered the plaster guard problem (the subject
of our mistake-prooﬁng device) to be a signiﬁcant issue. In this respect it
seems that the team may have had a different perception of the situation
and a diﬂ'erent prioritization of issues than did management — but this was

not clear.

264

APPENDIX V

PRESENTATION TO EXPERT PANEL

 

 

Study of Kaizen Events

 

John D. Hanson
Michigan State University
East Lansing, MI 48824-1122
Phone: (517)353-6381
e-mail: I'ﬂléﬂtﬂlQUEstﬂﬁil-ﬁﬁll

 

 

The Research Question:

 

I] How do ﬁrms access and apply
knowledge In their process
Improvement initiatives?

 

 

 

 

Why the question?

 

CI Much has been written, but very little
shop-ﬂoor observation.

C: Want to know more about how ﬁrms
generate competitive advantage.

 

 

 

Why Kaizen Events?”

 

D Bounded In scope and time - can
observe the whole process.
[I Data-driven, provides veriﬁable

results that allow us to say that
learning took place.

 

 

 

Reporting Preliminary Results

 

[‘3 May be interesting and useful.

‘3 “Interesting” academically - conﬂicts
with literature.

:3 Feedback and validation - are these
ﬁndings consistent with your
expeﬂence?

 

 

 

 

Deﬁnitions and Limitations:

 

D Context is high-volume productlon
environment under stable conditions.

I3 Know/edge narrowly deﬁned In
organizational sense: “situated
performance”.

Cl Learning is the “detection and
correction of error".

 

 

265

 

 

 

 

 

1. Experience Does Not Cause
Learning

 

as Key observation: ideas not rooted in
expenence.

I Innovations came from outside.

I Experience provlded constraints: (“that
won’t work because....“).

I The “learning curve” ls forward-looking,
not experience-based

 

 

 

1. Experience Does Not Cause
Learning (cont’d)

 

[3 Three types of learning:
I Autonomous - Traditional Ieaming
curve at the individual level.
I Induced Ieaming: planned, based on
future payback, draws from the outside
I Reactive learning - “insurance" against
a recurrence of past problems.

 

 

 

 

 

1. Experience Does Not Cause
Learning (cont’d)

 

:‘3 Why Is this Important?
I Productivity gains not automatic; based
on expected future volumes.
ll Firms learn from past mistakes, but the
results are mixed (insurance).
I Need to focus on external sources of
learning.

 

 

 

2. Cognition Is A Strongtimitui

 

I “Knowledge is knowing where to look it
up."

I People don’t look for what they don't
know about.

I “Tags“ are key to learning (meta-
knowledge).

I Stories are Important because they are
rich In tags.

 

 

 

 

2. Cognition 15 A Strong Limit
(cont’d):

:3 Why is this important?
I Knowledge management - oxymoron?
I Manage the tags, not the knowledge.
I Story telling is important - should be
encouraged.
I! Stories are how outside ideas are
imported In context.

 

 

 

 

3. “Absorptive Capacity” 15 Not An
Absorptive Phenomenon

 

III ACAP - ability to absorb and use
knowledge from outside the ﬁrm.

D Assumed to be cognitive: based on
related prior experience.

[1 Cognitive limits are not hard to work
around.

 

 

 

 

266

 

 

3. “Absorptive Capacity" Is Not An

Absorptive Phenomenen..Ssentisii-

I Explaining the concepts is not hard.

ti Problem is conﬂict with existing theory
of “why".

I The tools cannot be used until the old
theory Is displaced.

I Existing theory may not be clearly
articulated - an element of culture.

 

3. “Absorptive Capacity” Is Not An
Absorptive PhenOTEPPE-1.990.592...

 

:3 Why Is this important?
I Displacement problem Is often Ignored.
I Requires different tools.
I Culture change - known to be difﬁcult;
culture is good at perpetuating itself.
I Need new tactics - critical mass,
simultaneous changes.

 

 

 

4. Knowledge Isn’t “Accessed"; It's
Created As Needed,”-

EE Did not observe any instance of
knowledge being imported unchanged.

[3 Application was always unique.

[3 Two components: the facts at hand and
some organizing framework.

[3 These represent two modes of thought -
striking the right balance is important

 

 

 

4. Knowledge Isn't “Accessed"; It's
Created As Needed (cont'dm) ..,. .. .

D Why Is this Important?

I Can't manage knowledge; manage the
precursors: data and narratives.

I Gettlng good, clean data Is hard -
usually have to re-create It, or rely on an
Interpretation.

I Re-emphaslzes Importance of stories for
Importing frameworks.

 

 

 

 

Thank You For Your Attention

B These are the major points that I
wanted to emphasize

[3 Your feedback ls valuable and
appreciated:

:3 Do these observations ring true to
you, or do you feel that something is
missing?

 

 

 

267

 

Appendix VI

 

 

 

 

 

 

Validation Survey Results and Supplemental Comments
Question: 1 a) 1 b) 2) 3) 4a) 4b) 5)
m 2 4 5 4 5 3 4
3; 4 4 4 3 5 5 5
g 4 5 5 4 3 4 5
3 2 5 5 2 4 4 3
g 5 5 5 5 5 5 5
4 4 5 4 4 4 4
I Mean 3.5 4.5 4.8 3.7 4.3 4.2 4.3 I
Std.
Deviation 1.2 0.5 0.4 1.0 0.8 0.8 0.8

 

Additional Comments by Respondents:

 

 

 

 

Question Comments
1 a) - no comments
1 b) - Ieaming S_I_'I_O_QI_Q be evaluated on a forward looking basis
2) - no comments
3) - cognition not a “strong" limit - can go beyond it

- “people don’t want to move outside their box”

 

4 a) - idea makes sense, just not sure it’s true

- the longer the old paradigm has been in place the
tougher it is (age demographics a factor)

- leadership needs to force displacement

 

4 b) - no comments

 

5) - no comments

 

 

 

 

268

Appendix VII

Notes of Discussion with Expert Panel

The following is a loose transcript of the discussion that followed the
presentation. The primary emphasis was on capturing the essence of the
Individuals’ comments. To preserve their anonymity, participants in the

discussion are labeled by letter in the order in which they entered the discussion.

A: Liked the point about stories; agreed about their value, but noted that “there
are stories and there are stories” - meaning that some were credible, others not.
Felt that the most effective stories were those that reﬂected ﬁrst-hand

expeﬁence.

B: (reacting to the comment about knowledge being created at the point of use
from two components): saw the Kaizen methodology itself as the organizing
framework being applied to the speciﬁc company situation. Felt there was a third
component - the skill or personal experience of the facilitator in knowing how to
adapt the framework to the situation. This discussion made it clear that this view
can be applied at multiple levels of analysis, from the entire event as a unit, right
down to an individual participant wrestling with a small sub-component of the

problem.

C: countered this view by asking “have you ever been to the same Kaizen event
twice?” — meaning that the Kaizen event methodology didn’t really serve as a

uniform organizing framework; each event took on a life of its own.

269

D: Noted that he prepared for an event by selecting what stories he was going

tell.

This led to a discussion about matching the stories to the audience, knowing that
some audiences would relate better to some stories than others. Point was

made by several participants that the credibility of the stories was critical.

D: Commenting on the displacement argument: agreed that this was an
Important mechanism, but pointed out that this could not happen at the level of
an individual Kaizen event — had to involve direct action by senior management.
Said he had seen examples where this had happened (and many more where it

didn’t).

Discussion arose about the idea that knowledge is created uniquely in each
situation and how that related to attempts to replicate lessons learned in other
areas. The researcher suggested that both components would still be
necessary; the lessons learned somewhere else would be the organizing
framework, which might dominate, but would still need to take into account local

conditions.

E: Agreed with this point - said company 2 had learned this the hard way.
Having made a number of improvements at one site, they attempted to transfer
the results wholesale to another. They found they were unable to make it work

and had to break it up into pieces to get acceptance at new location.

270

The researcher recalled the example of GM building 3 identical plants in different
parts of the world with the idea that lessons learned would be transferable —

suggested that it probably wouldn’t work very well.
A: agreed and said “maybe that’s why you don’t hear about it any more.”

A: commented that this research had looked at Kaizen events In a way that none

of them had ever thought about - found it interesting.

A: reacting to point about how hard it is to get clean data — said he had been

ﬁghting that one for 40 years and had given up.

F: said very little during the discussion, but was generally supportive of the points
made. Body language however suggested that he wasn’t completely comfortable
with the narrow deﬁnitions of knowledge and learning — as a result, probably felt

that the ﬁndings were not as broadly applicable as they might have been.

271

Appendix VIII

DERIVATION OF THE LEARNING CURVE FROM AN EXPECTED UTILITY
MODEL:

Proposition 5 states that learning takes place as a result of some expected future
utility. This was based on repeated observations that no causal link could be
seen between past experience and present Ieaming activities. The alternative
view is that learning involves some sort of investment that must be justiﬁed by a
fonrvard-Iooking return model. This approach to formulation of a Ieaming curve
was proposed by Oi (1967), but was dismissed by Muth (1986) on the grounds
that in such a model, all improvements with a net present value (NPV) greater
than zero would be implemented right away and there would be no Ieaming

curve thereafter.

This view seems unrealistic. Potential Improvements are not all known in
advance; they must be searched for, and a priori estimates must be made as to
whether they are even worth searching for. When found, they cannot be
implemented instantly. Furthermore, ﬁrms do not have unlimited resources to
devote to search and implementation; it is not uncommon for positive NPV

projects to languish simply because there are more attractive ones available.

It seems more useful to denote a forward-looking factor M (which would normally
be expressed in units of currency) as the motivation for productivity
improvements, and postulate that the actual rate of implementation of the

Improvements (in the time domain) will be some company-speciﬁc function of M.

272

This rate would vary linearly with M over time if worthwhile productivity
improvements were inﬁnitely available, but would follow a power function as the
ﬁrm starts to exhaust a ﬁnite supply of ideas (per Muth’s model). Consider ﬁrst

the conventional formulation of a Ieaming curve:
y. =ax." (1)
where:
y, = the time or cost required per unit of production at time t
a = the Initial time or cost per unit of production
x, = cumulative production experience (units) at time t
b = a “learning” rate

In the revised, forward-looking formulation, the motivation would follow a

discounted cash ﬂow format:
M(t) = k, jV(r)e:"dr (2)

where:

M is the motivating force, typically expressed as the dollar value of

potential Improvements

k1 is a conversion constant to reduce the result to dollars or any other

desired unit
V(t) is the production rate in units per unit of time at time = t

r is the appropriate discount rate

273

In the steady-state, Inﬁnite horizon case, this reduces to a constant:

= ktV(t)

r

M (3)

We would then suggest that the actual rate of Improvement in, say, cost would

be:

dC__ .
E— Mf() (4)

where:

no) is a declining, convex power function of cumulative Improvements to

date, per the conventional formulation.

An interesting consequence of this formulation in theinﬁnite horizon case is that
if production rate changes, perhaps as a result of changes in the economy, to V’,
such that V’/V = a, M’/M will also equal a, as will (dC’/dt)/(dC/dt). However,
because of the change in production rate, when transformed to the volume
domain, dC/dV will appear to be unchanged. So, even though improvements are
evaluated and implemented according to varying constraints in the time domain,
the apparent Ieaming curve will appear to be a constant function of cumulative
production experience when viewed in the volume domain, consistent with

empirical ﬁndings.

The point of this exercise is to show that it is possible to make the argument that
learning is not motivated by past experience without thereby contradicting the
mass of empirical work that has been published to date. Assuming that

managers’ volume forecasts are more or less accurate, and that they make

274

rational choices, when viewed retrospectively, and in the volume domain,
process improvements will appear to follow a conventional learning curve,
creating an illusion that experience is somehow causal for productivity

improvement.

This formulation takes no notice of competitive dynamics and would be
appropriate in commodity markets where the ﬁrm is a price-taker. In a more
competitive market (speciﬁcally, one with a downward sloping demand curve),
managers’ forecasts may tend to become self-fulﬁlling prophesies. A higher
sales forecast than one’s competitors would lead to greater cost reductions
through learning, which would lead to a better competitive position, which would
lead to more sales, further reinforcing the apparent robustness of the empirically

observed learning curve.

Another interesting Insight from this formulation is that the motivation for learning,
and hence the rate of learning is dependent upon an opinion about the present
value of future improvements. In the case of a cost saving, this is (ceteris
paribus) relatively straightforward, depending only on a sales forecast and an
appropriate choice of discount rate. If the intended improvement Is in quality, the
situation is much more complicated, requiring an assessment of the present
dollar value of future improvements In quality. Clearly, this assessment will be a
function of the organization’s “shared logic” about the right way to compete and
the value of quality improvements. Thus organizations might be expected to

(reasonably) differ in their choice of k1 in Equation (2) above. In the more

275

conventional formulation of the learning curve In Equation (1), this would be an

explanation of variance in b, the so-called learning rate.

The idea of a learning curve being based on matters of opinion offers an
alternate explanation to that of Muth (1986) who noted a slight anomaly in many
learning curves. He noted that the very earliest stages of learning curves tended
to exhibit a slight concavity to the origin, unlike the convexity that prevailed over
the rest of the curve. He attributed this to a pull-ahead of some improvements
prior to launch. A plausible alternative explanation in light of the above
formulation Is that managers would tend to be skeptical of sales forecasts at the
time of launch and would discount them more heavily than they would once some

history was established.

276

lIjlijjljilijjjiIIIIII