mflbhnrb-thl

1%;

 

Ifl..{d...llr
.K‘xb ‘ a1
..3..... ..n ..
E‘ 3

3... ....

... ......zgikxul 2.33.2.3 .53“...
..3..}. A. n...» v ..ILSR
1...“.mflmmus..nz :
I... ha?!» a”. 5.?!
551 8.4.2.: ...-...!- 5.4982.
lax!!!“

Huh ; ...

‘3“...
z... z ..
.5... i ......u......

!
lis- .. .x.» 1.
lle-nivfltrali l.
(345‘. it. ”NIL“...

 

  

.... ......

....

 

 

. ..
...

.3... . fififl?tf#fl. ...rwunkfl.

..J

  
 

  

LIBRARY
“Tm Michigan State
20 .3 g UI'IIVBI'SIIY

This is to certify that the
dissertation entitled

AN EVALUATION OF THE IMPACT OF SUPPLY CHAIN
INFORMATION MANAGEMENT SYSTEMS ON
OPERATIONAL PERFORMANCE

presented by
Shawn Rocco Jones

has been accepted towards fulfillment
of the requirements for the

Doctoral degree in Supply Chain Management

 

 

wawfi

MajoTProf sor’s Signature
(103/443: 1 i 02 M 5/

Date

MSU is an affirmative-action, equal-opportunity employer

..—---.-o-n-—.---»—-.-.-.-.-.—.---»-.- .

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K‘IProlecc8Pres/ClRC/DateDue indd

AN EVALUATION OF THE IMPACT OF SUPPLY CHAIN INFORMATION
MANAGEMENT SYSTEMS ON OPERATIONAL PERFORMANCE

By

Shawn Rocco Jones

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Department of Supply Chain Management

2008

ABSTRACT

AN EVALUATION OF THE IMPACT OF SUPPLY CHAIN INFORMATION
MANAGEMENT SYSTEMS ON OPERATIONAL PERFORMANCE

By
Shawn Rocco Jones

Supply chain managers face increasing challenges to incorporate data from
suppliers, customers, competitors, and internal processes in a format to aid decision
making. As information management systems are developed to condense, contextualize,
and calculate the critical decision variables for a firm, there is a significant risk of
creating inaccuracies as the data is processed.

This study investigates the impact of variations in the accuracy of decision
information systems and completeness of data on managerial performance within a
supply chain management system. The Supply Chain Decision Operations Environment
simulation was adapted to provide a platform for this experiment. Afier randomly
assigning participants to one of four categories of corrected / uncorrected information and
condensed / uncondensed data, users executed several weeks of a simulated supply chain
operation. The final performance measures include raw material inventory status,
internal production efficiency, finished goods inventory status, market fulfillment rate,
net profit, and supply chain contribution to profit.

The accuracy of the information system had the most significant impact on the
performance of the managers and is recommended as the most critical investment
opportunity for information technology. Additional recommendations include key

decision variables under varying degrees of information system accuracy.

Dedicated to Beverly, Gwendolyn, and Evan.

iii

ACKNOWLEDGEMENTS

This finished dissertation reflects the combined efforts of a number of individuals
to whom I will forever be grateful.

Dr. David Closs served a role far greater than simply the Chair of the Dissertation
Committee. He opened the opportunity for a new phase of my life as the previous one
closed. In addition, he invested countless hours in simulation planning, career advice,
and personal development. For all of this and more, thank you.

Dr. Ken Boyer, Dr. Martha Cooper, and Dr. Cheri Speier comprised the
Dissertation Committee. I was truly privileged to have the opportunity to learn from
these distinguished experts in each of their respective fields. Despite their
overwhelmingly busy schedules, I always received the insight, direction, and advice that I
needed to continue work. I sincerely appreciate all that you have done to help me
achieve my goals.

Dr. Fred Rodammer serves as Director of the Center for Leadership of the Digital
Enterprise, which oversees the SCODE simulation as one of many successful projects.
His willingness to include me in the SCODE project, flexibility to allow me to adapt the
simulation to the experimental requirements, and trust to manage the project to fruition
means more to me than he understands. Thank you.

Cheryl Lundeen, Laurie Fitch, Sherry Danner, and Sheila Mead each took active
roles in my progression. In three years, I never had a question that one of you couldn’t
answer. You kept me on track, focused on the goal, and cheerful in the face of often

overwhelming obstacles. Thank you each for your camaraderie and advice.

iv

Alice Kalush, Jim Murphy, and Ping Liang were essential to development and
fielding of the SCODE simulation. Without your expertise, none of this would have been
possible. Congratulations on a job well done and thank you for helping me achieve my
goals.

Lieutenant Colonel Michael Weeks first introduced me to the opportunities and
challenges of academic research. I appreciate your guidance and advice before and
during the program.

Without the support of my family, none of this would have been possible. I thank
my parents who set a strong example of hard work and dedication to goals. I hope you
know how grateful I am for the tools, encouragement, and understanding you provided to
help me through this program.

My children, Gwendolyn and Evan, deserve a great deal of credit. They kept my
life balanced, never complained when we took another trip to the office, and were always
willing to reward my hard work with a trip to the MSU Dairy Store for ice cream.

My wife and best friend, Beverly gets the final and largest share of the credit.
Words cannot express my gratitude for everything you have done to make this dream
come true and I am thankful every day that I have the opportunity to share my life with

you.

TABLE OF CONTENTS

 

 

 

 

CHAPTER I: INTRODUCTION- 1
1 .0 INTRODUCTION .................................................................................................... 1
1.1 RESEARCH OBJECTIVE........................................ ................................................. 1
1.2 CONTRIBUTION OF RESEARCH ............................................................................. 3

I. 2. 1 Theoretical Significance ................................................................................. 4
1. 2. 2 Managerial Significance ................................................................................. 5
1.3 STRUCTURE OF THE DISSERTATION ...................................................................... 7

CHAPTER II: LITERATURE REVIEW - 8
2.0 INTRODUCTION .................................................................................................... 8
2.1 DEFINITIONS ........................................................................................................ 8
2.2 THE RESOURCE BASED VIEW OF THE FIRM ........................................................ 14
2.3 THE KNOWLEDGE BASED VIEW OF THE FIRM .................................................... 17
2.4 FACT-BASED KNOWLEDGE WITHIN FIRMS ......................................................... 19
2.5 INFORMATION COST, COMPLEXITY, AND PROCESSING ....................................... 22
2.6 DATA - IN FORMATION - KNOWLEDGE —- PERFORMANCE LINKAGES .................. 24
2.7 CONCEPTUAL MODEL AND RESEARCH HYPOTHESES ......................................... 26

CHAPTER III: RESEARCH METHODOLOGY- 34
3.0 INTRODUCTION ................................................................................................... 34
3.1 RESEARCH OBJECTIVE .............................. 34
3.2 SIMULATION MODELING ..................................................................................... 34

3.2.1 The Supply Chain Operations Decision Environment Model ....................... 36
3.2.2 Model Verification and Validation ............................................................... 40
3.3 RESEARCH DESIGN ............................................................................................ 45
3. 3. 1 Experimental Population .............................................................................. 45
3.3.2 Experimental Design ..................................................................................... 45
3.3.3 Experimental Variables ................................................................................ 48
3.3.4 Performance Variables ................................................................................. 5 7
3. 4 PERCEPTUAL MEASURES ................................................................................... 60
3.5 DATA ANALYSIS ............................................................................................... 63

CHAPTER IV: RESULTS - - -- 65
4.0 INTRODUCTION .................................................................................................. 65
4.1 EXPERIMENTAL POPULATION DESCRIPTIVE DATA ............................................. 65
4.2 STATISTICAL METHODS AND TESTS OF ASSUMPTIONS ...................................... 70
4.3 Two-WAY OVERALL RESULTS ........................................................................... 75
4.4 MAIN EFFECT RESULTS ...................................................................................... 88

vi

4.5 SURVEY AND NOMINAL GROUP TECHNIQUE RESULTS ....................................... 95

 

 

4. 5. 1 Survey Results ............................................................................................... 95
4.5.1 Nominal Group Results .............................................................................. 100

4.6 HYPOTHESIS TESTING ...................................................................................... 104

4. 6.1 Hypothesis 1 ................................................................................................ 104

4. 6.2 Hypothesis 2 ................................................... ‘ ............................................. 1 05

4. 6.3 Hypothesis 3 ................................................................................................ 107

4. 6. 4 Hypothesis 4 ................................................................................................ 108

4.7 DISCUSSION OF RESULTS .................................................................................. 109
CHAPTER V: CONTRIBUTIONS AND CONCLUSION 112
5.0 INTRODUCTION ................................................................................................ 1 12
5.1 RESEARCH CONTRIBUTIONS ............................................................................. 112
5.2 MANAGERIAL CONTRIBUTIONS ........................................................................ 1 13
5.4 LIMITATIONS AND OPPORTUNITIES FOR FUTURE RESEARCH ............................ 1 16
5.5 CONCLUSION .................................................................................................... 120
APPENDICES- - - 121
REFERENCES 146

 

vii

CHAPTER I: INTRODUCTION

1.0 Introduction

As supply chains continue to provide a challenging environment requiring agility
and flexibility to take advantage of new opportunities, executives are increasingly reliant
on management information systems to lead their companies. To keep pace with the
increasingly complex supply chains, these information systems collect data from many
sources to compile a hopefully meaningful set of decision metrics for managers. These
systems frequently enable managers in one continent to oversee operations across the
world. However, there are opportunities for the information to become corrupt,
inaccurate, or completely wrong. Does the manager usually detect incorrect information
prior to making decisions or is it only after developing a certain level of knowledge that
the manager questions the accuracy of the system? Even if the inaccuracy is detected,
does the manager know what other data are available to continue Operations? This
research investigates these questions through studying manager’s decision making

processes in a complex global simulation experiment.

1.1 Research Objective

The objective of this study is to improve understanding of the decision making
process in a supply chain environment and the balance between information accuracy and
data compression. Specifically, this research explores the impact of varied data quantity
and information quality in management information systems on a manager’s ability to

make sound decisions in a time-constrained environment. In order to explore these

questions, subjects participated in a structured experiment involving the Supply Chain
Operations Decision Environment (SCODE) computer simulation. The participants were
randomly assigned to one of four groups receiving different levels of feedback
information and quantity of feedback data. The users are unaware of any difference
between their feedback mechanism and any other players. After a series of rounds, or
simulated weeks, the users complete a detailed survey about their experience and
participate in a nominal group feedback session.

Four research questions are proposed with hypotheses explained in detail in
Chapter 11. Research Question 1 posits whether decision information systems mediate
the impact of data saturation on a supply chain manager’s performance. Tools such as
management dashboards are prevalent in industry to quickly highlight metrics or status
levels that require attention by a manager. One question this research seeks to answer
focuses on the manager’s ability to make micro-level decisions from the macro-level
aggregate metrics presented by the dashboard and whether the dashboard provides the
ability to process vast amounts of data into a quickly comprehendible Visual signal for
management action.

Research Question 2 inquires if the level of detail of the data available to a
manager mediates the impact of inaccurate information on that manager’s decision-
making ability. In the event that, whether purposely or accidently, the information
management dashboard presents incorrect information, is it better for the manager to
have access to vast amounts of data or compressed amounts of data? The study evaluates
manager’s performance in each instance and proposes the right amount of data to be

gathered and retained to minimize the impact of erroneous information systems.

The third research question asks whether there is a pattern Of incorrect
information provided by decision information systems before a manager seeks alternate
decision variables. By studying a manager’s ability to detect errors in the information
system and exposing the manager to various patterns of inaccurate information, insight is
gained into the trust relationship between a manager and the computer system.

The final research question inquires what decision variables managers identify as
the most important and least important to monitor in a supply chain. Direct inquiry
through a structured survey and a nominal group feedback period at the end of the
experimental session contribute to capturing the user’s experience. Analysis of the group
feedback provides Significant contributions to identifying the most critical data and
information for management systems to capture accurately as well as offering a listing of
the variables that may not need to be collected.

This dissertation reports the findings of experimental sessions utilizing the
SCODE computer Simulation within the design outlined in Chapter 3. These findings
make a number of significant contributions for supply chain managers to increase the
effectiveness and efficiency of their decision information systems as well as improving

understanding of the processes by which supply chain operational decisions are made.

1.2 Contribution of Research

This research provides important insight to both a practitioner and an academic
audience. For the academic audience, this research refines understanding of the
knowledge-based View (KBV) of the firm and expands previous research involving the

bullwhip effect, information sharing among supply chain collaborative relationships, and

information processing in a global supply chain. For the practitioners, this document
contains information on the behavior of managers operating a simulated supply chain in a
global environment. The insight gained assists practitioners in making both daily

operational decisions as well as strategic investment and policy decisions.

1. 2. 1 Theoretical Significance

In 1971, Andrews proposed that distinctive organizational competencies form the
basis of competitive advantage for a firm. Wemerfelt later (1984) refined this
proposition into the Resource Based View of the Firm. In both of these works, the
authors argued that firms are comprised of bundles of capabilities. These capabilities
could take a number of forms such as employees, equipment, methodologies, access to
markets, or other traits. The firm’s position relative to their competition within these
resources could make these a source of competitive advantage within the marketplace. In
the 119905, a number of researchers (Huber, 1990; Grant, 1996; Zack, 1999) identified
knowledge as the most Significant competitive advantage of the resources available to a
firm.

Davenport and Prusak (1998) propose processes for managers to gain knowledge
from information available within the firm. In turn, this information is gained from data
recorded throughout the firm. By progressing through the hierarchy, Davenport and
Prusak argue that simple records of transactions can form the basis for evaluating new
opportunities within the firm. However, this structure leaves two significant gaps
explored by this research. First, if the processes to translate data into information are

compromised, how is knowledge creation affected? Are managers expected to identify

discrepancies in the information based on prior experience, or is new knowledge formed
that incorporates the erroneous information? Second, if the processes are compromised,
is it possible for the managers to create knowledge directly from data without the
intermediate processing of information? The results proVide insight into the difficulty of
identifying the knowledge that contributes to competitive advantage versus the
knowledge gained through inaccurate information which could ultimately become a
competitive disadvantage.

In addition to building upon the work of Davenport and Prusak, this research
focuses on the bounds of the Knowledge Based View of the Firm. In particular, the
premise that more knowledge leads to greater absorptive capacity (Cohen and Leventhal,
1990) is evaluated under varying degrees of information accuracy and data compression.
Similarly, a vast literature stream supports information Sharing as essential to reducing
the phenomenon of order amplification throughout the supply chain, or the bullwhip
effect (Lee, Padmanabhan, and Whang, 1997; F iala, 2005; Croson and Donohue, 2006;
Disney and Towill, 2003). This research proposes sharing of specific information can be
helpful, but sharing of too much information or complete sharing of data sources could be
insufficient to mitigation of the bullwhip effect. Lastly, the research highlights
differences in trust behavior in a human-human interaction and a human-computer

interaction.

1. 2. 2 Managerial S ignificance
Supply chain managers make daily decisions affecting the profitability, customer

satisfaction, and supplier relationships of their firms. In order to make these challenging

decisions, the managers must be aware of hundreds of items impacting the supply chain.
These items include such diverse variables as raw material purchase prices, average
transit time from the supplier to the manufacturing plant, available labor hours,
distribution center finished goods status, and customer demand in a given market, to
name only a few. In a global marketplace, managers cannot personally oversee every
aspect of the supply chain. Instead, they rely on information systems to provide insight
into the operational status at specific points in time. Difficulties can arise when these
information systems are created and maintained by different agencies. The data may be
recorded for another use within the specific location, but also absorbed by the
information system. The data may also be modified without notifying the owner of the
information system. For example, a supplier in Singapore may record the number of
products available to sell to their customers. If a purchasing firm has a strong
collaborative relationship with the Singapore supplier and adheres to recommended best
practices in the industry, they likely share an information management system which
accesses this availability level and provides it to a manager in Canada. After a period of
time, the data system owner identifies missed sales Opportunities. The sales staff is
turning away prospective buyers for lack of inventory immediately available in the
warehouse. However, some buyers may request delivery in the subsequent week. In
order to rectify the problem and increase sales, the Singapore supplier changes the format
of the inventory status to reflect incoming shipments due within the week instead of those
currently available in the warehouse. If the Canadian information system owner was not
notified of the change, the manager could mistake the quantity immediately available for

Shipment and make decisions adversely affecting the supply chain’s customers.

This research highlights the potential for incorrect information presented by
management systems despite the presence of accurate data within a supply chain. By
understanding the key variables involved in decision making and the risk associated with
management information systems, practitioners can better balance the substantial
investment required to improve the data processing accuracy with the adverse effects of

inaccurate information.

1.3 Structure of the Dissertation

Chapter 1 provides an overview of the dissertation objectives, significance, and
purpose. Chapter 2 presents a review of recent literature as it pertains to the study. The
sections include definitions of essential terms used throughout this dissertation and
explanation of the Resource Based View of the Firm and the Knowledge Based View of
the firm, which form the theoretical basis for this work. Additional literature explores
managerial data requirements, information processing, computer simulation in research,
and details of the hypotheses supporting the research questions. Chapter 3 reviews the
research design, methodology, and tools. Chapter 4 contains the data analysis, results,
and research findings. Chapter 5 concludes the dissertation with implications of the
research findings, contributions to practitioners and academics, limitations of the study,

and opportunities for future research.

CHAPTER I]: LITERATURE REVIEW
2.0 Introduction
This chapter presents theoretical support for this research from the literature.
Included in this discussion are the definitions, bounds of the study, previous research
related to this work, and identification of specific gaps and opportunities explored in the
dissertation. Specific research questions and hypotheses to address these gaps are

presented as a precursor to the research design in Chapter III.

2.1 Definitions

AS the basis for topic development and establishing a frame of reference, it is
essential to present the definitions of key terms as they pertain to this study. In the field
of information management, three terms are consistently presented to explain the process
of transforming events into cognition. At the most basic level, data exists. Through
processes, this data transforms into information. At the highest level, information is
processed into knowledge. Various literature streams interpret the specific definitions
differently and there is frequently a lack of clarity in the usage of the three terms.
Therefore, further discussion is warranted to frame the definitions for this research.

In order to provide this frame of reference, contrasting definitions are presented to
illustrate the broad interpretation of these terms. Each paragraph discussing the usage of
the terms then focuses on typical business literature usage of the terms and concludes
with a clear statement of the definition used throughout this research.

Data are broadly accepted as a representation of fact. Within specific contexts,

data takes on slightly different characteristics. The Department of Defense specifies data

are a “representation of facts, concepts, or instructions. . .suitable for communication,
interpretation, or processing by humans or automatic means (DoD Dictionary, 2003).
Marketing dictionaries identify data as “facts that become useful information when
organized in a meaningful way or when entered into a computer (Imber and Toffler,
2007). In a more general form, the American Heritage dictionary presents the definition
as “factual information” (2007), adding a level of confusion and lacking distinction
between data and information. While various streams of literature adopt slightly different
interpretations of the term, for the purposes of this research, data is defined as “a set of
discrete, objective facts about events” (Davenport and Prusak, 1998). The explicit
inclusion of the term “objective” in this definition is significant. Without additional
input, data has no inherent meaning (Badii and Sharif, 2003). However, data is of vital
importance to firms and researchers. The data collected through various processes are the
raw materials critical to production of information or knowledge within an institution
(Chow, Choy, and Lee, 2003).

While some disparity exists in the definition of data, interpretation of the term
“information” is considerably broader. The American Heritage dictionary again lacks
specificity as it presents information as “knowledge derived from study, experience, or
instruction” (2007) as the primary definition. The Department of Defense crosses the
hierarchy in the other direction by including “facts, data, or instructions” to their
definition of information, though their dictionary does expand the explanation to include
“the meaning that a human assigns to data” (2003). Drucker emphasizes the human

interaction when asserting that information is “data endowed with relevance and

purpose” (1988). Drucker’s definition delineates the difference between data and
information for this research.

Davenport and Prusak propose five processes to add value to data in order to
create information. This dissertation focuses on these fiVe processes in progressing
through the hierarchy from data to information. The first is to contextualize the data in a
way for the user to understand why the data exist. Second, data are categorized into
components or similar units of analysis. Third, mathematical or statistical calculations
are performed to add meaning and simplify the data into a comprehendible set of
meaningful information. Fourth, errors are removed as the data are corrected. Lastly,
data are condensed into a format easier for the user to comprehend or analyze (Davenport
and Prusak, 1998).

As an example to clarify these processes, the simple case of recording data
obtained from a drive through window at a fast food restaurant is followed through the
five steps. Data consisting of order input time and order output time are recorded for a
month. A manager reviewing the data understands that these times are key performance
measures of customer satisfaction, which is critical to the success of her franchise.
Therefore the data has been given context to understand why it is important to the
manager. These sets of data are differentiated from the order times taken within the
restaurant, and so have been categorized by unit of analysis. The simple calculation of
subtracting the completion time from the input time yields the total customer wait time.
The manager can glean meaning from different wait times. If any of the wait times are
negative, this would clearly be an error that would be corrected or omitted from review.

Lastly, the manager would likely reduce the entire month’s customer wait times to an

10

average, standard deviation, maximum, and minimum. This step condenses the data into
a usable metric to present to the workers of the restaurant. By taking these five steps, the
manager was able to provide meaning and purpose to the data and, by the definition used

in this study, created information.

Figure 2.1: Processes to transform data to information to knowledge
(Davenport and Prusak, 1998)

 

 

 

 

 

 

 

I Information I

 

   

I I'J.‘ ll ' 'n _ . ' "' ' -ll' “'
,_ . _. ,
. . ,r
‘ .‘d ‘
' "'. ”in... wt. ... .

 

l Knowledge I

 

The highest level of the hierarchy is knowledge. Just as defining information
became more difficult than defining data, establishing a consensus on what it means to
know something is elusive. The dictionary attempt is circular and superficial by focusing
on such definitions as “the state of knowing” and “specific information about something”

(American Heritage Dictionary, 2007). The Department of Defense does not offer an

official definition of knowledge. Even as one of the pioneers of the Knowledge-Based
View (KBV) of the Firm, Grant evades a specific definition of knowledge. Instead, he
notes that the question, “What is knowledge?” has “intrigued some of the world’s greatest
thinkers from Plato to Popper without the emergence of a clear consensus” and, therefore,
asserts that “there are many types of knowledge relevant to the firm” (1996, p. 111).

While a clear consensus on the definition of knowledge is not found, business
literature focuses on two aspects that distinguish knowledge from information. The
necessity of human interaction with information is the first. Badii and Sharif (2003)
discuss the creation of knowledge by managers reflecting on previous experiences. This
reflection leads to a “justified true belief” (N onaka, 1994, p. 15) based upon the
experience of the person evaluating the information. The second differentiating factor for
knowledge stems from a person’s ability to comprehend new information and predict
future consequences based upon prior experience and evaluation of available information
(Davenport and Prusak, 1998). The definition used throughout this research combines
these two key points. For the purposes of this dissertation, knowledge is the person’s
justified true belief providing a framework for evaluating and incorporating new
experiences and information (N onaka, 1994; Davenport and Prusak, 1998).

As with processing data into information, Davenport and Prusak propose four
ways to transform information into knowledge. The first is comparison with previous
experience or standards. The second is for the user to understand the consequences of
acting on the information or what can be forecast from the information available. Third,

the user makes connections between one piece of information and others that are

12

available. Lastly, the user converses with other users with access to similar information
to gain a broader understanding of the meaning (Davenport and Prusak, 1998).

Continuing with the fast food drive through processing example, the information
the manager previously created from the data set of order input and order completion
times can form the basis of useable knowledge. Once she compares her average order
completion time to a standard, she understands whether her restaurant is performing well
or below standards. Additionally, her previous experience from other months as well as
realizing that the standard is based upon many years of historical customer satisfaction
records allows her to add significance to the month’s information. This is closely related
to understanding the consequences of the information. Her previous experience linked
profitability with higher customer satisfaction and linked higher customer satisfaction
with lower than standard wait times. Therefore, she knows the consequences of the
month’s drive-through transaction information. She connects this month’s performance
with other information available to her to determine if corrective action or rewards are
appropriate for the employees. Through conversations with other franchise owners, she
detects trends that may be affecting the industry as a whole and can better form a basis
for future decision making. By arriving at a final set of beliefs from the information, she
is prepared to incorporate the next month’s information and evaluate the ongoing
performance of her drive through as well as predict profit margins for the upcoming
months.

The definitions presented above for data, information, and knowledge help bound

this dissertation. In the following subsections, the theoretical framework supporting the

13

research questions is developed. The underlying theory is derived from the Resource

Based View (RBV) of the Firm.

2.2 The Resource Based View of the Firm

As firms develop strategies to compete in their industry, many will target certain
areas of strength within the company to exploit in the marketplace. In order to provide a
sustainable competitive advantage, these strengths must yield a unique position that other
firms within the industry cannot easily duplicate. These concepts form the premise of the
Resource Based View of the Firm (Andrews, 1971; Wernerfelt, 1984). RBV proposes a
causal relationship between the assets available to a firm and the products or services
offered to the marketplace. Additionally, there is capacity to exploit these resources to
improve performance and sustain competitive advantage (Conner, 1991). Evolving from
the general strategy of identifying strengths within a firm, specific traits are associated
with RBV as a strategic foundation. Specifically, the four resource‘traits are valuable,
rare, inimitable, and non-substitutable (Barney, 1991 ). Each of these four aspects is
discussed in greater detail below.

The theory of RBV provides a perspective to understand different levels of
performance among different firms with different asset portfolios. These assets available
to a firm are often separated into different categories for analysis. The first is physical
resources. These include geography, physical plants, equipment, technology, raw
material access, and distribution networks. The second focuses on the human factors

available to a firm. These comprise training, knowledge, opportunities, and experience.

14

The third category includes organizational routines such as processes, controls, and
practices regularly occurring within the firm (Barney, 1991).

To provide sustained strategic advantage, a resource must be of value to a firm.
The value could occur in a number of different means. The resource could be access to a
market allowing the firm to outperform the competition in speed to deliver new products
to the consumer. The resource could be an investment in technology that reduces internal
quality concerns that previously existed. Regardless of the specific resource, the value
added to the firm must be greater than the expense to gain that resource and be
compatible with the firm’s overall strategy (Barney, 1991; Mahoney and Prandian, 1992).

In addition to creating value within a firm, the resource should be rare to support
building long-term strategy involving the resource. As an early adopter of the rare
resource, a firm stands to control access to the resource within the industry. Access to a
mine yielding an extremely pure titanium source that requires significantly less
processing time or cost than the competition faces is one example. Employing a group of
scientists who bring together a unique set of skills to solve problems that other research
groups cannot is another example. As with the value aspect, the scarcity of the resource
should return a greater reward than the cost of acquiring the resource (Barney, 1991).

While rarity reduces the access of competition to a resource, a resource that is
inimitable cannot be economically repeated by others in the industry (Barney, 1991).
The reasons for inimitability could be diverse. In the previous example, it is possible that
the purity of the titanium as a raw material allows the firm to utilize a different method
for production than others in the industry. Because of the rarity of the raw material and

the unique processing capability of the firm, the processing resource is inimitable within

15

the field. There may be other processes that are inimitable due to the knowledge or
culture of the workforce (Conner and Prahalad, 1996). The Toyota Production System is
a widely understood process to empower employees, reduce waste in the system, and
paces production to a just-in-time supply system. Despite Toyota’s willingness to allow
competitors to observe their process, other automotive firms cannot duplicate Toyota’s
efficiency or quality of production. The cultural customs and knowledge the workforce
gained through years of experience are among two of the reasons this production system
has proven inimitable within the industry (Sugimori, Kusunoki, Cho, and Uchikawa,
1977; Womack, Jones and Roos, 1990). lnimitability of the production system supports
Dierickx and Cool’s work identifying characteristics of accumulating the resource
(1989). Specifically, the authors identify the value of time, mass efficiencies,
interconnectedness of similar resources, resource erosion, and causal ambiguity as critical
to establishing inimitability of a resource (Dierickx and C001, 1989).

The previous three traits are clearly important to return value to the firm owning
the resource. However, if the competition can easily attain a substitution to avoid the
rarity, value, or inimitability of the resource, the firm no longer holds a sustainable
competitive advantage (Barney, 1991; Peteraf, 1993). The existence of substitutes, as
established by Porter’s five forces model, reduces the monopoly’s advantage in the
marketplace (Porter, 1980). One example of a substitutable resource is an alternate
material or process that replicates the performance of the finished product without
requiring the rare, valuable, or inimitable resource. Altemately, a firm that creates a
completely different paradigm may eliminate the need for the resource entirely. Dierickx

and C001 (1989) present the example of Canon’s inability to compete with the dominant

16

position Xerox established in servicing copiers. By investing in a series of copiers that
no longer required significant servicing, Canon was able to substitute a better and lower
maintenance design for Xerox’s valuable and inimitable service network.

Viewed through the RBV perspective, a firm’s decision making process could be
a source of competitive advantage. In the case of managers compiling data through
commercially-available database management systems, the decision making process
would be easily imitated. However, if the process involves a unique manner of
processing data into management information systems and the manager’s analytical
process evolved from distinct corporate cultural or experiential knowledge, the process is
likely to be rare, inimitable, and valuable. While the actual decision making process
remains an individual-based action, firms leveraging their decision making processes as
competitive advantage are likely to have a reliable, sustainable, accessible, and highly

accurate information system to share among the workforce (Szulanski and Jensen, 2004).

2.3 The Knowledge Based View of the Firm

In the Knowledge Based View of the Firm, it is recognized that RBV identifies
opportunities for sustained competitive advantage by focusing on the strength of
resources within a firm. However, KBV extends this assertion one step further to identify
knowledge as the single most important resource available to a company (Grant, 1996;
Zack, 1999). Firms with the ability to identify and exploit specialized knowledge are
expected to adapt their resources more quickly and effectively in the marketplace than
their competitors. Additionally, the internalization of existing knowledge can foster

creation of new knowledge, thus reinforcing a dynamic progression of the critical traits

l7

(Zack, 1999). As in the RBV, KBV seeks to identify sustainable advantages for firms.
Two types of knowledge are identified as essential to sustaining performance: explicit
and tacit.

Explicit knowledge is defined by written procedures, recordable events, and
established processes that a firm employs. It is transferable from one location or section
of a firm to another. However, this trait also makes explicit knowledge easier for
competitors to imitate unless additional factors are involved (Zack, 1999). As soft drink
companies introduce a new packaging container for their products or an online news
service releases a new format for their material, the competition can swiftly adapt and
adopt a similar process without investing the research and development time and costs
into the new process. In the general case of all other variables being equal, unique
explicit knowledge is not likely to sustain competitive advantage for a firm.

Tacit knowledge refers to context-specific knowledge gained from a period of
experience within a company’s complex organizational routines (Zack, 1999). The
implied significance of experiencing time within the operating system, comprehension of
the underlying connectivity of resources, and causal ambiguity align with Dierickx and
Cool’s identification of traits critical to establishing inimitability of a resource (1989).
This tacit knowledge held within the workforce is extremely difficult to reproduce in
other settings due to its evolution as a complex resource requiring firm-specific assets in
order to be successful (Peteraf, 1993). Therefore, knowledge-based resources can yield
valuable, rare, and inimitable advantages to a firm.

The final significant facet identified in the RBV requires knowledge to provide a

sustainable advantage that rebukes efforts of competitors to find substitutes. In their

18

study of absorptive capacity, Cohen and Leventhal (1990) find evidence that the more a
firm knows, the greater its capacity to learn and incorporate additional knowledge.
Therefore, it is predicted to be more difficult for competitors to absorb and process new
information or acquire new knowledge than for a firm with unique tacit knowledge and
experience in the arena (Zack, 1999).

This stream of literature supports knowledge as a critical resource capable of
sustaining competitive advantage. At the individual worker level, the tacit knowledge
gained from experience with an operational system increases the opportunity to process
new information and gain additional knowledge. After reaching a level of proficient
knowledge about the job requirements, the individual is capable of evaluating which
additional knowledge is critical to success and what order to learn new processes. This
self-reinforcing cycle leads to a generalization that accumulation of increasing amounts
of knowledge is beneficial and should lead to higher performance in the marketplace

(Zack, 1999).

2.4 Fact-based Knowledge Within Firms

Companies operating within a global supply chain setting rely on information
technologies to monitor performance of plants, warehouses, transportation, personnel,
and various other aspects of the business from a centralized location due to constraints on
time and travel budgets. Numerous studies support the assertion that companies who
capitalize on their information technology capabilities can achieve improved performance
among their collaborative partners and increase coordination within their own
organizations (Shin, 1999; Min, Roath, Daugherty, Genchev, Chen, Amdt, and Richey,

2005). As companies cross traditional firm boundaries with information systems, real-

19

time electronic data interchange allows manufacturers to forecast retailer demand patterns
and promote sharing and consolidation of forecasting responsibilities (Challener, 2000;
Xu, Dong, and Evers, 2001). This centralization of information processing reduces
workload requirements on both firms and should lead to reduced amplification of order
uncertainty by essentially eliminating one link from the supply chain (Disney and Towill,
2003). The firms must ensure they have compatible information technology systems,
have sufficient quality controls in place, and resolve cultural or legal concerns with
sharing data across firm boundaries (Kahn, Maltz, and Mentzer, 2006).

Within companies, centralization of certain functions realizes significant cost
savings and control of scarce resources. An unintended consequence of consolidation for
some companies is the improved opportunity to acquire and disseminate implicit
knowledge among workers performing similar tasks. During a presentation to the
Michigan State University Operations Management Summit (2007), Sara Lee executives
discussed the recent alignment of purchasing fimctions into a centralized location.
Despite requiring an immense investment in information management systems, the
company is saving substantial costs through economies of scale and improved bargaining
power in specific markets due to the centralized structure. Sara Lee is also able to better
share best practices among purchasing agents by co-locating them within a centralized
unit. These purchasing agents share solicitation strategies, pricing information, and
promotional opportunities at a significantly increased rate when they are physically
working within the same office. In a similar move, the United States Air Force is moving
toward a centralized purchasing structure estimated to save over $1 billion per year. In

addition to the dollar savings and reduced workload at the individual Air Force bases,

20

leadership cites increased asset visibility, improved accountability in the contracting
system, and better support to the customer as key benefits made possible by a centralized
information management system. The purchasing agents share best practices, operating
information, and knowledge of successful methods to prOcess and track their orders at a
greater rate than when operating from geographically-separated bases (Holmes, 2008).
These are two examples of large organizations operating in a global supply chain
environment with centralized designs to better facilitate knowledge creation and transfer.

In their 2003 work, Lee and Choi investigated knowledge management factors
such as collaboration, trust, learning, centralization, formalization, skills that are both
deep and broad, and information technology support to better understand their impact on
organizational performance. Their model identified four specific areas of knowledge
creation processes, socialization, extemalization, combination, and internalization, to
measure the effects. Of particular significance to this dissertation, Lee and Choi found
support for information technology systems as facilitators of statistical knowledge
development.

Similar results have been attained in Britain as radio fiequency identification
(RFID) tags are added to increasing quantities of inventories of food and drug products.
The RF ID tags are performing their expected function of providing management with
near-instantaneous monitoring of inventory, logistics timeliness, and freshness factors.
Some unexpected benefits have been to focus traceability for theft, recalls, and other
unforeseen problems with the food or drug supply chain. Workers monitoring the RFID

tracking stations have become knowledgeable about expected patterns and are able to

21

rapidly detect variance from the norm (Jones, Clarke-Hill, Comfort, Hillier, and Shears,

2005)

2.5 Information Cost, Complexity, and Processing

Based upon the seemingly clear benefits of accurate and encompassing
information management systems for improved operational performance, it would seem
that every business would possess the highest level of information processing systems.
The most significant barrier to this level of information availability is the complexity and
cost of acquiring and operating such systems. As with most business decisions, a balance
must be attained to maximize the operational performance contribution of an information
management system at the most economical cost possible.

In terms of cost and complexity, the United States government manages one of
the most challenging networks of information. In Fiscal Year 2006, they invested $66.2
billion in information technology systems. Focusing solely on the Department of
Transportation during the same period, the government purchased a total of 74 budgeting
and accounting systems and 38 operational performance systems to track metrics within
the workplace (www.whitehouse.gov/omb/ budget/fy2008). The immense investment
required to purchase and maintain the systems is apparent. In addition to the cost, the
complexity of maintaining the vast number of systems is daunting. With this number of
new systems introduced in a single year, confusion is certain to occur when updating
older systems or integrating newer ones. The opportunities to inadvertently present

inaccurate information to the user are plentiful. The users are presented with a

22

challenging environment of diverse computer interfaces and knowledge to manage in
order to successfully complete his or her job.

In addition to the cost and complexity of the system at the firm level, the
individuals processing the information have a separate level of constraints. The human
mind is limited in its ability to process information. Bounded rationality is the term
Simon (1957) proposes to identify a number of dimensions that limit human rationality.
Due to the financial and comprehension limits of solving complex problems or storing
information, Simon recommends use of heuristics for daily decision processing rather
than making efforts to optimize each decision. By employing these useful tools to
approximate optimal solutions within a certain degree of accuracy, the computational
time and cost may be significantly reduced.

Consideration of time allocated for decisions is a critical factor in many supply
chain settings. Managers are frequently saturated with human resources requirements,
meetings to attend, production reports to review, and internal performance improvement
processes. The time available to oversee seemingly routine purchasing orders is likely
minimal. Given this restricted time, the manager must be afforded the tools to make
quick and accurate decisions. One method for compensating for the negative
performance impact of time pressure is information compression (Spurrier, Topi, and
Valacich, 1994).

In the contemporary management arena, company dashboards accomplish the
same task as the heuristics proposed by Simon (1957) and serve as the means for
information compression as proposed by Spurrier et al (1994). The analogy to the

dashboard in an automobile is intentional as managers are encouraged to focus on the

23

road ahead while frequently glancing at a few key indicators to provide status updates,
warning codes, and verification of progress. If the gauges are remaining within the
expected range, no action is required. Once the warning indicators show a trouble area,
the manager can divert attention to the necessary item and set a corrective course
(Dennis, 2006).

The ability of management to respond quickly to accurate decision variables is
significant in the global supply chain setting. Daily decisions frequently represent
actions occurring months into the future due to lengthy lead times, transportation
limitations, customs clearance times, and intermodal final delivery requirements. In
order to provide the manager the best opportunity to succeed, the quality of the
information provided as well as the quality of the decision making process must be of the
highest quality due to their direct relationship on the performance of the organization

(Huber, 1990).

2.6 Data — Information — Knowledge - Performance Linkages

As discussed in Section 2.1, Davenport and Prusak (1998) present a testable
framework for this study. The addition of a link from knowledge to performance, as in
Figure 2.2, completes the bounds of this study. A number of researchers established

strong linkages between knowledge and performance. A brief discussion follows.

24

Figure 2.2: Processes to transform data to information to knowledge (Davenport and
Prusak, 1998) with performance link (Choi, Poon, and Davis, 2008)

 

 

 

     

II, _ a Lin, ..1’)’ '-

(Davenport and Prusak, 1998)

 

 

 

I Information I

 

 
   

F.-

(Davenport and Prusak. 1998)

I Knowledge |‘ l Performance

 

m
n - . --_‘_.' , - ‘2 ', ;.\-',1U' .‘siu ‘ .. --"'.- '- '= a . ‘n'r-r" . "'-,_‘-‘-'i'..-!‘I Wm “'u: if"! ‘ I.’ -'a '- .' -' _
1 , ' _ ‘ ‘1 ' , . ' ‘ .. ‘ ' I ‘ ‘ ' I. ' H
.... ' . t‘ ....' ,“~ " -' ..‘L‘| a ‘
I“. ‘ .1. . 1 ‘c'i . .4» x, . t It; - 'L‘ ‘1 x a. I 'u ' m1 i m. ,v .. u“. -J'L v- a» - [all u

 

 

Choi, Poon, and Davis (2006) utilize a novel theoretical approach involving
complementarities of knowledge management strategies and operational performance.
The primary supported finding involved firms using a combination of a tacit-internal
strategy and an explicit-extemal strategy outperformed any of the other combinations.
Their evaluation of 131 firms engaging in knowledge management practices revealed
performance improvements in those firms capable of combining a strategy based upon
following documented best practices in relationships outside the organization and
capitalizing on intemally-acquired knowledge for within-firm decisions.

In addition to the significance of the complementarities in Choi, Poon, and Davis’

(2006) work, the authors found support for differentiating operational performance based

25

upon knowledge management strategy. For a setting in which all parties have the same
structure and opportunity for developing a knowledge strategy, the performance link is
expected to be present and equal for all of the four groups in the study. Through different
methodologies, researchers such as Sharma, Levy, and Evanschitzky (2007) and van
Buren (1999) found similar linkages between knowledge management strategy
employment and operational performance.

Kanawattanachai and Y00 (2007) identified links between knowledge and
operational performance through study of virtual teams. These researchers found the
three factors of expertise location, task-knowledge coordination, and cognition-based
trust all had varying impacts on operational performance over time. While virtual teams
took longer to develop trust than co-located teams, once a member demonstrated a
sufficient level of knowledge for the task, the location bias was eliminated. Additional
research in the area includes Lee and Yang’s (2000) work on the knowledge value chain
and the need to reach an established level of knowledge value in order to yield
knowledge performance. Pathirage, Amaratunga, and Haigh (2007) evaluated the tacit
knowledge present within construction workers and again identified a strong link

between a sufficient level of task-oriented knowledge and operational performance.

2.7 Conceptual Model and Research Hypotheses

With the research framework supported through extant literature and established
theoretical grounds, the final evolution of Davenport and Prusak’s (1998) hierarchy
positions this research. As seen in Figure 3, the research questions are established by

holding three of the methods for transforming data to information constant as well as all

26

four of the methods for transforming information to knowledge. The result is a 2x2
experimental design in which the information is either corrected or includes a random
erroneous factor and the data are either condensed into aggregate form or provided in full
detail. This design yields four categories of information. Evaluation of the research

questions involves hypotheses for each as explained below.

Figure 2.3: Processes to transform data to information to knowledge (Davenport and
Prusak, 1998) with performance link (Choi, Poon, and Davis, 2008) and experimental
design.

 

Data
(Accurate)

** 2

' Information
1, (Four categories)

 

 

 

 

 

 

 

 

 

 

3
1 4

 

   

‘ 1v ~n_,n_5'.-'..‘.. -' “"’ ‘ --*
i
‘3
I

 

 

 

 

Performance
| Knowledge l ‘ I (Financial/Operating)

Using Davenport and Prusak’s (1998) model as a basis, the research model was
developed to capture the experimental factors, environmental factors, operational
decisions, and overall performance as it relates to the Supply Chain Operations Decision

Environment (SCODE). The administrator set the supply chain environment for the

27

duration of the exercise. This environment directly impacts the criteria utilized by the
participants for operational decision making as well as affecting the operation of the
SCODE model during the weekly iterations of the scenario. Participants entered weekly
decisions impacting their supply chain model and entered those into the SCODE
simulation. After processing the input from both the administrative environmental
settings and the participant’s operational decisions, SCODE provides the users with
performance feedback. The subjects were randomly assigned to one of the four
experimental factor groups which impacted whether the data was condensed or not and
whether the information was corrected or not. The weekly performance measure then
sets the baseline for future decisions as processed through the experimental factors for the

subjects. Specifics of the methodology are presented in Chapter 3.

28

Figure 2.4: Research Model

 

Supply Chain Environment

- Location of Plants

 

- Base and Variation of Demand
- Reliability of Transportation
- Raw Material and Labor Requirements

1

_(_)_perational Decisions
- Select Supplier(s)
- Purchase Raw Materials
_, - Determine Speed of Production
- Select Product Mix
- Identify Markets to Service Demand
- Select Transportation to Destination

l

 

 

 

 

 

 

 

 

 

 

SCODEF

l

 

 

Performance
- Inbound - Market
- Internal - Net Profit Margin
- Outbound - Supply Chain Contribution

 

 

 

l

, Experimental Factors
- Condensed / Uncondensed Data
- Corrected / Uncorrected Information

 

 

 

 

 

29

Research Question 1 asks whether decision information systems mediate the
impact of data saturation on a supply chain manager’s performance. Tools such as
management dashboards are prevalent in industry to quiCkly highlight metrics or status
levels that require attention by a manager. This research focuses on the dashboard’s
ability to process vast amounts of data into a quickly comprehendible visual signal and
evaluate the manager’s ability to make micro-level decisions from the macro-level
aggregate metrics.

The first Research Question explores the ability of information systems to
mediate the impact of data saturation. Specifically, it states:

RQI .' In a time-constrained setting, do decision information systems mediate the

impact of data saturation on a supply chain manager’s performance?

H1: Supply chain managers performing in an environment with uncondensed

data, time constraints, and uncorrected decision information systems perform at a

lower level than those with corrected decision information systems.

This hypothesis delineates operating performance expectations for the two groups
with uncondensed data treatments. Given complete data availability, the group with
accurate decision information systems is expected to perform at a higher level than the
group with inaccurate decision information systems. This finding should be expected,
but becomes more interesting in light of the second research question and hypothesis.

Research Question 2 focuses on the level of detail of the data available in the

presence of inaccurate decision information systems. If the information management

dashboard presents incorrect information, is it better for the manager to have access to

30

vast amounts of data or compressed amounts of data? The next two research questions

and hypotheses follow:

RQZ: In a time-constrained setting, does the level of detail of the data mediate
the impact of inaccurate decision information systems on a manager’s decision
making ability?

H2: Supply chain managers operating in an environment of uncorrected decision

information systems, time constraints, and uncondensed data perform at a lower
level than those with condensed data.

RQ3: Is there an interaction effect between the accuracy of the information
system and the level of detail of the data?
H3: There is an interaction between the level of detail of the data and the
accuracy of the information system. The effect will result in the following
performance levels for the various treatment groups:
1) Highest performance level: Corrected information / Uncondensed data
2) 2d highest performance level: Corrected information / Condensed data
3) 3d highest performance level: Uncorrected information / Condensed data
4) Lowest performance level: Uncorrected information / Uncondensed data
The hypothesized interaction effect between the two variables marks a departure
from Davenport and Prusak’s (1998) model, but finds support in literature. Simon’s
assertion that bounded rationality limits the amount of data able to be processed supports
the first hypothesis. In conditions where the information management dashboard is
accurate, any additional insight the user receives directly from the data will augment
performance. However, when the user is forced to abandon the information management

dashboard due to inaccuracies, the additional data becomes burdensome and the user

cannot distinguish the important variables from the noise of the data.

31

The fourth research question focuses on the user’s reaction to incorrect
information in the dashboard. By manipulating the patterns of exposure to incorrect

information, the following hypotheses are tested:

RQ4: Is there a pattern of incorrect information provided by decision
information systems before a manager seeks alternate decision variables?

H4a: Managers provided consecutive reports from a decision information system
containing incorrect information cease utilization of the information system and
made future decisions based on another approach.

H4b: Once managers seek an alternate approach to decision making, they do not

return to utilizing the decision information system regardless of the accuracy of

the system.

These two hypotheses are based upon the user’s learning as the simulation
progresses. The first time the user is presented with incorrect information, it is expected
to take considerable time to identify alternate decision variables and establish a method to
incorporate the raw data into a decision making framework. After the information
management dashboard is consistently wrong enough to provide the user with sufficient
motivation, the user identifies other variables, regardless of whether they are as useful as
the corrected information could have been, and continues to use those variables
throughout the remainder of the exercise.

The final research questions focus on the crux of the question facing supply chain

managers today. It states:

RQ5: What do managers identify as the most important and least important data
to monitor in a supply chain?

32

The final research question asks which decision variables the majority of
managers identify as the most important and least important to monitor in a supply chain.
Analysis of the group consensus contributes to identifying the most critical data and
information for management information systems to capture accurately as well as
offering a listing of the variables that may not need to be displayed. This opportunity to
discuss the results of the experiment with the users immediately following the session
contributed to capturing a more accurate image of the user’s experience. Further, by
highlighting the most useful and least useful data, recommendations can be developed to
focus information technology spending in improving accuracy of those critical areas

rather than dispersing the funds across a wider and less useful collection strategy.

33

CHAPTER III: RESEARCH METHODOLOGY

3.0 Introduction

This chapter describes the approach used to test the hypotheses proposed in the
previous chapter. Following a pilot study to verify both the simulation environment and
the experimental instruments, several sessions of the main study were conducted. The
following sections describe the simulation program, validation and verification processes,
participant selection, identification of variables, and the approach to data analysis. The

analysis of data is presented in Chapter 4.

3.1 Research Objective

This research examined the challenges faced by supply chain managers operating
in an environment characterized by an overwhelming quantity of available data,
information technology systems with varying degrees of reliability, and constrained time
to complete all necessary tasks. Specifically, this study investigated the effects of
varying quality of information and quantity of data on a manager’s ability to make
decisions in a time-constrained setting. The questions to be answered by this research are
listed in Chapter 2. The following sections explicate the experimental design focusing on

these four questions.

3.2 Simulation Modeling
Supply chain management research presents unique challenges when designing an
experiment affecting financial and operational performance in a firm. Investigating

phenomena in their naturally occurring environment is preferred, but not always possible.

34

The intricate nature of a supply chain, the inability to control the environment to
completely account for confounding factors, and the potentially adverse effect on a firm’s
performance make supply chain experiments difficult to accomplish in real world settings
(Kelton, Sadowski, and Sturrock, 2003). Research involVing issues as complex as supply
chain management can best be evaluated using dynamic approaches in a controlled
setting (Ridalls, Bennett, and Tipi, 2000). Computer simulation is one recommended
dynamic modeling method enabling the researcher to provide a controlled environment in
a cost-effective manner without adversely impacting a firm’s livelihood (Helo, 2000).
Providing research evidence to evaluate the above hypotheses in a global supply chain
setting can therefore be appraised through use of a verified and validated computer
simulation program.

An area of less precedent in literature is the use of simulation to stimulate theory
development. Davis, Eisenhardt, and Bingham (2008) explore the use of simulation
methods to develop theory. These authors provide a number of guidelines as a roadmap
for using simulation in theory development. Additionally, they present advantages
simulation offers over competing methods. By starting the study with clearly defined
research objectives rooted in established literature, simulation provides a unique
opportunity to isolate specific interactions over time. Additionally, the precision and
ability to measure numerous performance variables simultaneously lends simulation to
refining theory. A common criticism for simulation studies stems from researchers
viewing the setting as too artificial and the outcomes as predetermined. However, by

following the recommended roadmap, these concerns are sufficiently addressed. Table

35

3.1 summarizes the steps to ground simulation studies while pushing the research

envelope to expand or develop new theory.

Table 3.1: Roadmap for Developing Theory Using Simulation Methods (Davis,
Eisenhardt, and Bingham, 2008, p. 482)

 

Step

 

1. Begin with a research question
2. Identify simple theory
3. Choose a simulation approach
4. Create computational representation
5. Verify computational representation
6. Experiment to build novel theory

7. Validate with empirical data

 

This dissertation follows the recommended roadmap to build theory. Step 1 is presented
in the introduction. Step 2 is discussed in Section 11. Steps 3 — 5 are discussed in this

section. Steps 6 and 7 are explored in Section IV.

3. 2.1 The Supply Chain Operations Decision Environment Model

The Supply Chain Operations Decision Environment (SCODE) is a “simulation
training program to help executives and students learn the Digital Enterprise intricacies of
international supply chain management” (www.bus.msu.edu/clodeo. Through this
simulation, teams create supply networks to acquire, store, manufacture, market, and

distribute goods throughout an international setting.

36

The general environment is depicted in Figure 3.1. The simulation administrator
creates an environment designed to focus participant learning in a specific supply chain
objective. During the training and experimental sessions for this research, the objective
is to gain a better understanding of purchasing raw materials in a global supply chain
environment. In general, the users may have the opportunity to manufacture up to four
products, with each requiring a specific number of labor hours and a mix of up to four
raw materials to produce one unit of finished goods. These raw materials may be
purchased through contract or spot-market buys and may be found in any of four global
regions designated by “North America,” “South America,” “Europe,” and “China.”
There may be up to four manufacturing plants for the users to select from, which include
co-located raw material warehouses and a finished goods warehouse, with a single plant
in each of the previously identified four regions. Adding the region “India” to the
previous list completes the five markets providing demand for the finished goods. Raw
material and finished goods are transported throughout the supply chain by air, land, or
sea channels. The participants have the opportunity to select the mode of travel based
upon the speed required and price he or she is willing to pay to meet his or her objectives.

The specifics involved in the experimental scenario are presented in Section 3.3.

37

Figure 3.1: SCODE Supply Chain Network

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Suppliers Suppliers Suppliers Suppliers
(North America) (Europe) (South America) (China)
l
i
v v v - 1
North America Europe South America China
RM warehouse RM warehouse RM warehouse RM warehouse
Plant Plant Plant Plant
DC DC DC DC
v j i v
v v V r r
DC (North DC DC DC DC
America) (Europe) (South (China) (India)
America)
1 r r
v
v
Market
(N. A.)

 

 

 

 

The simulation operates at three distinct levels. The highest architectural level
involves the “Administrator” decisions. These options define the environment for the
users. The scenario designer can tailor any variables ranging from raw material price or
availability to the market demand function curve and including over 1,000 other variables
impacting all aspects of the supply chain. The second level is the supply chain “Design.”

The design worksheets allow the users to set the initial conditions for their supply chain.

38

While some variables can be altered later in the game, usually at a substantial cost, others
remain fixed for the duration of the exercise. The operational level involves the
“Weekly” decisions. Each round simulates one week of elapsed time where the users
make decisions about the product quantity and mix to manufacture, the transportation
mode for raw materials or fixed goods, production acceleration to meet immediate market
demand, and investment decisions for research and development or promotions, to name

a few variables. The User’s Manual Executive Summary is included in Appendix A.

Figure 3.2: SCODE model framework

 

 

 

 

 

Strategic Decisions Administrator Decisions
(One time team decisions at start) (Controlled by administrator)
. Plant (own or contract) 0 Demand variation
. Plant capacity . Production economies
0 DC (own or contract) . Transportation lead times
a DC capacity oSupplier availability
0 Market & DC allocation oReport detail
0 Initialization Decision
0 Initial inventory
0 Scheduling

 

 

 

 

 

 

l SCODE

 

 

 

Operations Decisions
(Week/y team decisions)

oSourcing (plant, supplier, raw material,
mode, quantity)

oProduction (what to produce where ‘ RESULTS
and when, quantity)

oMarketing (R&D, promotion, price)

oDistribution (source, DC, mode,
quantity)

 

 

 

 

 

 

 

 

 

 

 

 

39

Feedback from the simulation occurs on two worksheets. The first is a financial
report containing data ranging from revenue to detailed supply, plant, and distribution
center costs for both the current week and year to date. This financial data lists aggregate
data, data separated by region, and data ordered by product. The second form of
feedback is the “Output” reports providing detailed analysis of the supply chain
performance. The report contains status on seven different areas including raw material
quantity and locations, production output and schedules, distribution center performance,
location and expected receipts of finished goods, sales report, lost sales report, and
system messages. As an additional aid to users, the output worksheet contains a decision
information system which captures relevant data and recommends key variable input for
the next week of operational decisions. For example, the interface detects the team
scheduled production of a product requiring 100 units of a raw material, but there are
only 20 units of this raw material in the warehouse. The display would recommend
ordering 80 additional units of the raw material to meet the production schedule. The
portions of SCODE directly affecting the experimental variables are discussed in greater

detail in Section 3.3.

3. 2.2 Model Verification and Validation

When creating a simulation model, two items are critical for success: verification
and validation. Verification involves certifying the computer program is correctly
designed and implemented (Sargent, 2000). Validation tests ensure the computer
simulation is producing results that are consistent with possible results from the actual

system operating in the business world (Sargent, 2000). Four categories of validation

40

include conceptual model validity, computerized model verification, operational validity,
and data validity (Sargent, 2000). These are addressed individually below.

Conceptual model validity requires evaluation of the theories and assmnptions
upon which the model is built as well as determination that the model is reasonable for its
experimental purpose. Two typical methods employed in conceptual model validity are
face validity and exploring traces (Law and Kelton, 2000). In the case of SCODE, a
panel of experts comprised of industry executives and academic professors envisioned
the simulation environment. The vision was implemented by computer programmers and
supply chain management students. This implementation team closely followed trace
technique to evaluate every step of the simulation program while processing input and
output. Comparison with hand calculations and expected outcomes highlighted
discrepancies. The program was continually refined until the implementation team was
confident with the operation and results SCODE provided. Throughout the creation of
SCODE, the team was closely advised by the panel of executives and academics. The
expert panel reviewed the operational environment and tested SCODE with realistic
inputs and confirmed the model was based upon valid constructs and implemented in a
way that met the purpose of creating SCODE. Therefore, conceptual model validity was
achieved.

Computerized model verification requires analysis of the computer programming
and implementation of the model. Traces and error log reviews are typical methods
utilized in computer model verification (Sargent, 2000). As previously discussed,
meticulous trace examination of SCODE input and output screens provided evidence of

an accurate computerized model. Further, SCODE outputs error messages to the user

41

whenever input is inconsistent with expectations. Close scrutiny of these error messages
led to further refinement of SCODE. Once the trace examinations and error messages
confirmed proper execution of the program, the computerized model verification
requirement was met.

Operational validity is attained when the output has sufficient accuracy for the
input conditions. A number of tests exist for operational validity including extreme
condition tests, internal validity checks, parameter variability-sensitivity analysis, and
degenerate tests (Kelton et al, 2003). In determining the operational validity of SCODE,
all four of these methods were used and explained in detail below.

Extreme condition tests require input of values at the limit of operational
conditions to ensure the simulation system reacts appropriately (Kelton et al, 2003). For
example, exceedingly large or small quantities of orders were processed within the same
week, large changes in warehouse capacity entered, and unusually long transportation
delays created to examine the effects on SCODE. In each case, the program responded as
expected.

Internal validity checks are implemented by examining the internal stochastic
variability across several model replications (Kelton et al, 2003). SCODE offers the
ability to set a fixed random seed or variable random seed. For conducting internal
validity checks, the random seed was varied and the coefficient of variation measured
with the same set of input data. The model performed as expected, with the coefficient of
variation remained less than three percent across all measured variables. This low degree

of variation lends credibility to the operation of the simulation.

42

Parameter variability-sensitivity analysis involves varying known parameters to
determine the impact on the output (Kelton et al, 2003). This process was performed in
SCODE at numerous levels. By fixing the random seed and varying input, the expected
results were compared with the experimental results. In One case, the likelihood of
experiencing a transportation-related delay was significantly increased. As a result, the
supply chain experienced shortages of all raw materials, production was reduced, and lost
sales skyrocketed. Another case lowered labor hours required to produce each product.
SCODE output showed the profits soared as all demand was fulfilled at a fraction of the
previous production operating cost. Over a number of these test cases, SCODE
performed as expected, lending another degree of operational validity support.

The final technique employed degenerate tests to manipulate input parameters in a
known way to create constraints or extremes in the system (Kelton et al, 2003). As
SCODE’s warehouse capacity was decreased below the level of demand, the
manufacturing plants lacked storage for finished goods. The demurrage charges
escalated and lost sales increased. This display of expected results indicated passing
marks for SCODE’s ability to handle degenerative inputs.

The fourth and final validation category involves data validity. Expert review and
internal consistency checks are the two most common methods of ensuring data validity
(Kelton et al, 2003). The administrative settings in SCODE are obtained and updated
according to publicly available statistics and values. As one example, the costs of
transportation are checked monthly against figures released through the Department of
Transportation. Expert analysis of all variables present in the administrative settings

supported face validity of the constructs. Internal consistency checks similarly ensured

43

the absence of outliers or incorrect information encoded in the administrative settings of
the program.

SCODE underwent extensive pre-testing prior to initiating the experiment.
During development, SCODE evolved incrementally from basic design to its current
operational state. Separate teams of experts in computer programming, supply chain
management, computer simulation, and research design alternately implemented and
tested each facet of SCODE as well as validating other teams’ work. As the computer
programmers introduced a new feature to SCODE, the simulation and supply chain
management teams tested the functionality, incorporated the new features into existing
scenarios, and validated the feature performed as designed. Periodic presentations to a
research board and peer review also contributed to the verification and validation of
SCODE.

Experimental validation occurred with a group of users who had not previously
utilized SCODE, but were considered highly qualified in supply chain management
experience. These executive participants completed the entire experimental design as
outlined below with one additional feedback session focused on comprehension and
interpretation of the survey questions and the perceived effectiveness of the experiment to
capture research variables. Following this validation, minor questionnaire and scenario

design changes were implemented.

44

3.3 Research Design
This section focuses on the technical aspects of the study. Sections include the target
sample population, the pilot study, identification of the variables and controls, and

method of analysis.

3. 3.1 Experimental Population

The experimental participants include Master of Business Administration (MBA)
students attending one of three programs offered through Michigan State University. The
first session involved full-time traditional MBA students. The second included Executive
MBA students attending supply chain management courses in the evening. The third and
fourth sessions were comprised of Weekend MBA students attending a supply chain
management course over a series of weekends. All of the MBA students were pre-
screened to ensure a sufficient SCM background to have meaningfirl participation in this
study.

Power analysis and sample size calculations were performed using a statistical
calculator designed by Lenth (2006) based upon the calculations of Cohen (1975). Given
the 2 x 2 factor design of this study, a minimum sample size of 20 participants per block

is required to provide a power of at least 0.9 given an alpha level of 0.01.

3. 3. 2 Experimental Design
All SCODE sessions began with a one-hour presentation about a specific topic of
supply chain management interest. In this study, the topic is Global Sourcing. The

briefing (Appendix B) provided participants with a baseline level of knowledge about the

45

subject. The Global Sourcing training expanded upon the participant’s general
management knowledge by identifying current trends in research, demonstrating common
benefits and concerns about purchasing in a global environment, and enumerating best
practices seen in industries. Classroom discussion was encouraged for the users to share
their experience in the context of the session.

Following the topic lesson, users were introduced to SCODE through a hands-on
training session. This training session reinforced the topic area while introducing the
users to SCODE through a reduced-scale practice scenario. Again following the Global
Sourcing example, the scenario was tailored to include only four decision variables,
simplified assignment of manufacturing locations and capacities, fixed sales prices of raw
materials and finished goods, and additional assumptions to provide users an acclimation
opportunity with the new software. Users were guided through the design input interface,
the feedback information and location, and the decisions that are required to sustain their
supply chain. The decision variables followed the classroom discussion to enhance the
understanding of benefits and challenges to operating a supply chain in a global market.
Following repeated rounds, or simulated weeks, of training, users had an opportunity to
ask questions or discuss any issues with the software, scenario, or focused training area.
After this session, the users reach a level of proficiency to understand the cause-effect
relationship between their inputs and the feedback SCODE provides. The computers are
reinitialized and the recorded experiment begins.

The second scenario is more complex than the orientation problem and includes
ten decision variables, dynamic market demand, supplier shortages, production quality

deficiencies, and numerous other obstacles simulating industry challenges. This

46

increased level of difficulty provides users with a more realistic supply chain
environment facilitating generalization of the study results to broader supply chain
practices. A ten minute lecture introduces the specifics of the scenario and advises the
teams of the strict time table to be followed for this simulation setting. The scenario
hand-out is included in Appendix C.

The participants operate a production facility in North America with the
opportunity to manufacture four finished goods. Individuals may choose to limit the
quantity of finished goods in favor of a simplified supply chain, but at the expense of
potentially greater income. Such decisions are noted during data analysis and
incorporated into the performance evaluation and data analysis portion of the research.
Each participant has twenty minutes to create and input his or her supply chain design.
Facilitators provide enough assistance to ensure the users are able to input the supply
chain as designed, but do not provide advice on network creation. At the end of the
design phase, participants have ten minutes to enter his or her decisions for the first week.
The members simultaneously execute the first week’s run and again have ten minutes to
analyze the results, make decisions, and input the next week’s operational instructions.
This process continues until round six, when the time is reduced to five minutes between
rounds for all remaining rounds. As with the design, facilitators provide help identifying
the proper places to input the weekly operational decisions, but will not help locate

information from the output or interpret the data.

47

3. 3.3 Experimental Variables

Designing an experiment to measure differences in performance within an
environment requires manipulation of certain variables while holding others constant.
This approach increases the likelihood that any differences in measurement are due to
actual differences in the desired variables rather than outside random fluctuations. A full
factorial experiment involves manipulation of all experimental factors simultaneously
(Rodrigues, 2004). In this research, two factors are varied in two dimensions each. The
first is quantity of feedback data provided. The second is the quality of feedback
information provided. Each of the SCODE participants is randomly assigned to one of
four experimental groups. .

The first treatment group received all available SCODE “Output” and “Financial”
feedback data. Additionally, the decision information system on the “Weekly” worksheet
always provides accurate information to assist with the next week’s raw material ordering
requirements, labor force utilization, and warehouse inventory status. This group is
identified as the “Uncondensed / Corrected” group.

The second treatment group received the same feedback data as the first group.
However, the decision information system provides incorrect information to the team 30
percent of the time. The information may include the wrong quantity of a material to
order, mislead the team about the quantity of products capable of being produced in the
next week, or incorrectly report the usage rates within the warehouses. This group is
identified as “Uncondensed / Uncorrected” for complete data, but questionable
information. Figures 3.3 and 3.4 illustrate the management information dashboard with

correct and incorrect results for the same data set.

48

 

Figure 3.3: Management information dashboard with accurate feedback. When viewed
in color, the areas requiring immediate attention of the user are highlighted in red. For
example, this screen is telling the user to order 24,899 units of “Raw Material A” for the
South American plant immediately to meet the production requirements for the week.

 

 

lnformat ion Technology Interf ace

 

 

Needed to Meet Scheduled Production for Next Week

 

Raw Material North Amgrig _E_u_r_gpe Sgum Amgrigg Chin;
A 0 0 24899 27090
B 0 0 -269045 -9525
C 0 0 403238 4091
D 0 0 -8948 0
Percentage of Labor Force Scheduled for Next Week
North America Eurog South America China
0.0% 0.0% 88.4% 100.0%
Percentage of Raw Material Warehouse in Use for Next Week
02mm gums mm M
0.0% 0.0% 110.8% 96.4%
Percentage of Finished Goods Distribution Center in Use for Next Week
North America Eurog South America China India
0.0% 0.0% 0.0% 0.0% 0.0%
Percentage of Demand Satisfied This Week
North America Eurou South America China India
By Region: 62.4% 24.7% 25.8% 0.0% 0.0%

nguid Regular Liguld Premium Dy Regular Du Premium
By Product: 89.2% 4.4% 0.0% 57.9%

 

Performance Metrics

 

Weeks of Raw Material in the Warehouse (weekly) 0.6
Percentage of Production Completed (weekly) 33.4%
Weeks of Finished Goods in Warehouse (weekly) 0.0
Order Fulfillment (year to date) 47.5%

49

 

 

Figure 3.4: Management information dashboard with inaccurate feedback. The data
supporting this dashboard is identical to that in Figure 3.3. However, a single factor used
for creating the dashboard has been randomly altered. Instead of placing the correct
order (as in Figure 3.3 above), the user is recommended to place an order for 105,899
units of “Raw Material A” for the South American plant.

 

lnformat ion Technology lnterf ace

 

Needed to Meet Scheduled Production for Next Week

Raw Matgrial North America Europg Sguth Amgrig Chlng
A 0 0 105899 127090
B 0 0 -169545 90475
C 0 0 -64738 104091
D 0 0 9552 0
Percentage of Labor Force Scheduled for Next Week
North America Eurom South America China
0.0% 0.0% 132.7% 150.0%
Percentage of Raw Material Warehouse in Use for Next Week
Loam m ME! 91m
0.0% 0.0% 110.8% 96.4%
Percentage of Finished Goods Distribution Center in Use for Next Week
North America Europ_e_ South America China India
0.0% 0.0% 0.0% 0.0% 0.0%
Percentage of Demand Satisfied This Week
W M MM; 2111—"! Me
By. Region: 62.4% 24.7% 25.8% 0.0% 0.0%

Liguld Rggulgr Liguld Premium Du Rggular m PQmium
fly Product: 89.2% 4.4% 0.0% 57.9%

 

 

Performance Metrics

 

 

 

Weeks of Raw Material in the Warehouse (weekly) 0.6
Percentage of Production Completed (weekly) 33.4%
Weeks of Finished Goods in Warehouse (weekly) 0.0
Order Fulfillment (year to date) 47.5%

The error rate factor for the information management dashboard is based on

refinement through the validation process as well as previous research. Kang and

50

Gershwin (2008) evaluated the accuracy of inventory systems as part of their study to
reduce stock outs. In their review of 500 stores in various industries, the authors found
the highest level of inventory accuracy to be between 75 — 80%. The inventory
information systems averaged 51% accuracy across the stores. In a previous study,
Raman, Deoratius, and Ton (2001) investigated over 370,000 stock keeping units within
a single retailer and found over 65% of the inventory systems reflected inaccurate
quantities. The initial experimental validation session for this research utilized a 50%
error factor for the frequency of projecting incorrect information on the management
information dashboard. Participants provided feedback indicating this level seemed
unusually high and created the impression that the simulation program was not working
correctly. The users expressed doubts about. the accuracy of the data, the input screens,
and the delivery of raw material in addition to the information dashboard. The second
experimental validation session reduced the error rate to 30%, placing it at the lowest
level of Kang and Gershwin’s findings. The second experimental validation participants
were better able to differentiate between the accuracy of the data and the inaccuracy of
the management information dashboard than the first group. The second group did not
relate incorrect information on the dashboard with simulation problems. During the
feedback session at the end of the validation, the majority of users with incorrect
information correctly assumed this phenomenon was part of the learning environment
and expressed trust in the operating performance of the input, order processing, and
delivery mechanisms of the simulation.

The magnitude of the effect also combined experimental validation feedback with

previous information systems research. In their 1995 work, Rohleder and Scudder

51

investigated methods to maximize the net present value as a scheduling performance
measure. They hypothesized a significant concern with using such measures stemmed
from inaccuracies in two of the key factors comprising the metric. Their research
concluded that these performance metrics proved robust to error levels below 40%
magnitude of change. In other words, even though the contributing factors to the metric
varied by as much as 40% from accuracy, the metrics proved sufficiently resilient to
recommend the proper course of action. Therefore, an initial error factor of 50% of the
magnitude was selected for the experimental validation studies. This factor went largely
unnoticed by participants in the validation studies. Therefore, a final range of 50% -
200% of the correct value was selected. Excel selects a random number as a multiplier
for the dashboard and can provide erroneous information in the positive or negative
direction. SCODE processes the data required for the management information
dashboard by taking the week’s performance output data and comparing it with the next
week’s scheduled production and expected demand. For example, to calculate the raw

material required within a region for the next week, the equation is:

4
QAN = Z PNyRANy — WAN '- DAN
y=1
where: Q AN is the quantity of raw material A required in North America

Py is the quantity of finished good y scheduled for production for
the next week (Dry/Liquid and Premium/Regular)

RANy is the quantity of raw material A required to produce one unit
of finished good Py in North America

WAN is the quantity of raw material A currently available in the
North American warehouse

DAN is the quantity of raw material A forecast for delivery at the
start of the next week at the North American warehouse

52

The error term is introduced by altering RANy by a factor of 50% and rounding to the
nearest whole unit in the results. In a simplified example, if the user chose only a single
finished good to produce, “Dry Regular,” and operated a plant in South America, the
following calculations would be made. The order quantity to meet next week’s
production is the amount of product to be produced, say 10,000 units, multiplied by the
amount of raw material A required to produce one unit of “Dry Regular” finished good,
say 2. Therefore, the quantity of raw material A needed is 20,000 units. If the user
carried over 5,000 units from the previous week and expects a delivery of another 5,000
units, the management information dashboard will tell the user to order an additional
10,000 units to meet the next week’s schedule. For the users with an inaccurate
dashboard, SCODE multiplies the raw material requirement by 1.5 to yield a raw material
requirement of 30,000 units for the same 10,000 units of finished goods. Subtracting the
10,000 units in the warehouse and delivery status, the management information
dashboard tells the user to order 20,000 units to meet the next week’s schedule.

While the users in the experimental validation study were concerned about the
apparent system problems triggered by the management information dashboard
presenting erroneous information too often, the magnitude of the error was not a concern.
Therefore, it was unchanged for the second experimental validation session. This group
responded favorably to both the pattern and magnitude of the errors. The feedback
indicated the error magnitude was sufficient to trigger awareness, but not too exaggerated

to be unrealistic.

53

The third treatment group of users receives only condensed feedback data. Rather
than having full access to the data provided in the “Financial” and “Output” worksheets,
the teams receive a condensed version of the data only. The financial information by
region or product is not available and the specific cost centers are shown in aggregate
rather than in item-by-item detail. For example, Figures 3.5 through 3.8 presents the
uncondensed and condensed data from two different financial and output screens. The
decision information system is always accurate for this team. This group is identified as

“Condensed / Corrected” for condensed data and a correct information dashboard.

Figure 3.5: Production output data for team with uncondensed data.

 

r—__Te7m Number ' 1]

~‘* C4...) .4...,..~,..,;,1._,.;»,_.-.. 9.14.! PRODUCTION REPORT

 

 

 

Premium

Premium
Subtotal

 

Figure 3.6: Production output data for team with condensed data.

 

I Team Number 1]

PRODUCTION REPORT

 

 

 

 

 

Premium

Premium

 

54

 

Figure 3.7: Financial output data for team with uncondensed data.

  

SCM Budget

Less: Supply
Order Placement
Purchase Raw Materials
Purchase Price Variance
Transportation ( Supplier-Plant)
Inventory Carrying

Contract

Not Fulfilled

Total Supply

Less: Plant

Acceleration Charge

Excess P

roduction Charge

Production Variance

Charges to Alter Production Capacity
Warehousing

Demurrage

Plant Pro

duction Facility

Plant Facility
Charges to Alter Warehouse Capacity
Total Plant

Less: DC

Order Placement
Transportation
Inventory Carrying

Facility

Warehousing...
Demurrage
Charges to Alter Warehouse Capacity

Total DC

Less: Market
Transportation

Total SCM Expenses

SCM Contri

buuon

55

 

wrathmw
. I ' a. I‘ ‘ 1."

" M20 '

$0.00
$0.00
$0.00
$0.00
$25,61 1.72
$0.00
$25,61 1.72

$0.00

$52,548.05
$91,281.09

$143,829.15
$0.00

$0.00
$17,711.30

$2,465.46
$0.00

$20,176.76
$4,200.00
$193,817.63

$96,682.18

 

‘.a.

" $911

$119.70
$365,253.79
$31,762.47
$35,610.86
$64,318.03
$0.00
$497,064.84

$4,878.47

$126,851.51
$165,948.88

$297,678.86
$0.00

$0.00
$40,343.99

$7,422.72
$0.00

$47,766.71
$16,800.00
$859,310.40

$52,089.76

  

Figure 3.8: Financial output data for team with condensed data.

SCM Budget
Less: Supply
Total Supply
Less: Plant
Total Plant
Less: DC
Total DC
Less: Market
Transportation

   

Total SCM Expenses

SCM Contribution

,,
" .‘J’F‘. ’

$2901.81,848 '

$66,911.77
$155,524.75
$22,584.59
$0.00
$245,021.10

$1,567,827.80

8.0

$1,065,801.17
$444,220.14
$162,202.17
$0.00
$1,672,223.47

$4,669,935.43

The last treatment group receives only condensed information from the

“Financial” and “Output” reports as well as occasionally inaccurate recommendations

from the decision information system. This group is identified as “Condensed /

Uncorrected” for condensed data and questionable information. A summary of the

groups is shown in Figure 3.9.

56

Figure 3.9: Experimental groups.

 

 

 

 

 

 

 

 

'O
93
9
E .
g 8 Uncondensed Condensed/
«E / Corrected Corrected
.2
a
E
.9
E
C
.3 8 Uncondensed Condensed/
g g /Uncorrected Uncorrected
9
__2
Uncondensed Condensed
“Financial” and “Output” Feedback Data

3. 3. 4 Performance Variables

Six performance variables are recorded for every week of the experiment. These
metrics were previously developed in business literature (Martin and Stanford, 2007;
Robinson, Sahin, and Gao, 2008; White, 1996; Wacker, 1996; Bendoly, Rosenzweig, and
Stratman, 2007) as common metrics monitored by supply chain managers and validated
by members of the CLODE Research Board. These six are summarized in Table 3.2 and

explained more completely following.

57

Table 3.2. Dependent variables, their equations, and significance.

 

 

 

 

Performance Equation Measures condition of:
Variables

Weeks of raw Units RM in warehouse . .

. . . . . - Efficrency of inbound
materral 1n Unrts RM used 1n prevrous .
supply charn

warehouse week

Percentage of Units produced Effectiveness of

production completed

 

Units of production scheduled

production scheduling

 

Weeks of finished
goods in warehouse

Finished units in warehouse
Finished units shipped

Efficiency of outbound
supply chain

 

Sales to market

 

Demand satisfaction

 

 

 

 

Order Fulfillment Sales to market + Lost sales effectiveness
Net profit Amount of each sales
Net Profit Margin C ' dollar left after all
ompany revenue .
expenses have been pard
Ability to operate supply
SCM Contribution SCM budget — SCM expenses chain efficiently from end-

 

 

to-end

 

The first performance variable is the number of weeks of raw material available in

the warehouse. Excessive raw material inventory is one indicator of problems in the

supply chain. This safety stock incurs unnecessary expenses for inventory carrying costs

and variable warehouse costs. The most likely causes of inflated raw material storage is a

poorly managed inbound transportation system, a struggling manufacturing plant, or a

miscalculation of the anticipated demand for finished goods (Martin and Stanford, 2007).

The percentage of production completed metric compares the scheduled

production with the actual production. This ratio directly reports the ability of the team

to execute its production plan and indirectly reflects the efficiency of the supply chain. In

this simulation, there are only two reasons for the scheduled production to be greater than

58

 

the actual production. The first is if insufficient labor exists to manufacture the quantity
scheduled. The second is a shortage of raw material to complete the product. In either
case, the manufacturing plant is incurring expenses by planning a higher level of
production than it can complete. This metric is reflective of difficulties in forecasting
operational requirements with operational goals (Robinson, Sabin, and Gao, 2008; Joshi,
1982).

The third metric is reflective of the first. The lower number of weeks of finished
goods in the warehouse, the more efficient the supply chain performs. High levels of
finished goods also incur greater amounts of inventory carrying costs and variable
warehouse costs. The most likely cause of inflated finished goods is overestimation of
the demand for the product (White, 1996; Gunasekaran, Patel, and Tirtiroglu, 2001). It is
possible for situations of variable demand where the peaks exceed operational capability
that the safety stock could be mistaken for poor performance in this metric. However, in
the design of the scenario for the SCODE users, the plant production capacity exceeds all
demand throughout the simulation.

The order fulfillment metric records the percentage of total demand that is met
each week. At first glance, this appears to mirror the previous metric. However, there
are a number of settings where order fulfillment could be low despite high numbers of
finished goods available in the warehouse. Since there are multiple products offered in
the marketplace, a team could have high quantities of premium products in the
warehouse, but insufficient regular products to satisfy demand. Additionally, the teams
could have the finished goods in a warehouse serving different markets from those with

unfulfilled demand. The ability to meet customer demands is a central metric throughout

59

management and business literature (Kaplan and Norton, 2000; Wacker, 1996; Slater,
Olson, and Reddy, 1997).

The final two metrics are financial performance measures. The first is the supply
chain management contributiOn to the margin. This value represents the amount of the
supply chain budget that is returned to the firm. The second is the net profit margin. The
net profit margin represents the money remaining after all expenses are deducted.
Therefore, these two values represent the efficiency and effectiveness of the entire supply
chain (Blasberg, Vishwanath, and Allen, 2008; Bendoly, Rosenzweig, and Stratman,

2007)

3. 4 Perceptual Measures

An important portion of this experiment is to measure the supply chain manager’s
perceptions within SCODE. Following the final round of the scenario, the computers
utilized for the SCODE training are turned off and memory sticks containing the input
and output data are collected. While the data are being collected, each individual is
provided a survey (Appendix D) to report their experience. The survey contains two
distinct parts. The first contains open-ended questions utilized to facilitate nominal group
technique discussion at the end of the session. The second is adapted from previous
works in the information technology arena.

The open-ended questions in the first part of the survey elicit group discussion
and consensus. Nominal group technique is a brainstorming method encouraging input
by all members of a group. By recording the written responses of all participants, trends

become apparent and foster interaction among the individuals. By waiting until after the

60

experiment and questionnaires are complete for the group discussion, the researcher is
able to reveal the different levels of feedback data and accuracy of the decision
information system to the entire group, thus facilitating a rich discussion about the
required information to effectively make the next week’s operational decisions. This
group discussion is documented through recording information on a white board and on
the individual surveys. The experimental session is concluded following the group
discussion.

The perceptual measures recorded in the second part of the survey capture the
individual’s impression of the experiment. These items are derived from Speier (1996),
Bailey and Pearson (1983), and Spurrier, Topi, and Valacich (1994). The questions focus
on both the information qualities as well as the user’s confidence in his/her performance
using the information. This portion of the survey captures feedback essential to
determining the usefulness of the condensed data versus the uncondensed data, the
perception of reliability of the decision information system, and the user’s ability to grasp
the information and utilize it effectively. The constructs, specific questions comprising

the construct, and derivation source are presented in Table 3.3.

61

Table 3.3. Survey Constructs

 

 

 

 

 

 

 

 

 

 

 

 

Constructs Questions Derived from
25:22:;22 7A 7327“: 7D Bailey and Pearson (1983)
Information Precision 8A 8B 8C 8D Bailey and Pearson (1983)
[aggrllfigfyn 9A 9B 9C 9D Bailey and Pearson (1983)
Conigfzhrengiisgness 10A 10B2210C 10D Bailey and Pearson (1983)
Information Format 11A 11B 11C 11D Bailey and Pearson (1983)
13:31:22: 12A 13133 1226C 12D Bailey and Pearson (1983)
Confidence 61‘8‘13613166C176D 81:11:11;ng IP99?)1d
Rational Approach 15 20 25 $3121.12: at? €113 91,9?)1d
Intuitive Approach 18 21 85:12:; a: €11) 91,923“!
13:22:18,123: 14281929223024 Speier (1996)

 

 

 

62

 

3.5 Data Analysis

Data analysis occurs in three parts. The first is multivariate analysis of variance
(MANOVA) of the weekly decisions. The second phase is analysis of variance
(ANOVA) for the scaled survey questions. Third, the short-answer and group discussion
information are examined by post-hoc analysis for trends. Statistical analysis of results
divided by executive or MBA status captures and controls for any potential differences
between the groups.

MANOVA enables analysis of variance for multiple dependent variables. The
dependent variables in this study include the weekly output variables provided at each
SCODE week’s end. Since there is no single indicator of a “health of the supply chain”
available in SCODE, evaluation of variables is used to identify whether the supply chain
is operating efficiently and effectively. Table 3.2 summarizes the six dependent variables
evaluated in Phase 1 of data analysis.

ANOVA is performed on questions from the survey. By retaining group identity
in the analysis of these responses, the four clusters representative of condensed /
corrected, condensed / uncorrected, uncondensed / corrected, and uncondensed/
uncorrected are compared for statistical differences from and similarities to the entire
population. Comparing the variance between the groups with the variance within each
group will provide insight to the perceived difference in information availability and
accuracy.

The post-hoc analysis captured the remaining experimental data in a holistic

perception of the four groups. The richness of information provided by open-ended

63

questions and group discussion significantly augments the quantitative data collected

throughout the experiment. The detailed analysis of the data follows in Chapter 4.

64

CHAPTER IV: RESULTS

4.0 Introduction

This chapter presents the results of the study. The first section introduces the
experimental population and reveals the sample size, demographics, and assignment to
experimental groups. The next sections explore the statistical methods for analysis, the
assumptions required to utilize these methods, and the verification checks within the
sample data sets. After ensuring the data fits the assumptions for the methods, the
hypotheses are tested and post-hoe analysis highlights further stratification of the

variables under study.

4.1 Experimental Population Descriptive Data

The subjects in this study are currently enrolled in a MBA program at Michigan
State University. The students represent three variants of the MBA options offered
through Michigan State University. One session included traditional MBA (F-TMBA)
students enrolled full-time in courses. Another session involved Executive MBA
(EMBA) students attending courses in the evening. The final two sessions enrolled
Weekend MBA (WMBAl and WMBA2) students who attend courses on the weekend.
All participants completed at least one course in supply chain management prior to
inclusion in this study.

A total of 172 subjects participated in this study. Twelve of the subjects were
excluded from analysis due to various causes. Two participants failed to complete
enough weeks of SCODE to provide meaningful results. Six subjects did not return a

questionnaire to complete the required data collection. Two subjects did not properly

65

identify his/her group number on the survey for identification with the computer

simulation. The final two subjects were excluded because he/she did not submit the

correct files from the computer simulation for study. The division of participants by

session is in Table 4.1.

Table 4.1. Participants by Session

 

 

 

 

 

 

 

Session
Number of Participants Percent
Executive MBA , 42 ,, , 26.3
Full-Time MBA 29 18.1
Weekend MBA 1 42 26.3
Weekend MBA 2 47 29.4
Total 160 100.0

 

 

The included 160 subjects were randomly assigned to one of the four treatments within

the experiment. The division of subjects by session is presented in Table 4.2.

Table 4.2. Assignment of Treatment by Session

 

 

 

 

 

 

 

 

 

 

 

Group Assignment
EMBA F-T MBA- WMBA1 WMBAZ Frequency Percent
Uncondensed / Corrected _9_ _ 6 15 8 _ 38 23.8
Uncondensed / Uncorrected 13 1 1 8 12 44 27.5
Condensed / Correct 12 5 7 11 35 21.9
Condensed / Uncorrected 8 7 12 16 43 26.9
Total 42 29 42 47 160 100.0

 

Chi-square tests investigated possible differences among group assignment and

format of MBA education. The chi-square statistic was 9.82 (d.f. = 9) with a p-value of

66

 

.365. This indicates no significant differences among the treatments based the
population’s method of MBA education.

Of the 160 subjects in the study, 48 females and 112 males are included. The
treatment by gender appears in Table 4.2. As with the MBA format, chi-square tests
(1.99, d.f. = 3, p-value = .575) revealed no significant differences among the treatments

based upon the gender of the participants within each group.

Table 4.3. Assignment of Treatment by Gender

 

 

 

 

Group Assignment
Female Male Frequency Percent
Uncondensed / Corrected 9 29 38 23.8
Uncondensed / Uncorrected 16 28 44 27. 5
Condensed / Correct 13 22 35 21.9
Condensed / Uncorrected 10 33 43 26.9
Total 48 112 160 100.0

 

 

 

 

 

 

 

Participants represented a wide variety of backgrounds. The self-reported fields
of the survey included questions about the current position, years of supply chain
experience, and highest degree earned to date. The diverse job positions represented in
Table 4.4 are then clarified to show a more equal diversification of job-related experience
among the participants. Table 4.5 condenses the current position to supervisory levels
within the participant’s respective organization. The results follow in Tables 4.4 and 4.5.
The chi-square tests for supervisory position within the treatments resulted in a test

statistic of 6.76 (d.f. = 12, p-value = .873).

67

 

 

Table 4.4. Participant Employment Position

 

Participant Position

 

 

 

 

 

 

 

 

 

 

 

 

Frequency Percent
Accountant 4 ' 2.5
Analyst 11 6.9
Associate 1 0.6
Banker 1 0.6
Buyer 1 0.6
CEO 1 0.6
CIO 2 1.3
Consultant 3 1.9
Director 6 3.8
Engineer 22 13.8
Finance 1 0.6
Leader 71 7 g 0.6
Management 47 29.4
Marketing fl ”1 7 70.6 7
President 1 0.6
Sales 4 2.5
Specialist 1 1 6.9
Student 37 23.1
Supervisor 1 0.6
VP 4 2.5
Total 160 100.0

Table 4.5. Participant Employment Level
Participant Position By Supervisory Level

Frequency Percent
Senior Management 14 8.1
Management 49 30.6
Employee 60 37.5 .
Student 37 23.1
Total 160 100.0

 

68

 

 

The self-reported experience level of participants varied greatly. The vast
majority of subjects, 65.6%, reported no supply chain management experience. The
division of subjects by experience across the treatment groups again reflects no statistical
differences (chi-square statistic = 35.21, d.f. = 33, p-value = .364). The results are

summarized in Table 4.6.

Table 4.6. Participant Years of Supply Chain Experience

 

 

 

 

Participant Years of Supply Chain Experience
Frequency Percent
0 105 65.6
1 11 6.9
2‘ ,, 8 7 5.0
3 7 4.4
4 4 2.5
5 10 6.3
6 2 1.3
7 2 1.3
8 5 3.1
10 2 1.3 ‘
15 3 1.9
20 1 0.6
Total 160 100.0

 

 

 

 

 

The final report of demographic information represents the highest degree earned
by the participants. The chi-square statistic of 6.13 with d.f. = 9 and a p-value of .727
indicates no statistical difference among the treatments based upon educational level

attained to date.

69

Table 4.7. Participant Years of Supply Chain Experience

 

 

 

 

 

 

Participant Highest Degree Earned
Frequency Percent
High School 5 3.1
Bachelor's 87 g 54.4
Master's 65 40.6
Doctoral 3 1.9
Total 160 100.0

 

 

 

 

4.2 Statistical Methods and Tests of Assumptions

MANOVA can be a powerful tool to compare groups created through assignment
of different treatments to different groups. There are five assumptions for MANOVA
results to be accurate. The first is that observations must be statistically independent
from each other. Second, the dependent variables should follow a multivariate normal
distribution. Third, the variance-covariance matrices of the dependent variable should be
equal for all treatment groups. Fourth, the dependent variables must exhibit linearity and
multicollinearity. Fifth, there should not be significant outliers in the population.

In the case of this study, each observation is unique to the participant using the
SCODE simulation. This is one of the more stringent assumptions for MANOVA.
Typical violations of independence occur if repeated measures are taken for a single
participant, as in a time-series study, or if a subset of the population is likely to have
similar traits that are distinct from other members. As shown in the demographic tables
in the first section of this chapter, the participants are in similar status within their MBA
programs and each treatment represents a statistically uniform cross-section of the
population based on gender, type of MBA program, and supervisory level. Therefore,

independence of observations is met.

70

Multivariate normality refers to the assumption that the joint effect of two
variables results in a normal distribution. For the case of MANOVA with large sample
sizes, violations of this assumption have minimal impact on the effectiveness of the tests
(Hair et al, 1998). Concern is warranted if non-normality is caused by outliers rather than
skewness. The skewness and kurtosis statistics (Table 4.8) as well as frequency

histograms (Figure 4.1) illustrate the variance from normality within the dependent

 

 

 

 

 

variables.
Table 4.8. Skewness and Kurtosis Statistics

Skewness and Kurtosis of Dependent Variables

Inbound Internal Outbound Market Profit SC Contrib
Skewness 1.318 -1 .245 2.276 -0.586 -2.461 -1.581
Std. Error of Skewness 0.192 0.192 0.192 0.192 0.192 0.192
Kurtosis 0.886 0.331 4.936 -0.685 ,. 7 7.891 9.3047
Std. Error of Kurtosis 0.381 0.381 0.381 0.381 0.381 0.381

 

 

 

 

 

 

 

 

The third assumption of MANOVA is the equality of variance-covariance
matrices. The assumption addresses substantial differences in the degree of variance
within a single group compared to the others. As with the multivariate normality
assumption, MANOVA is robust to violations of the variance-covariance matrices
equality for equal sample sizes. For these data, Levene’s Test of homogeneity of
variances returns significant results for all dependent variables at p<.05. Additionally,
Box’s M-test results (Box’s M = 317.8, d.f. 1 = 63, d.f. 2 = 53,614, p < .001) are
significant indicating the covariance matrices of the dependent variables are not equal
across all treatment groups.

The frequency histograms in Figure 4.1 graphically depict the absence of

significant outliers from the population. This satisfies the fifth MANOVA assumption.

7]

Figure 4.1. Frequency Histograms for the Dependent Variables

 

Frequency

 

 

 

 

 

 

 

 

l
0.0 2.0 4.0 6.0

 

 

 

 

 

 

Inbound
1005
80-
a
c 60'
e
:l
e-
2
u.
40-
20—
.. r":
V I I I I T
o 20 40 60 60
Internal

72

Figure 4.1 (cont’d).

 

w-

60—

Frequency

40...

 

 

 

 

 

v I I I

 

 

 

 

 

 

T
l 2 4 6 3
Outbound
fro-1
@—
6
r:
O
3
6' 20-
O
L-
ll.
10-
3 r
o o

 

 

Figure 4.1 (cont’d).

 

Frequency

 

 

 

 

 

 

 

 

 

Frequency

 

 

 

 

 

 

I I 0
4mm 5000000 I

 

SC Contrib

74

l
50111300

1
10000000

 

The final assumption of MANOVA is the linearity and multicollinearity among
dependent variables. For general linear hypotheses models to correctly predict
relationships among the variables, it is important that the variables have a linear
association with other variables. Bartlett’s test of sphericity is used to evaluate the
significance of intercorrelation among the dependent variables. The result indicates that
significant intercorrelations do exist within the variables. Bartlett’s factor of 196.37 with
d.f. = 15 yields a p < .001 for the experimental results.

As a result of the assumption testing in this section, MANOVA is an appropriate
method to evaluate the effect the treatment had on the subject’s ability to perform in a

supply chain environment.

4.3 Two-way Overall Results

This section progresses from the tests of the data ensuring assumptions of the
statistical tools to analysis of the results from the experimental sessions. The four subject
treatment groups are based on the manipulation of two variables: corrected information
and condensed data. Each of the experimental variables has two levels as discussed in
Chapter 3. The participant’s performance was evaluated on six distinct metrics
measuring the effectiveness of various aspects of the supply chain. The six metrics
include the efficiency of the inbound supply chain (Inbound), the effectiveness of
manufacturing operations (Internal), the efficiency of the outbound supply chain
(Outbound), the market satisfaction level (Market), net profit margin (Profit) for the

entire operation, and the supply chain contribution to the overall profit (SCContrib). A

75

seventh overall metric utilized a standardized calculation of performance across all six
metrics to determine overall supply chain performance by utilizing a weighted z-score.
Table 4.9 provides more detailed description of the measurement method and the

interpretation of each of the variables.

Table 4.9. Performance Variable Scoring

 

Performance

Variable Measurement Method Interpretation

 

Weeks of raw material in
Inbound warehouse based on current Lower (zero) is better
production requirement

 

Percentage of production
Internal scheduled for current week that Higher (100%) is better
was actually completed.

 

Weeks of finished goods in
Outbound warehouse base on current Lower (zero) is better
market demand requirement.

 

Market Percentage .Of demand met in Higher (100%) is better
prevrous week.

 

Net profit margin from the
Profit beginning of the simulation to Higher (100%) is better
the ending state.

 

50% of the firm’s revenue is
allocated for supply chain
expenses. This measures the
amount under budget (positive)
or over budget (negative) that
the supply chain returns.

SCContrib Higher is better

 

 

Equally weighted scaled
combination of the above six
metrics. Provides an overall . .

. H1 r1 tter
performance metric based on ghe S be
deviation from the average
scores of each metric.

Overall

 

 

 

76

 

A final overall supply chain score (Overall) is computed based upon the equally
weighted sum of the deviation from average values of each of the six metrics. This
seventh metric provides an overall performance measure for the behavior of the entire
supply chain over the simulation and provides an accurate gauge for supply chain
behaviors beyond the experimental setting. For example, participants identified a supply
chain goal at the beginning of the session. Most (55.0%) selected “Meet all demand” as
the goal of their supply chain. This approach would be expected to earn the highest
profits by meeting a higher percentage of demand, but meeting greater challenge with
managing the inbound, internal, and outbound supply chain performance which could
lead to lower supply chain contributions. Others created a more focused or efficient
approach that might sacrifice profits for simplicity. It is expected that those groups may
return a lower profit margin, but should have less difficulty maintaining an efficient
inbound and outbound flow of goods and therefore returning greater supply chain
contributions. Because of these trade-offs inherent in operating a supply chain, there is
merit to comparing the performance of the subjects under different treatments with the
six performance variables as well as separately with an overall combined supply chain
variable. The participant’s supply chain goals were spread statistically equally across the

treatments (chi-square value = 22.98, d.f. = 15, p = .085).

77

Table 4.10. Participant’s Supply Chain Goals

 

 

 

 

 

Supply Chain Goal
Frequency Percent
All 2 1.3
Demand 88 55.0
Efficiency 34 I 21 .3
Focused 22 13.8
None 1 0.6
Profit 13 8.1
Total 160 100.0

 

 

 

 

Analysis of the data was performed utilizing the SPSS software package, version
15.0, with conversion to Excel for presentation of tables and some figures. There are four
criteria presented for testing multivariate differences between groups. The criteria
include Pillai’s Trace, Wilks’ Lambda, Hotelling’s Trace, and Roy’s Largest Root.
Pillai’s Trace is the sum of variance that can be explained through calculation of
discriminant variables and is robust in conditions where MANOVA assumptions have not
been met. Wilks’ Lambda determines whether there is an overall difference between the
levels of the independent variable relative to the dependent variables when considered as
a whole. Wilks’ Lambda is also generally utilized when there are more than two
dependent groups under study. Hotelling’s Trace is more commonly used for two
dependent groups and includes the pooled ratio of effect variance to error variance.
Roy’s Largest Root is generally most sensitive if population centers differ along a single
dimension and is least sensitive otherwise. However, in cases where sample sizes are
large, all four of these become asymptotically equivalent (Hair et al, 1998). As can be
seen in Table 4.11, all return significant (p < .001) results for the difference in

performance variables among the four experimental treatment levels.

78

 

 

 

Table 4.11. Multivariate Test Results for Treatment Levels
Multivariate Test Results
F Hyp df Error (If Sig. Part Eta Sq Power(a)
Pillai's Trace 2.675 18 459 <0.001 0.095 0.998
Wilks' Lambda 2.726 18 428 <0.001 0.097 0.997
Hotelling's Trace 2.767 18 449 <0.001 0.100 0.998
Roy's Largest Root 5.847 6 153 <0.001 0.187 0.997

 

 

 

 

 

 

 

 

 

Computed using alpha = .05

The additional columns of the table include the Partial Eta Squared and Observed
Power of the test. The Partial Eta Squared statistic measures effect size by identifying the
proportion of total variability in the dependent variables that are accounted for by
variation of the independent variable. The possible values range from 0, no effect, to 1, a
very strong effect (Cohen, 1988). The Observed Power presents the ability of the test to
reject the null hypothesis when false. Power in excess of .9 represents a strong test
(Cohen, 1988).

The multivariate test results in Table 4.10 represent the highest level of analysis.
The significance of the multivariate tests indicate a statistical difference in performance
across the six dependent variables based on the four different treatments of corrected or
uncorrected information and condensed or uncondensed data. To better understand the
effects within each of the treatment groups, additional analysis is required. The next step
in analysis involves evaluating the between-subjects effects. This series of tests identifies
the specific dependent variables affected by treatment level. As can be seen in Table

4.12, four of the seven metrics show significant differences in performance based upon

79

treatment. The inbound and internal metrics are significant at p < .01 level. The profit

and overall metrics indicated significance at p < .05 level.

Table 4.12. Tests of Between-Subject Effects for Treatment Levels

 

 

 

 

 

 

 

 

 

Tests of Between-Subjects Effects
Dependent Variable F at Sig. Part Eta Sq Power(a)
Inbound 4.409 3 0.005 *** 0.078 0.868
Internal 4.377 3 0.005 *** 0.078 0.865
Outbound 1.626 3 0.185 7 0,030 7 0.421
Market 1.670 3 0.176 . 0.031 0.431
Profit 2.844 3 . 0040?: 7 70.052 0.673
SC Contrib 1.769 3 0.155 0.033 0.455
Overall 3.720 3 0.013 ** 0.067 0.799

 

 

a Computed using alpha = .05

For these four dependent variables with significant differences at the treatment level, the
mean value of each metric by treatment is graphed and further analyzed to identify which
specific treatments are statistically different from others. Figure 4.2 displays the inbound

dependent variable by treatment.

80

Figure 4.2. Mean Value of Inbound Dependent Variable by Treatment

Inbound Metric

2.5 _______._..__.. ..______.____ . ..___ .----____...___ ......----_-_-.-_...-._....__-._2.-..__-_-_____.___..._..-_.--.___.-._.._....__.----.__.__2;

1.5

Weeks of Raw Material In Warehouse

 

 

Uncondensed / Uncondensed/ Condensed I Corrected Condensed /
Corrected Uncorrected Uncorrected

From visual examination, Figure 4.2 shows a substantial difference between the
uncondensed / corrected treatment and the condensed / corrected-treatment. To validate
this statistically, the Tukey HSD test is used. The Tukey HSD test performs a series of
pair-wise comparisons of the mean values of each treatment. For the inbound dependent
variable, the results are presented in Table 4.13. The Tukey evaluation confirms the
visual indication from Figure 4.2, that the mean values are statistically different for the
uncondensed / corrected treatment and the condensed / corrected treatment. In addition,
the Tukey HSD test confirms the uncondensed / corrected treatment returns significantly
different performance on the inbound metric than the condensed / uncorrected treatment.

This series of findings indicates both uncondensed data and corrected information are

81

efficient management of the inbound supply chain.

necessary to statistically improve the performance of the ordering, inventory levels, and

Table 4.13. Tukey HSD Test for Inbound Dependent Variable by Treatment

 

 

Tukey I-ISD Multiple Comparisons

 

 

DV (1) Group (J) Group Sig.
Inbound Uncondensed / Corrected Uncondensed / Uncorrected 0.185
Condensed / Corrected 0.005 *“
Condensed / Uncorrected 0.022 **
Uncondensed / Uncorrected Uncondensed / Corrected 0.185
Condensed I Corrected 0.439
Condensed / Uncorrected 0.797
Condensed / Corrected Uncondensed I Corrected 0.005 ***
Uncondensed I Uncorrected 0.439
Condensed / Uncorrected 0.922
Condensed / Uncorrected Uncondensed / Corrected 0.022 **
Uncondensed / Uncorrected 0.797
Condensed / Corrected 0.922

 

 

 

 

 

 

 

 

Next, the internal metric is analyzed in greater detail. Figure 4.3 provides the

82

graphical depiction of the values of the dependent variable at different treatment levels.

Figure 4.3. Mean Value of Internal Dependent Variable by Treatment

Internal Production Metric

90 l
80
70
60
50
40
30

20

Percentage of Production Completed

10

 

 

Uncondensed / Uncondensed/ Condensed I Corrected Condensed I
Corrected Uncorrected Uncorrected

As with the inbound metric, the internal production metric appears to differ
between the uncondensed / corrected treatment and both the condensed / corrected and
the condensed / uncorrected treatments. Table 4.14 confirms both visual findings. The
test indicates that a firm would need to expand the level of detail of the data and improve
the quality of the information in order to make a significant improvement in performance

for scheduling labor and raw material within the production facility.

83

Table 4.14. Tukey HSD Test for Internal Dependent Variable by Treatment

 

Tukey HSD Multiple Comparisons

 

 

 

 

DV (1) Group (J) Group Sig.
Internal Uncondensed / Corrected Uncondensed / Uncorrected 0.528
Condensed / Corrected 0.026 **
Condensed ’/ Uncorrected 0.008 ***
Uncondensed / Uncorrected Uncondensed / Corrected 0.528
Condensed / Corrected 0.382
Condensed / Uncorrected 0.220
Condensed / Corrected Uncondensed / Corrected 0.026 **

Uncondensed / Uncorrected 0.382
Condensed / Uncorrected 0.996

 

 

 

 

 

 

Condensed / Uncorrected Uncondensed I Corrected 0.008 ***
Uncondensed / Uncorrected 0.220
Condensed / Corrected 0.996

 

The profit dependent variable showed significant values in Table 4.11. The graph
of the mean value of the profit metric is presented in Figure 4.4. For this metric, the
figure demonstrates the uncondensed / uncorrected and condensed / corrected treatments

have mean values the furthest apart.

84

Figure 4.4. Mean Value of Profit Dependent Variable by Treatment

Profit Metric

50 »
45
4o
35
3o
25

20

Net Profit Margin

15

10

 

 

Uncondensed / Uncondensed I Condensed I Corrected Condensed /
Corrected Uncorrected Uncorrected

Performing the Tukey HSD test for the profit dependent variable across all
treatment levels returns significant differences between the uncondensed / uncorrected
and the condensed / corrected treatments. This finding is interesting because it provides
some evidence that a firm may be able to increase performance level for net profit margin
by providing supply chain managers with a less detailed level of data as long as the

information system provided accurate decision making metrics.

85

Table 4.15. Tukey HSD Test for Profit Dependent Variable by Treatment

 

Tukey HSD Multiple Comparisons

 

 

 

 

DV (I) Group (J) Group Sig.
Profit Uncondensed / Corrected Uncondensed / Uncorrected 0.125
Condensed ‘/ Corrected 0.954
Condensed / Uncorrected 0.924
Uncondensed / Uncorrected Uncondensed / Corrected 0.125
Condensed I Corrected 0.037 **
Condensed / Uncorrected 0.361
Condensed / Corrected Uncondensed / Corrected 0.954

Uncondensed / Uncorrected 0.037 **
Condensed / Uncorrected 0.661

 

Condensed / Uncorrected Uncondensed / Corrected 0.924
Uncondensed I Uncorrected 0.361
Condensed / Corrected 0.661

 

 

 

 

 

 

The overall combined performance metric also was significant in between-subject
effect testing. Figure 4.5 illustrates the overall scaled performance across all treatments.
As with the first two metrics, the overall dependent variable shows uncondensed/
corrected and condensed / uncorrected as the treatments with the greatest difference in

means.

86

Figure 4.5. Mean Value of Overall Dependent Variable by Treatment

Overall Performance Metric

0.8

0.6

0.4

0.2

Scaled Performance

 

 

Uncondensed / Uncondensed / Condensed I Corrected Condensed /
Corrected Uncorrected Uncorrected

Further analysis using the Tukey HSD test confirms the uncondensed / corrected
and condensed / uncorrected treatments have a statistically significant difference in
means. There is no surprise that the largest difference in overall performance would be
between the treatment possessing more extensive data availability and accurate
information feedback systems and the treatment possessing less data and inaccurate
information systems. However, the lack of statistical difference of the other two
treatments is surprising. This series of discoveries warrants further investigation through

exploring each of the two factors separately. This analysis is covered in Section 4.4.

87

Table 4.16. Tukey HSD Test for Overall Dependent Variable by Treatment

 

 

 

 

 

 

Tukey HSD Multiple Comparisons
DV (1) Group (J) Group Sig.__
Overall Uncondensed / Corrected Uncondensed / Uncorrected 0.106
Condensed/ Corrected 0.344
Condensed / Uncorrected 0.007 ***
Uncondensed / Uncorrected Uncondensed / Corrected 0.106
Condensed / Corrected 0.957
Condensed I Uncorrected 0.726
Condensed / Corrected (Uncondensed / Corrected fig 0.344 7 J
Uncondensed / Uncorrected 0.957
Condensed I Uncorrected 0.449
Condensed / Uncorrected Uncondensed / Corrected 0.007 ***
Uncondensed / Uncorrected 0.726
Condensed / Corrected 0.449

 

 

 

 

 

 

4.4 Main Effect Results

Section 4.3 provided insight into the effects the combined treatment had on the
seven performance variables. This section focuses on each of the treatment variables
separately to evaluate the specific effects that condensed / uncondensed data and
corrected / uncorrected information systems have upon the seven dependent variables.

As with the two-way interactions, the analysis begins with the multivariate tests.
Table 4.17 provides the multivariate test results for the main effects. The condensed /
uncondensed treatment reports significant differences across the seven performance
metrics. The corrected / uncorrected effect does not show significant impact across all of
the metrics simultaneously. However, further investigation reveals significant effect
within certain metrics as will be investigated later in this section. Due to the sample size
and degrees of freedom of the test, the four multivariate tests asymptotically approach the

same values.

88

Table 4.17. Multivariate Test Results for Main Effects

 

Multivariate Tests c)

 

 

 

IEffect F Hyp of Error df Sig Part Eta Sq Power(a)
Corrected I Pillai's Trace 1.589 6 151 0.154 0.059 0.598
Uncorrected Wilks' Lambda 1 .589 6 151 0.154 0.059 0.598

Hotelling's Trace 1 .589 6 '151 0.154 0.059 0.598
RoYs flesf Root 1.589 6 151 0.154 0.059 0.598
Condensed I Pillai's Trace 5.344 6 151 <0.001 0.175 0.995
Uncondensed Wilks' Lambda 5.344 6 151 <0.001 0.175 0.995
Hotelling's Trace 5.344 6 151 <0.001 0.175 0.995
Roy's Largest Root 5.344 6 151 <0.001 0.175 0.995

 

 

 

 

 

 

 

 

 

a. Computed using alpha = .05

Further analysis through tests of between-subject effects provides additional

clarification to the significant differences in dependent variables for each of the main

effects. Figure 4.18 presents these test results.

Table 4.18. Tests of Between-Subject Effects for Main Effects

 

 

 

 

 

Tests of Between-Sub'ects Effects
Source Dependent Variable F df Sig. Part Eta Sq Power(a)
Corrected I Inbound 0.912 1 0.341 0.006 0.158
Uncorrected Internal 1.230 1 0.269 0.008 0.197
Outbound 3.622 1 0.059 0.023 0.473
Market 0.009 1 0.926 0.000 0.051
Profit 5.591 1 0.019 ** 0.035 0.652
SC Contrib 2.387 1 0.124 0.015 0.336
Overall 7.061 1 0.009 *** 0.043 0.752
Condensed I Inbound 9585 1 0.002 *** g 0.058 g 0.868
Uncondensed Internal 11.500 1 0.001 *** 0.069 0.921
Outbound 0.592 1 0.443 0.004 0.119
Market 2.534 1 0.113 0.016 0.353
Profit 2.221 1 0.138 0.014 0.316
SC Contrib 0.127 1 0.722 0.001 0.065
Overall 3.734 1 0.055 0.023 0.484

 

 

 

 

 

 

 

a Computed using alpha = .05

89

 

The two main variables show very different effects on the dependent variables.
The corrected / uncorrected variable creates significant differences in the outbound,
profit, and overall metrics. The condensed / uncondensed metric produces significantly
different results in the inbound, internal, and overall metrics. Reviewing the mean values
of each group on the same graph better illustrates the relationships. Figures 4.6 — 4.10
graphically depict the dependent variables with one or both main effects showing

significant differences in the mean values of the results.

Figure 4.6. Mean Value of Inbound Dependent Variable by Main Effect

 

 

 

Inbound Metric

2.5 ,9 , , , . vvvvvv
0
§ 2
2
f
3
E 1.5
(I
t
3
‘9‘
3 1
II
I!
"6
in
'§ 0.5
3

o

No Yes

+Condensed? - I- Corrected?

The graph illustrates the significant difference with the condensed-yes versus the

condensed-no (uncondensed) weeks of raw material in the warehouse. The accuracy of

90

the information system had little effect on the performance by itself. However, the
participants receiving condensed data for ordering raw materials were able to Operate the
supply chain more efficiently on the inbound side of the process. By reducing the
extraneous data available to the uncondensed users, the condensed users performed at a

higher level on this metric.

Figure 4.7. Mean Value of Internal Dependent Variable by Main Effect

Internal Production Metric

90

G)
(II

(D
O

 

\l
0"

Percentage of Production Completed
V
O

 

 

65
No Yes

+Condensed? - I- Corrected?

The internal production metric shows a significant difference between the
condensed / uncondensed treatment while the corrected / uncorrected treatment again
shows no statistically significant difference in means. As in the raw material ordering

and storage metric, the removal of extraneous data should allow the subjects to focus

9]

attention on the critical data needed for decision making regardless of the accuracy of the
management information system. The average participant in the condensed data
treatment completed over 88% of the scheduled production as compared to under 75%

for the average user in the uncondensed data treatment group.

Figure 4.8. Mean Value of Outbound Dependent Variable by Main Effect

Outbound Metric

12

0.8 is.

0.6

0.4

Weeks of Finished Goods In Warehouse

 

 

No Yes

.—e—Condensed? - I- Corrected?

The graph of the outbound metric reflects the opposite trend as seen in the last
two figures. The corrected / uncorrected treatment groups performed statistically
differently within the outbound metric while the condensed / uncondensed treatment

groups did not. The participants with corrected information in their management

92

dashboard performed statistically better at managing their finished goods inventories and

matching production with demand.

Figure 4.9. Mean Value of Profit Dependent Variable by Main Effect

Profit Metric

40

 

Net Profit Margin
N N 00
O 01 O

..s
01

10

 

 

No Yes

+Condensed? - l- Corrected?

The graphical depiction of the profit metric also shows a greater difference in the
mean values of the corrected / uncorrected treatment than the condensed / uncondensed
treatment. The participants in the group with corrected information on the management
dashboard performed significantly better in terms of net profit margin than those with

uncorrected information.

93

Figure 4.10. Mean Value of Overall Dependent Variable by Main Effect

Overall Performance Metrics

0.5
0.4
0.3
0.2
0.1

-0.1

Scaled Performance

-0.2
-0.3
-0.4

 

 

No Yes

.—e—Condensed? - I- Corrected?

For the overall supply chain performance metric, the condensed / uncondensed
treatment group and the corrected / uncorrected treatment group performed significantly
differently. The participants with uncondensed data recorded an overall higher score
across all six metrics than the participants with condensed data. Further, subjects with
corrected information on the management dashboard performed better across all six
metrics than the subjects with uncorrected information.

The study of main effects individually provides some greater insight into the
behavior of the participants within each treatment group. The last analysis required prior
to evaluating the hypotheses involves evaluating the written survey and nominal group

technique discussion at the conclusion of each experimental session.

94

4.5 Survey and Nominal Group Technique Results

The survey and nominal group technique portion of the session provided
additional insight into the perception of the subjects and revealed additional significance
to the numerical analysis presented in previous sections of this chapter. By
understanding more about the decision making process of supply chain leaders, the
argument for condensed versus uncondensed data and corrected versus uncorrected
information becomes less of a binary decision and more of an understanding of the
balance of the two variables along with complexity and cost of the decision information

system.

4. 5.] Survey Results

As presented in Table 3.3, the survey items werederived from constructs
established in previous studies (Speier, 1996; Bailey and Pearson, 1983; Spurrier, Topi,
and Valacich, 1994). Analysis of the survey results followed the same pattern as the
simulation results. Namely, the surveys were first evaluated at the overall treatment
level. Table 4.19 presents the significant survey items based upon ANOVA at the
treatment group level. The constructs Information Correctness, Information Precision,

Information Reliability, and Amount of Information are found in the significant list.

95

Table 4.19. Survey Significant Items at Treatment Level

 

 

 

 

Treatment Level Effects
W Mean
Constructs Squares df Square F 8L
Confidence - 1.477 3 g 0.492 0745 g -, g 0.527
Information Conectness 13.312 3 4.437 3.437 0.018“
Information Precision 17.127 3 5.709 3.442 ., 0.0138"
lnfonnation Reliability 12.159 3 4.053 2.453 0.065
Infomiation Comprehensiveness 3.370 3 1.123 0.977 0.405
lnfonnation Format 2.542 3 0.847 0.480 0.696
Information Usefulness 4 4.021 g 3, 1.340 ,. 1.604 0.191 g
Rational Approach 4.153 3 1 .384 1 .848 0.141
Intuitive Approach 2.515 3 0.838 0.828 -. 0.481
Amount of Information 1.891 3 0.630 1.003 0.393

 

 

 

 

 

 

 

 

Further evaluation of the significant items through Tukey HSD test reveals the
specific pair or pairs of treatments with significant differences. Figure 4.11 presents
these findings in a 4x4 matrix to facilitate comparison of the results.

The corrected / condensed and uncorrected / condensed treatment groups
produced statistically different responses to ten survey items. All of these questions
focused on information correctness, information reliability, and quantity of information.
The number of significant results from this pairing was expected due to the treatment
providing the subjects with less data on the Output Report and Financial Report.
Therefore, the participants were more reliant on the dashboard to make weekly
operational decisions. For the groups with the corrected dashboard, the absence of
detailed data was less critical than for the groups with an uncorrected dashboard. Further,
for each of the ten questions returning significant differences, the corrected /condensed
treatment groups had a more favorable impression of the accuracy, sufficiency,

consistency, and precision of the dashboard. At the same time, the corrected / condensed

96

treatment groups expressed less concern with the quantity of data available on the Output
Report and Financial report.

The corrected / condensed treatment group also reported statistically different
survey responses on three questions from the corrected / uncondensed treatment group.
These three questions were subsets of the information precision and amount of
information constructs. As with the previous pairing, the corrected / condensed treatment
group expressed higher satisfaction with the precision and less concern about the quantity
of data presented in the session.

The corrected / condensed treatment group had two additional significantly
different responses from the uncorrected / uncondensed treatment group. These two
questions reflected statistically different perceptions about the reliability of the
information presented in the dashboard as well as the quantity of data presented in the
Output Report and Financial Report. The uncorrected / uncondensed treatment group
found the reliability of the information presented on the dashboard to be less sufficient
and the quantity of data presented to be more than he or she could use to make weekly
operational decisions.

The final statistically significant difference within the questions results from the
corrected / uncondensed and the uncorrected / uncondensed treatment groups’ divergent
perceptions of the confusion caused by the quantity of data presented in the Output
Report and Financial Report. This difference was likely caused by the corrected /

uncondensed treatment group relying on the dashboard for weekly decisions.

97

Figure 4.11. Tukey HSD Significant Results by Treatment Level

 

 

 

 

 

 

Post-Hoe Test Sigiificance Summary (Tukey HSD)
Corrected I Condensed “mm”
Precision
Uncorrected I Correctness
Condensed Precision
Corrected I
Uncondensed
Uncorrected I
Uncondensed
Corrected I Uncorrected I Corrected l Uncorrected I
Condensed Condensed Uncondensed Uncondensed

 

 

 

 

 

 

 

The final analysis of the survey data involved reviewing the two treatments
separately to reveal distinctions within the main effects. Table 4.20 presents this group of
significant results. As can be seen in the table, the corrected / uncorrected variable
provided significantly different responses on 22 of the questions in three of the constructs
while the condensed / uncondensed variable only provided a single difference in
response. One possible reason for the disparity is the focus of the questions on the
behavior of the dashboard.

For the corrected / uncorrected variable, the range of constructs returning
statistically significantly differences in the question responses included information
correctness and information precision. As presented in the overall treatment analysis, the
direction of the difference was the same in each case. The mean values of the participants
in the corrected treatment groups indicated higher satisfaction with the information

presented in the dashboard and less concern about the quantity of data presented in the

98

Output Report and the Finance Report. The corrected group means reflected a more

rational approach and less intuitive approach to their decision making process. Of

additional significance to the findings, the users were able to detect inaccuracies in the

information provided to them with the uncorrected information management systems.

Table 4.20. Tests of Between-Subjects Effects for Independent Variables

 

 

 

 

 

 

 

 

 

 

Tests of Between-Subjects Effects
Type In SumW mean
Constructs Squares df Square F 8L

Corrected / Confidence 0.565 1 7 0.5675A7 0.858 0.356
Uncorrected Information Correctness 7.559 1 7.559 7 5.765 0017“
Information Precision 6.800 1 6.800 3.993 0.047”
Information Reliability 7.942 1 7.942 4.790 0.030”

lnforrnation Comprehensiveness77 7 3.137 1 3.137 7 7 772.760 77 0.099

Information Format 2.387 1 2.387 1.370 707.244

Information Usefulness 2.447 1 2.447 2.7930 77#9.7089 7

Rational Approach 3.459 1 3.459 4.649 7 0.033

7 Intuitive Approach 0.058 77777 7 1 7 ' 7 0.058 0.057 7 770.812

Amount of Information 0.039 1 0.039 0.062 0.804

Condensed / Confidence 0.729 1 0.729 1.109 0.294
Uncondensed lnforrnation Correctness 0.000 1 0.000 0.000 0.991
Information Precision 1.050 1 1.050 0.603 0.439

Information Reliability 2.263 _ 1 2.263 1.336 0.250

Information Comprehensiveness 0.150 1 0.150 0.130 0.719

Infonnation Format 7 777i 0.181777 7 1 0.181777 0.103 0.748

Information Usefulness 0.590 1 0.590 0.696 0.405

Rational Approach 0.031 1 0.031 0.040 0.841

Intuitive Approach 0.070 1 0.070 0.069 0.794

Amount of lnfonnation 0.534 1 0.534 0.848 0.358

 

 

99

 

4.5.2 Nominal Group Technique Results

The experimental sessions concluded with collection of the surveys and
discussion among the participants concerning the three open-ended questions at the
beginning of the survey. These three questions solicit the subjects to provide the three
most important pieces of output data used to make decisions, the three least important
pieces of output data, and three items the participants would like to have available to
make the operational decisions. Once the surveys have been collected, the groups were
encouraged to talk openly about the differences in condensed or uncondensed data in the
Output Report and Finance Report as well as the corrected or uncorrected displays of
information on the management dashboard. After refinement from the surveys and the
discussions, the lists of responses by frequency are presented in Tables 4.21 — 4.23.

By overwhelming majority, the participants identified raw material inventory as
the most critical item to monitor when making the weekly operational decisions. Based
on the global sourcing focus of the scenario, the reliance of the production system on raw
material to execute the schedule, and variability in delivery timeliness, the fact that 120
of the 160 participants identified raw material inventory as the most critical item is
expected. The next three highest items include monitoring the demand forecast, tracking
the finished goods inventory, and the weekly production status. These items provide
valuable insight to understanding which variables supply chain managers identify as

critical.

100

Table 4.21. Most Important Decision Making Items as Reported by Participants

 

Most Useful Items

 

 

 

 

 

 

Uncondensed Uncondensed Condensed I Condensed I
I Corrected I Uncorrected Corrected Uncorrected
Count Count Count Count Total

Raw Material Inventory 29 291 25 37 120
Demand Forecast 191 1 7 1 2 17 65
Finished Goods Inventory 1 5 14 1 1 19 59
Production Status 9 17 7 10 43
Costs 10 14 9' 3 36‘
Profit 7 6 8 6 27
Sales Report 4 3 9l 8 24
Lost Sales Report 4 5 3 8 20
Transportation 4 6 4 5 1 9
Dashboard 4 2 5 2 1 3
Calculators 4 1 OI 0 5
Total Responses 109 1 14 93 1 15I 431

 

 

 

 

 

Interestingly, the responses are statistically undifferentiated across the four

treatment groups. Out of a total sample of 431 responses from the 160 participants, with

each participant listing up to three answers for each of the open ended questions, none of

the treatment groups yielded a significant difference in responses from the others.

The same process identified the items that the participants found to be useless or

distracting while making weekly operational decisions. These results are found in Table

4.22. There was far greater diversity in the answers provided to this question both on the

written surveys and in the group discussion at the end of the session. The highest

response rate involved the finished goods distribution center status. One of the metrics

provided to the users identified the percentage of the warehouse currently in use. The

responses from participants indicated that they were less concerned about the inventory

101

 

carrying costs or demurrage costs than they were about meeting the demand within the

regions. The next highest response, market information, was a surprise to be identified

on the least important list. The group discussion involving production status indicated

that the participants focused on the raw material inventory and therefore were less

concerned with the response from the market.

Table 4.22. Least Important Decision Making Items as Reported by Participants

 

Least Useful Items

 

 

 

 

 

 

 

 

 

 

 

 

Uncondensed Uncondensed Condensed I Condensed I
I Corrected I Uncorrected Corrected Uncorrected
Count Count Count Count Total

Distribution Center Status 6 5 2 8 21
Market Information 6 4 2 6 1 8
Production Status 2 6 1 8 17
Investments 7 2 2 4 1 5
Sales Reports 2 2 3 8 15
Warehouse Utilization 2 3 1 8 14
‘ System Notifications 0. 3 6 5 14
Financial 3 5 1 4 1 3
Lost Sales Report 2 1 2 5 10
Too Much Data 2 2 3 2 9
Raw Material Inventory 3 2 1 1 7
Not Enough Time 3 2 0 1 6
Dashboard 3 2 0| 0 5
Transportation 1 1 0 3 5
Other 5 1 0- 0 6
Total Responses 47 41 24 63 1 75

 

The next highest three items include the production status, investments to bolster

demand, and the sales report. In each case, the participants felt there was enough

102

 

complexity to manage across the supply chain without adding these variables to their
weekly operational decisions. Again, analysis by treatment revealed no statistical
differences among the respondents.

The third open-ended question on the survey inquired about additional
information the participant would have found helpful to making the weekly operational
decisions. Table 4.23 presents the complete listing of responses. The most requested
item was a more precise forecast of the demand. In the discussion, participants admitted
that they have the same challenges when dealing with real-world supply chains and
identified demand uncertainty as a major obstacle to efficiency. The next highest request
was for more precise costs of transportation, production, and warehousing so they could
make more informed trade-off decisions about sourcing. Of the remaining requested
items, more time for decision making, more time for training the users on the details of
managing this supply chain, and more precise information in general comprised the
majority of the requests. Participants indicated the precise costs could have been helpful,
but also included demand forecast, reliability of transportation, reliability of suppliers,
and quality defect rate for the production facility as useful items. As with the previous
two open-ended questions, there were no statistically significant differences among the

treatment groups for the responses to this question.

103

Table 4.23. Items Needed to Improve Decision Making as Reported by Participants

 

 

 

 

 

 

 

Needed Items
Uncondensed Uncondensed Condensed I Condensed I
I Corrected I Uncorrected Corrected Uncorrected
Count Count Count Count Total

POW“ 1o 11 8 11 40
0°33 4 9i 3 7 23
More training 1 8 3 5 17
More time 2 2 2 3 9
Transportation 0 2 01 2 4
01"" o o 2 2 4
7°“ ““9“” 17 32 18 3o 97

 

 

 

 

 

 

4.6 Hypothesis Testing
This section presents evaluation of the hypotheses based upon the analysis of the
previous sections in this chapter. The hypotheses are presented in the same order as they

were introduced with a summary table at the end of this section.

4. 6.1 Hypothesis 1

The first hypothesis stated that in an environment of uncondensed data, supply
chain managers with corrected information would perform at a higher level than those
with uncorrected information. By isolating only the treatment groups with uncondensed
data, performing ANOVA for the corrected / uncorrected treatment with the seven
dependent variables reveals significant findings for five of the dependent variables as
presented in Table 4.24. This hypothesis is strongly supported at a 95% confidence level

for the overall supply chain metric and partially supported for the remaining six metrics.

104

 

Table 4.24. ANOVA for Corrected and Uncorrected Effects in an Uncondensed Data

 

 

 

 

 

 

 

 

 

Environment
ANOVA
Sum of Squares df Mean Square F Sig.
Inbound Between Groups _ 9.1951 17 9.195. 6.6581 0.011“
Within Groups 110.4897 80» ' 1.381’_
Total ' 119.685 81 '
Internal Between Groups 7 1223.40? 1‘ 12234047 3802 0.072
Within Groups 29643045 80 370.538; 9
ITotal ’ 30888449 ‘81! if i i i
Outboundpetwesneroupsp _ _ 1&1, 11 _____1..46_1;_9_-§§5i ..__9.417
Within Groups _ 175.7737 807 7 2.197
Total 177.234, 811 7 7
Market Between Groups 7 645.0237 17 645.023 1.490T 0.226
Within Groups _ 34530-217; 801 432.8787 7
Total _ 35275240. 81 i 1
Profit Between Groups 1 2544.417: 1 2544.417, 2.9951 0.087
Within Groups f 67971399. 80; 849.842l *
Total ‘ 70515815 81 ‘ . ‘
sc ContrilBetween Groups_ 3.1753137 1; 3.175913? 5.986 0.016”
Within Groups 4.2441314 80 537049127 f 7 77
Total I " 4.561E+14: 81lm” i i
Overall Between Groups 7 23.774T7 1 23.7747 4.9287 0.029“
Within Groups 7 385.9497 807 4.824
Total 1 409.723 81 . 1

 

 

 

4.6.2 Hypothesis 2

The second hypothesis stated that managers operating in an environment of
uncorrected information would perform at a higher level with condensed data than with
uncondensed data. Isolating the uncorrected information, ANOVA was performed to
analyze the condensed versus uncondensed performance. As seen in Table 4.25, the
investigation yields no statistically significant results. Additionally, within the perceptual
results, the respondents-provided only moderate feedback of less certainty in their

responses with condensed data and a moderately more difficult time performing the

105

weekly operations due tO excessive data in the Output Report and Finance Report. In the

group discussions following the simulation portion of the experimental session, the

respondents expressed satisfaction with the amount Of data provided even at the

condensed level. One respondent from an uncondensed group explained that she

occasionally found herself spending too much time sorting through the vast amounts of

data. However, she further explained that the short amount Of time available between

executing weeks of the simulation forced her to rely on the summary data. The

hypothesis is not supported for any of the seven metrics.

Table 4.25. ANOVA for Condensed and Uncondensed Effects in an Uncorrected Data

 

 

 

 

 

 

 

 

 

 

 

 

Environment
ANOVA
—§finot
Squares df Mean Square F Sig.
inbound [Between Groups 7 1.88611; 1.8861 0.788I 777707.377
Within Groups 203.5861 851 2.395 1 77
Total 7 205.473 86
. Internal Between Groups ‘ 2473.719+ 1i 2473719 3.3451 0.071
Within Groups ‘ 62863618 85 739.5721 ,
Total 65337.337 86' ’ 1
Outbound BetweenGroups 1 2.8091 717 2.809 77 0.958: 270.3301
Within Groups ' 249.1617 85‘ 279311 1
Jotal 251.970 867 1 1
Market Between Groups I 3.5631 1: 3.5637 0.00612 70.938
. -_w1thin-9;<>ups.. 419789115. 851- -. 985-7541 ,___
Total . 49792678, 86. 1 ;
Profit Between Groups 7 13933801 17 13933801 1.8721 0.175
Within Groups 63270177 85 744.355 1
Total f 646635577 861
SC Contrib Between Groups 1 5.545E+121 17 5.545E+121 0.8397 0.362
1WIthin Groups 5.618E+141 857 6.609E+12‘
Total 5.673E+14 86 1
Overall Between Groups 1 4.9317 11 4.931[ 0.8181 0.368
Within Groups ‘ 512.1497 85} f 6.025
Total 517.080 86T

 

106

4.6.3 Hypothesis 3

The third hypothesis stated there in an interaction between the level Of detail Of
the data and the accuracy Of the information provided to supply chain managers.
Specifically, the hypothesis suggested that the overall performance level would be ranked
as follows:

1) Highest performance level: Corrected information / Uncondensed data

2) 2d highest performance level: Corrected information / Condensed data

3) 3d highest performance level: Uncorrected information / Condensed data

4) Lowest performance level: Uncorrected information / Uncondensed data

Evaluating the analysis Of the seven dependent variables, there was a significant
difference in the overall performance of the corrected / uncondensed treatment group and
the uncorrected / condensed treatment group. However, the corrected / condensed
treatment group and the uncorrected / uncondensed treatment group did not show
significant differences between themselves or among the other groups.

In the perceptual portion of the study, the respondents clearly identified
differences between the corrected and the uncorrected information presented on the
dashboard. However, as explored the Hypothesis 2, the respondents were less concerned
about the quantity Of data presented because the summary data was available in all of the

treatment groups. Therefore, this hypothesis is partially supported.

107

4. 6. 4 Hypothesis 4

The fourth hypothesis consisted of two parts pertaining to the behavior of supply
chain managers when presented with incorrect information systems. The first hypothesis
stated that managers presented with consecutive reports containing incorrect information
would seek alternate decision variables to make his or her weekly operational decisions.
In the experimental sessions, the uncorrected treatment groups received randomly
incorrect information. From post-hoc analysis of the uncorrected treatment groups, 51 of
the 87 participants in the uncorrected treatment groups received at least one instance of
consecutive weeks of incorrect information. Analysis of this subset revealed a
statistically significant (p < .05) difference in the responses to the perceptive data in the
areas of information correctness, information precision, information reliability, and
information usefulness than expressed by the remaining 36 participants receiving the
uncorrected treatment but not experiencing consecutive weeks of incorrect information.
Further, 24% (12 of 51) of the participants receiving consecutive weeks of incorrect
information identified the dashboard as one of the three least useful items available for
forming their weekly operation decisions. Conversely, 0% of the subjects identified the
dashboard as one of the three most important items for making weekly operational
decisions. The evidence supports the user’s dissatisfaction with the dashboard and
reliance on alternative decision making variables to complete the weekly operations
input. While there is not an empirical test to evaluate this hypothesis, the perceptual
feedback and the post-hoe analysis of the participant’s performance after receiving

consecutive weeks of incorrect information lends support to this hypothesis.

108

The second portion of the fourth hypothesis stated that once managers seek
alternate decision variables, they do not use the uncorrected information system
regardless of the accuracy of the system. As with the first part of this analysis, the
perceptual feedback provides the insight into the subjects’ behaviors. The 87 participants
in the uncorrected treatment groups received incorrect information on the dashboard in a
total of 161 out of the 522 weeks exercised in the simulation for an overall 31%
occurrence of incorrect information. In the remaining 69% of the weeks, the participants
received corrected information on the dashboard. As presented earlier in this section, the
uncorrected treatment groups reported statistically significant differences in the question
responses included information correctness, information precision, information
reliability, information comprehensiveness, information usefulness, rational approach,
intuitive approach, and amount of information. The uncorrected treatment group
participants indicated lower satisfaction with the dashboard and relied on a more intuitive
approach than a rational approach to the decision making requirements. The group
discussion revealed that the groups who identified the dashboard as misleading did not
refer back to the dashboard in future weekly decision making. Again, there is not an
empirical test to evaluate this hypothesis, but the perceptual evidence provides support

and warrants further investigation of this behavior.

4.7 Discussion of Results
The main objective of this research is to gain greater understanding into the
impact that accurate decision information systems and compression of data has on the

performance of a supply chain. In an environment typified by costly and highly complex

109

hardware and software requirements to detect, store, process, and display data and
information about a supply chain’s status, there is great value to exploring the items
managers identify as vital to operational decision making as well as those that they find
least important. .

Through both the detailed analysis of the seven dependent variables and the
perceptive feedback of the users, the treatment groups with corrected information
performed at a higher level. This treatment had a substantial impact on the decision
making ability, perception of the simulation as a whole, and perception of being able to
make rational operational decisions than the treatment with inaccurate information.

The condensation of data by itself did not have a significant impact on the
performance or perceptions of the participants in the study. Across all seven performance
metrics and 48 survey questions, the treatment groups with condensed or uncondensed
data differed in only two areas. One possible explanation for the lack of significant
findings in this area is the inclusion of aggregated data for both treatment groups. If one
group had only complete data without calculated summaries and the other treatment
group received only the summary data, the study may have produced different results.
This is addressed further in the limitations and opportunities for future research section of
Chapter V.

Despite the insignificant findings of the condensed treatment groups, the
combination of corrected information and condensed data into four groups of treatments
significantly affected the performance when viewed as a whole. The treatment groups
with corrected information benefited most from the opportunity to add uncondensed data

to augment their decision making variables. Conversely, the uncorrected information

110

treatment groups did not realize a substantial benefit from receiving condensed data to
aid the decision making process.

Collection of perceptual feedback from the users provides a useful list for
managers. By identifying the raw material status, demand requirements, production
status, and finished goods status as the most essential decision making items, the groups
were able to condense nearly 1,000 variables into a few critical items for successful
operational decisions.

Equally important, identification of items that were unnecessary for operational
decision making can help managers eliminate excessive data collection. As a whole, the
items on the least useful variables list included areas traditionally outside of the supply
chain environment. Inclusion of such items as financial reports, marketing information,
production status, utilization rates, and sales figures on the list indicate the narrow focus
sometimes adopted in managing complex supply chains in constrained timeframes. By
broadening user’s scope of decision variables to include some of these items, greater
returns in performance may be realized.

The list providing the items participants requested to make better operational
decisions is similarly valuable to researchers as well as managers. Identifying better
methods for forecasting demand behavior, providing improved precision of costs and
decision variables, and offering managers the luxury of time to consider all of these
variables before making operational decisions are all areas that researchers and
practitioners can collaborate to improve supply chain management as a science and a

occupation.

111

CHAPTER V: CONTRIBUTIONS AND CONCLUSION

5.0 Introduction

This chapter presents the implications of the research findings. The first section
presents the research contributions of the study. The second discusses managerial
contributions. This section is followed by a discussion of the limitations of this study and

recommended directions for future research.

5.1 Research Contributions

This study makes academic contributions to both theory refinement and
experimental design for future research opportunities. The Knowledge Based View of
the Firm provided a foundation for this study through identification of critical implicit
knowledge required to perform complex functions and then detect and adapt new
information into an existing framework. The study largely upheld the idea that greater
levels of data converted into corrected information will help managers make operational
decisions resulting in higher operational effectiveness. However, this study also raised
questions about Davenport and Prusak’s (1998) processes to transform data to
information and then to knowledge (Figure 2.1). The authors asserted condensing vast
amounts of data was essential to knowledge creation. In this study, there was no
statistical difference between the users in the uncorrected treatment group with
uncondensed data and those with condensed data. This creates potential for future studies
to resolve the apparent paradox.

The approach undertaken in this study is unusual within the supply chain

management arena. The triangulation of techniques to capture complex simulation data,

112

a formal experimental design, and essential perceptual feedback is not a common practice
in the field. However, the insight gleaned from putting the empirical results together with
human impressions and behaviors is critical to better understanding the research
questions. This process also ensured the research remained pertinent to managers due to
the close interaction with the population representing diverse levels of supply chain
knowledge and experience. While some areas for refinement are addressed in the
limitations section, future studies should incorporate similar design to provide the human

behavioral reporting in conjunction with empirical results.

5.2 Managerial Contributions

The opportunity to work closely with managers in the development of SCODE,
the testing of the experimental design, and the experimental sessions ensured the
pertinence of the study to managers. By design, SCODE is intended to bring together
practitioners in the supply chain arena of multiple levels and create a cooperative learning
environment. In addition to preparing for the experimental sessions, the validation
opportunities with SCODE yielded numerous improvements for application of the
simulation to the classroom and Executive Education environments. A few examples of
the evolution of the program included development of a simpler user interface to control
SCODE, deployment of the program from USB drives rather than full installation on a
computer, and the addition of material requirement calculators to reduce the
mathematical computation time between operational decisions. Through the refinement

in conjunction with this study, SCODE has become a valuable tool to integrate with

113

traditional classroom learning in developing the skills and knowledge necessary for
managers to successfully compete in the complex and challenging supply chain field.

As a result of this research, a great deal of insight was gained into the behavior of
managers under dynamic conditions within a supply chain. Though the experience
involved a simulation in this case, these same phenomena have been realized at
subsequent Executive Development sessions. As managers have the Opportunity to
design or revise an existing supply chain, strategic goals of the company must be a
priority consideration. Through identification and publication of goals focused on certain
customer satisfaction levels, specific profit margins, return on assets, or other measures,
the personnel employed within the supply chain environment are immediately
conditioned to certain behaviors. Senior management must ensure that the metrics, goals,
and reward system within the company promote broad-visioned employees rather than
people who optimize local metrics at the expense of the overall supply chain. For
example, in this research, participants could easily ensure 100% production efficiency for
their metric by increasing safety stock and setting a desired fill rate of 80% of customer
demand. However, the efficiency and effectiveness of the entire supply chain suffers.
Similarly, senior managers must carefully balance the employees’ span of control reward
system with the overall needs of the company.

Further developing the previous recommendation, broader training requirements
within the supply chain environment are clearly needed. Discussions with managers and
researchers support the assertion that many employees in supply chain related positions
have limited comprehension of tactical operational requirements across the supply chain.

Many employees have a broad understanding of how procurement, transportation,

114

warehousing, production, distribution, and marketing fit into the overall health of the
organization, but few grasp the effect that a decision in one area can have throughout the
supply chain. Numerous participants in this study discussed the eye-opening experience
of failing to enter a raw material order correctly and then watching production,
transportation costs, and demand satisfaction suffer over the next few weeks of the
simulation. Development of supply chain professionals at all levels must be a recurring
process within firms. A third recommendation focuses on the approach to this training.
By integrating the simulation into structured lessons with specific objectives, the
participants were able to reinforce the concepts from the lessons in a first-hand
environment without jeopardizing the health of the firm in the real world. Completing
the learning loop by sharing lessons and insight among groups, the managers had the
opportunity to express their perceptions at different challenges of the simulation and learn
from other participant’s successes and failures. Whether the simulation is SCODE,
another computer-based simulation, or a completely different hands-on learning tool, this
educational process has strong merit and should be adopted by managers seeking return
on their training investments.

The fourth area of research contributions focuses on the linkage of accurate
information and uncondensed data with higher performance levels. While the most
complete data sets combined with accurate decision information systems yielded
significantly higher performance in some areas, there was not a statistically significant
difference between the corrected / uncondensed, corrected / condensed, and the
uncorrected / uncondensed groups at an overall supply chain level. Given an unlimited

budget and the ability to integrate extensive quantities of data into a management

115

dashboard, a firm would be wise to get the most accurate and complete system available.
However, firms that must make trade-offs due to costs, computing skills, storage
capacity, time required to record data, or any variety of other constraints affecting most
businesses should make investments into the most accurate decision information systems
available, even at the expense of completeness of data.

The fifth managerial contribution is discussed in greater detail at the end of
Chapter IV (pp. 109 — 110) and will be only briefly mentioned in this section. The
creation of lists identifying the most important, least important, and desired items for
managers to make supply chain decisions should be invaluable. By focusing on these
items and gaining a better understanding of significance to a particular firm, managers
should be able to make immediate impact in operational performance through addressing
the shortfalls, reinforcing the most important items, and making a decision to either

eliminate recording or improve education related to the least important items.

5.4 Limitations and Opportunities for Future Research

As with any research involving a simulation study, certain assumptions are
adopted in the creation of the simulation to limit the breadth of topics that can be covered
in a short amount of time. The assumptions also serve to focus the variability of the
interaction with respondents in an effort to detect significant differences in a few effects.
In the SCODE simulation, there are numerous assumptions created to provide an
environment where the participants can quickly adapt to the interface and learn valuable
supply chain management lessons within a short amount of time. The operating

environment was created to capture the most critical aspects of managing a supply chain

116

without becoming overburdened with the thousands of details occurring every day in
real-world supply chains.

SCODE provides a dynamic learning environment for users to be exposed to a
variety of supply chain challenges. While it is Excel-based for ease of interface with the
users, it shares a common limitation with many software packages. Specifically, SCODE
has a steep learning curve for new participants. All of the subjects in the study were new
users of SCODE. A 3-hour timeframe was extremely compact to provide a global
sourcing lecture, distribute USB drives of the software and ensure compatibility with
diverse personal computers, train users on the input-output cycle of SCODE, execute the
scenario, complete the survey, and participate in the nominal group session. With
additional time or multiple sessions, the results of this study may have changed,
particularly with the condensed / uncondensed treatment groups. If the users were more
proficient at reviewing the available data and integrating more of the variables into their
weekly decision making routine, there would likely have been greater significance in the
findings between the groups with only condensed data and those with extensively
detailed data.

Starting with the basic design of SCODE, there are realistic, but simplified,
parameters affecting the interactive environment. The structuring of the simulation into
weeks of play accelerates the input-output effect, but forces assumptions about behavior
and performance. Since each set Of operational weekly decisions requires the participants
to invest considerable time and effort in the process, some of the details that could
normally occur in a firm are overlooked. The raw material orders are placed at the

beginning of the week. However, the users will not receive feedback about exact

117

quantity to arrive or the timeliness of the shipment until the following week. Further, if
the user schedules production which exceeds the labor force capabilities, the feedback
occurs in the following week when shortages are apparent. When there are shortages
affecting the user’s ability to meet demand in a market, there are no backorders. All of
these assumptions create an environment suitable for understanding and developing the
concepts critical to operating an effective supply chain network, but will not mimic all
aspects of real-world supply chains.

Another limiting aspect when compared to real world supply chains is the
architecture of the scenario. In this case, the users operated a single production facility
and could source raw materials from four locations to distribute finished goods to five
markets. This process is an enormous simplification of a real-world supply chain, but
again creates a manageable framework for the participants to operate within.

The characteristics of the products created in this scenario could have a limiting
factor on the results. The general design incorporates a chemical company
manufacturing artificial sweetener for bulk consumers. The design includes liquid and
dry powder variants of a standard grade and a premium grade of the product. Each of the
products requires four or fewer raw materials to create. The weight and packaging
requirements favor shipment by land or sea by making air transportation significantly
more expensive. Given the design, these are reasonable assumptions. However, users
who operate in a drastically different supply chain in the real world will bring certain
preconceptions to SCODE. If the participant’s background involves high-cost electronics
with a short shelf life and intense competition in the global market, he or she will be

accustomed to utilizing air transportation for inbound and outbound shipping, will meet

118

all demand at any cost to preserve market share, and is likely to schedule the production
mix to favor the premium lines of products. This may be in stark contrast to a participant
with a background in low value, bulk goods with an extended shelf life and a steady
market share. While the design cannot simultaneously mimic both extremes of a supply
chain in a single setting, it could be a limiting factor to the broader application of the
results.

At the same time, the design of the scenario provides opportunities for future
studies. The first recommendation for future research includes developing a range of
scenarios with different product focus to investigate the changes in behavior over diverse
supply chains. One iteration could incorporate the lightweight, high value item discussed
in the previous paragraph with a design favoring air shipment and complete satisfaction
of demand. Another scenario could include more complex supply chains consisting of
multiple production facilities, complex demand curves, and supply chain interruptions.
The behavior of participants under the different circumstances may be significantly
different than under the scenario provided for this study.

A second research opportunity involves refinement of the current study. By
creating additional layers of correctness of information and condensation of data, a more
refined model of the trade-offs could be developed. In future iterations of this study,
levels could include corrected information with no data, condensed data with no
information dashboard, a between-group level of data condensation, and completely
uncondensed data without including the condensed summary data. Repeating the study
with these additional levels of treatment would require substantially more participants,

but could shed additional light on the balance required between costly and complex data

119

collection and information management systems and the absence of quality information
or enough data to make informed operational decisions.

Another opportunity for future research involves testing additional aspects of
Davenport and Prusak’s (1998) model. By controlling the quantity of data and the
quality of information throughout the simulation, variables addressing the
contextualization, categorization, and calculations of the data could be manipulated and
the study repeated. While less focused on the information technologies required to
process data, these variables would lend insight into the cognitive processes of
integrating supply chain data into the information required to make knowledgeable

Operational decisions.

5.5 Conclusion

This study set out to investigate the process of converting supply chain data into
useable information yielding the knowledge necessary to achieve a high level of
performance. This chapter summarized the study as it pertains to the goal. This study
also provided empirical results and perceptions about the variables in question. This
chapter summarized the previous four chapters and recommended opportunities to further
develop the framework into additional research. The research and managerial
implications of this study provide the opportunity to benefit from this study as well as
create internal improvement opportunities to build a stronger sense of supply chain theory

and improve operational decision making.

120

APPENDIX A: SCODE USER’S MANUAL EXECUTIVE SUMMARY

IllI LINIFRICDRIIAIMR>1HP
C31 1111 DICE ITAI INIIIF’RISI

SCODE SIMULATION MODEL

EXECUTIVE SUMMARY

BROAD

[I INISK

ftllSL.-.I.IKLAN STATE
I N‘s \ E It s I 1 \

 

April 8, 2008

121

INTRODUCTION

SCODE is a computerized logistics simulation designed to provide realistic training on
integrated decision making in a competitive supply network environment. SCODE is
executed as a supply chain in which users compete against an environment controlled by
game administrators.

SCODE demonstrates the importance of integrated decision making. Participants make
several decisions including sourcing, supply base collaboration, production scheduling,
marketing, and distribution.

SCODE also provides a learning environment to apply the concepts presented in
educational seminars to a realistic setting. Learners are grouped into teams. Each team is
expected to discuss and make strategic and operational decisions in order to compete
effectively in SCODE. Due to time constraints for making and entering decisions, each
team is required to work efficiently. Moreover, SCODE goes beyond a simple training
exercise -— it provides users an opportunity to enjoy the challenge of achieving top
performance in a highly competitive environment.

DECISION OVERVIEW

SCODE teams are expected to develop and execute an integrated supply chain strategy in
a competitive environment. Each team makes two sets of decisions: strategic and
operational. Strategic decisions relate to supply chain network design. Simulation
initialization decisions (intended to get the simulation process started) are made during
strategic decision phase. Operational decisions involve weekly supply chain activities.
SCODE administrators control the simulation environment by manipulating production
’ economies, material availability, transportation lead times, and demand variation.

Thus, as Figure 1 shows, there are three categories of decisions fed into SCODE: (i)
Strategic decisions, (ii) Operational decisions, and (iii) Administrator decisions. (Refer to
Sections 2.0, 3.0, and 5.0 of the User’s Manual for a detailed discussion of each decision
phase). The administrator runs SCODE simulation and then gives results to teams to help
them make their next weekly decisions. The results are in form of reports (Refer to
Section 4.0 for detailed discussion of SCODE output reports and interpretation).

122

Figure 1. SCODE model framework

 

 

Strategic Decisions
(One time team decisions at start)

0 Plant (own or contract)

Administrator Decisions
(Controlled by administrator)
0 Demand variation

e Production economies

'0 Transportation lead times
oSupplier availability
oReport detail

0 Plant capacity
0 DC (own or contract)
0 DC capacity
0 Market & DC allocation
0 Initialization Decision
0 Initial inventory
6 Scheduling

 

 

 

 

 

 

 

 

SCODE

 

 

V

 

 

 

 

 

Omrations Decisions
(Weekly team decisions)

oSOurcing (plant, supplier, raw material,
mode, quantity)

0 Production (what to produce where A
and when, quantity)

oMarketing (R&D, promotion, price)

oDistribution (source, DC, mode,
quantity)

 

 

RESULTS

 

 

 

 

 

 

 

 

 

 

GEOGRAPHIC OVERVIEW

SCODE environment consists of five geographic regions, namely, North America,
Europe, South America, China, and India. The team’s manufacturing plant will be
located in one of these regions. Raw materials can be sourced from any of the available
suppliers. A distribution center is required to be open in the region where the
manufacturing plant is located and a team may open additional DCs in other regions.
There are five market segments, representing the five geographic regions. Teams can
locate DCs in each market segment or in select segments. A DC can serve more than one
market segment. Figure 2 provides a pictorial depiction of the entire SCODE supply
chain network. In any given scenario, portions of the network may be unavailable in
order to emphasize certain learning objectives.

123

Figure 2. SCODE Supply Chain Network

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Suppliers Suppliers Suppliers Suppliers
(North America) (Europe) (South America) (China)
1 1
L T v W
North America Europe South America China

 

RM warehouse

 

RM warehouse

RM warehouse

RM warehouse

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Plant Plant Plant Plant
DC DC DC DC
6 v t J
1
DC (North DC DC DC DC
America) (Europe) (South (China) (India)
America)
1 v
11
v
Market
(N. A.)

 

 

124

 

 

PRODUCT OVERVIEW

SCODE consists of two product categories: bulk industrial product (liquid) and packaged
consumer product (dry). Each of these products has a regular and premium brand. Each
product brand has different raw material requirements. Teams may decide to deal with
one product line (industrial versus consumer product), one brand of each product
category (regular versus premium), or all of them (both product categories and brands of
each category). The brands are:

Table 1: Products and Raw material components

 

 

Product Category Brand Raw Materials
Industrial (Liquid) Regular Each product will
Premium consist of a unique
Consumer (Dry) Regular mixture of raw materials
Premium from the four types

 

TRANSPORTATION OVERVIEW

Three Options are available for to transport raw materials to the manufacturing plant,
transport finished goods to additional DCs, or in transport finished goods to the market.
The options, cost, reliability rates for on-time deliveries, and expected delivery times are
shown in Table 2.

Table 2: Transportation

 

 

 

Mode Cost Reliability Expected Transit
Time (weeks)
Air High High 0 (arrives in same
week shipped)
Land Low High 1
Sea Low Low 2
FEEDBACK

SCODE will provide the teams feedback in a number of Excel worksheets. This data
includes raw material levels, production output, DC finished goods levels, quantities sold
to market, lost sales, financial performance, transportation costs, plant Operating costs,
and raw material costs to name but a few. Successful interpretation of these indicators
will affect the next week’s decisions and help produce a sustainable supply chain.

ADMINISTRATOR ROLE

SCODE administrators will identify key training objectives for an educational session.
Based upon these learning objectives, scenarios will be designed to mimic real-world

125

situations. The teams will have the opportunity to apply their personal knowledge as well
as integrating the information provided during the classroom education portion of their
training in the SCODE environment. Through the use of the “Administrator” workbook,
administrators will have the ability to alter the very nature of the SCODE experience to
focus attention on one or two topics. The learning cycle is closed by providing both in-
depth debrief on the decisions and outcomes of a team’s performance as well as a group-
wide debrief on the learning objectives from the lesson. Additional information about the
variables, scenario design, and expected outcomes are available in the Administrator’s
Manual.

126

APPENDIX B: SCODE GLOBAL SOURCING BRIEFING

 

Global Sourcing with the
Supply Chain Operations
Decision Environment

6 — 7 June 2008
Shawn Jones

 

 

 

 

 

E'L‘s'fi'té‘é Introduction

- SCODE Overview

- Executive Education Scenario:
Training with Global Sourcing Focus

. Execution

- Metrics / Feedback

 

 

 

127

 

Installation Of the Software

- Insert USB drive
- Copy entire SCODE folder to your C:\
- Double click on C:\SCODE\SCODE.exe
- Should open SCODE window
- If so, close the window

- If not, let me know

- Pass the drive on

 

 

 

 

_ Introduction: SCODE Simulation

- Interactive learning tool designed for
Executive Education and MBA courses

- Setting includes:

- Individuals or teams Operating a global
supply network

- Up to four raw materials

- Up to four finished goods

- Over 600B possible approaches

- Over 1,000 administrator-controlled
variables

 

 

 

128

 

1115mm SCODE Product Overview

- Industrial (Bulk) - Consumer (Package)
— Liquid — Dry
- Premium - Premium
- Regular - Regular

 

 

 

Why Source Globally?

 

Access to markets

Cost/Price Benefits

Access to Technology

Quality

Access to Only Source Available

Introduce Competition to Domestic Suppliers

React to Buying Patterns of Competitors

Establish a Presence in a Foreign Market

 

 

 

129

Barriers to Global Sourcing

 

° Lack of understanding Of International
procedures

- Lack of knowledge of documentation
- Resistance to change

- Domestic market nationalism

Longer leadtimes and material pipelines

Increased uncertainty and risk

 

 

 

Barriers tO Global Sourcing

 

- Logistical, political, financial risks

- Planning, coordination, information sharing

- Lack of knowledge of foreign business practices
- Language and cultural differences

- Expertise and labor

. Negotiations can be difficult

- Engineering changes, changes in general difficult

 

 

 

130

 

 
   

1:; Global Sourcing Futur
-

Continued refinement/development of the
global sourcing process

Develop or obtain human resources with
higher-level skills and a willingness to view
the supply network from a worldwide
perspective

Develop global performance measures

Establish integrated systems between
worldwide units and with suppliers

 

 

aumm Training (Global Sourcrng) Scenario

 

Learning Objectives

— Introduce SCODE

- Orientation to input/output

— Increase understanding of cause-effect in SCODE

— Secondary objective: learn about trade-offs of global
sourcing

Simulation Model Proxies

— Variable transportation reliability

— Significant price difference for regions

— Bulky, heavy goods expensive to transport
- Variable product availability

 

131

 

 

 

- Supply chain goal:

- Selected suppliers because:

- Marketing approach:

- Other notes:

 

 

 

lil'tlk'té‘é Operations Commentary

 

- Challenges to meet your goal:

- How were the challenges handled?

- What worked well?

- Other notes:

 

 

 

132

APPENDIX C: SCODE SIMULATION SCENARIO

fl “\d

IIIF C IPxITIRI<DRIIAIDlRSIIIF
")1“ I ll I)IC3 IIAI [Nillxll’lsl

SCODE SIMULATION MODEL

GLOBAL SOURCING
SCENARIO

MIC! IKJW STAN

\‘r'tlllsl

 

June 6 - 7, 2008

133

GLOBAL SOURCING TRAINING SCENARIO
Scenario: International Sourcing and Manufacturing in Uncertain Conditions

Congratulations! Today is your first day on the job as the Vice President for
Operations and Supply Chain Management at Spartan Technologies, an artificial
sweetener manufacturing firm. You have access to a global network of suppliers to
provide raw material to your manufacturing plant in North America. Your product is
available in four bulk varieties consisting of a “regular” grade and a “premium” grade of
each dry products and liquid products. Four raw materials are involved in production in
the following ratios:

AMERICA

 

Your raw materials are available globally in a spot-buy market. Due to increased
competition in the marketplace, your raw material orders may not always be completely
fulfilled. Also be aware that transportation reliability may vary among regions. Your
suppliers offer significant discounts to purchase in quantities that are easier for them to
ship, handle, and store. Raw material costs from various regions are as follows:

Supplier Unit Prices

An Quantit Price Quanti Break

NORTH
AMERICA

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

RM A $ 3.20 20,000 3 295
RM B 3 3.25 2,000 3 3.10
RM c 3 0.55 20,000 5 0.50
RM D s 6.25 1,000 3 6.00
. , .
RM A $ 3.40 10,000 3 3.25
RM B 3 3.20 2,000 3 3.15
RM 0 $ 0.55 2,000 $ 0.50
RM D $ 6.40 1.000 s 6.20
0
RM A 3 3.30 40,000 8 2.90
RM B 3 3.30 75,000 3 2.75
RM c 3 0.75 50,000 $ 045
RM D $ 6.10 100,000 3 5.00
RM A 3 4.00 5.000 3 3.80
RM B s 3.50 2,000 $ 3.45
RM c s 0.55 20,000 $ 0.50
RM 0 $ 6.00 11.000 3 5.90

 

 

 

 

 

134

 

 

 

 

Your objective is simple: keep Spartan Industries profitable. You may choose to
specialize in one product and dominate the world market. You might concentrate your
efforts on one region across all products. Some might attempt to meet all worldwide
demand. Still others may combine these tactics in pursuit of the most efficient supply
network. The choice is yours.

Some key facts and figures about your operation are:

l.

The North American manufacturing plant has a capacity of 43,000 labor-hours
per week.

. You have a raw material storage warehouse with a capacity of 300,000 units.

. Your predecessor leaves you with enough raw materials to meet worldwide

demand for all products for one week (You begin with 1 week of raw material
inventory).

. You have a one week supply of finished goods in your distribution center.

. Initial sales forecast predicts the following weekly demand

 

. Your marketing research department provided you with the following

anticipated demand patterns for different regions of the world. These patterns
are established for your regular dry powder and the three new products are
expected to follow a similar pattern:

135

 

Demand Pattern
North America, South America, and Eu rope

 

 

 

 

V ' ' 1' U V I v r I r I V I Y

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

 

Demand Pattern
China and India

1 2 3 4 5 6 7 8 9 1O 11 12 13 14 15 16 17 18 19 20

136

 

 

7. The average sales price for your product in each region is:

, LIQUID REGULAR LIQUID PREMIUM

DRY REGULAR

1 n. . nit

DRY PREMIUM ‘

 

 

 

$100.00 $150.00

 

$300.00

 

 

$350.00

 

The decisions you must make each week are:
1. MATERIAL PROCUREMENT AND SPOT BUY: decide product mix and

quantities you’d like to produce

a. Input raw material order quantity
b. Select the raw material location
c. Select the mode of transport for the raw materials
2. PRODUCTION SCHEDULE: decide whether accelerated production is
desired and the quantities for each priority

DJ

DISTRIBUTION NETWORK: decide how to transport your goods to market
4. INVESTMENT: decide if you would like to invest between $0 and $20,000

in Sales Force, Promotion, Convenience, or R&D factors to improve your

demand in specific markets

137

APPENDIX D: SCODE FEEDBACK SURVEY

Group Number: Date:

 

 

Please take a few minutes to provide us with feedback about your experience with
SCODE. Please indicate in the space provided the degree to which each statement
applies to you. There is no right or wrong answer. Many of the statements are similar to
other statements. Do not be concerned about this.

Demographics:

Male Female Years of Supply Chain Experience

Current Position

 

Highest degree completed: High School Associates Bachelors Masters Ph.D.

Key Definitions (Davenport and Prusak, 1997; Drucker, 1988):
Data: Discrete, objective facts; structured records of transactions
Information: Data endowed with relevance and purpose

1. If your team divided responsibilities among members, what was your role?

 

2. What was the team’s supply chain goal (circle all that apply)?
Efficiency Meet all demand Focused Product Mix Other:

3. Please tell us the three most important pieces of output data you used to make
decisions for the next week’s operations in order of importance to you:

1.

 

2.

 

3.

 

4. Please tell us three pieces of output data you found useless or a distraction to
making your decisions for the next week’s operations

1.

 

2.

 

 

138

5. Is there any information you needed, but did not have available, to make your

decisions for the next week’s operations? Please explain.

6. Rate the difficulty you experienced when performing this task.

1
Convenient
1

Good

1

Easy

1

Efficient

7. Rate the correctness of the information on the dashboard

1

Accurate

1

High

1
Consistent
1
Sufficient

8. Rate the precision of the information on the dashboard

1
Sufficient
1

High

1
Consistent
1

Superior

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

139

5

5

5

5

5

5

7
Inconvenient
7
Bad
7
Difficult
7
Inefficient

7
Inaccurate

7
Low

7
Inconsistent

7
Insufficient

7
Insufficient
7
Low
7
Inconsistent
7
Inferior

9. Rate the reliability of the information on the dashboard

1 2 3 4 5 6 7
Consistent Inconsistent
1 2 3 4 5 6 7
High Low
1 2 3 4 ‘ 5 6 7
Superior Inferior
l 2 3 4 5 6 7
Sufficient Insufficient

10. Rate the comprehensiveness of the information on the dashboard

1 2 3 4 5 6 7
Consistent Inconsistent
l 2 3 4 5 6 7
Complete Incomplete
l 2 3 4 5 6 7
Adequate Inadequate
1 2 3 4 5 6 7
Sufficient Insufficient

11. Rate the format of the information on the dashboard

l 2 3 4 5 6 7
Good Bad
1 2 3 4 5 6 7
Simple Complex
1 2 3 4 5 6 7
Readable Unreadable
1 2 3 4 5 6 7
Useful Useless

12. Overall, rate the information on the dashboard

1 2 3 4 5 6 7
Useful Useless
1 2 3 4 5 6 7
Relevant Irrelevant
1 2 3 4 5 6 7
Clear Hazy
1 2 3 4 5 6 7
Good Bad

140

13. I am confident that my solution is the best possible one

1 2 3 4 5 6 7
Strongly disagree Strongly agree

14. The amount of data on the Output and Financial worksheets made this task
difficult ‘

l 2 3 4 5 6 7
Strongly disagree Strongly agree

15. I used an organized plan to solve the problem

1 2 3 4 5 6 7
Strongly disagree Strongly agree

16. Completing the weekly operation correctly was important to me

1 2 3 4 5 6 7
Strongly disagree Strongly agree

17. There are better answers to this task than the ones I reached

1 2 3 4 5 6 7
Strongly disagree Strongly agree

18. I completed weekly operations input based upon overall impressions

1 2 3 4 5 6 7
Strongly disagree Strongly agree

19. I would prefer to use information displayed in dashboards in order to be
successful

1 2 3 4 5 6 7
Strongly disagree Strongly agree

20. My answers were based on a logical process of making choices

1 2 3 4 5 6 7
Strongly disagree Strongly agree

21. My answers were based on intuition

1 2 3 4 5 6 7
Strongly disagree Strongly agree

141

22. Information displayed in dashboards increased the time required to obtain an
answer

1 2 3 4 5 6 7
Strongly disagree Strongly agree

23. Information displayed in dashboards increased the accuracy of my answers

1 2 3 4 5 6 7
Strongly disagree Strongly agree

24. The amount of data on the Output and Financial worksheets made answering
the question confusing

I 2 3 4 5 6 7
Strongly disagree Strongly agree

25. I was unable to use a rational strategy in arriving at my answers

I 2 3 4 5 6 7
Strongly disagree Strongly agree

26. Did the dashboard help you interpret the weekly operations output?

1 2 3 4 5 ~ 6 7
Always Never

27. Did the dashboard provide you with accurate information to help decision
making?

1 2 3 4 5 6 7
Always Never

28. Did the dashboard provide you with enough information to help decision
making?

1 2 3 4 5 6 7
Always Never

29. Did the Output and Financial worksheets provide enough data to help decision
making?

1 2 3 4 5 6 7
Always Never

142

30. Did the Output and Financial worksheets provide more information that you
could use to make weekly operational decisions?

1 2 3 4 5 6 7

Always Never

31. Did you feel SCODE provided a realistic simulation of supply chain operations?

1 2 3 4 5 6 7

Always Never

143

APPENDIX E: RESEARCH PARTICIPANT INFORMATION

AND CONSENT FORM

You are being asked to participate in a research project. Researchers are required to provide a
consent form to inform you about the study, to conveythat participation is voluntary, to explain
risks and benefits of participation, and to empower you to make an informed decision. You
should feel free to ask the researchers any questions you may have.

Study Title: An Evaluation of the Impact of Supply Chain lnforrnation Management Systems on
Operational Performance

Researcher and Title: Shawn R. Jones, Ph.D. Candidate, Logistics

Department and Institution: Department of Supply Chain Management, The Eli Broad Graduate
School of Management, Michigan State University

Address and Contact Information: 471 North Business Complex, East Lansing, MI 48824-1 122;
517-432-6457

1. PURPOSE OF RESEARCH:

You are being asked to participate in a research study of Supply Chain Management
leaders attending educational seminars at Michigan State University

You have been selected as a possible participant in this study because your experience in
Supply Chain Management provides a unique perspective on the subject.

From this study, the researchers hope to learn how the accuracy of decision information
systems and the volume of supporting data available to a manager affects the knowledge
creation process and ultimately impacts operational performance of a supply chain.
Your participation in this study will take about 1 hour beyond the required classroom
portion of the exercise. If you choose not to participate in the study, no data is retained
from the classroom portion of the exercise.

If you are under 18, you cannot be in this study without parental permission.

2. WHAT YOU WILL DO:

After completing the required classroom portion of the exercise, participants will
complete a survey about their experience and participate in a nominal group discussion to
identify the most important and least important decision variables you experienced during
the supply chain simulation.

Your answers will be collected and analyzed in aggregate. The summary results will be
available to you in approximately one week after the session through your ANGEL site.

3. POTENTIAL BENEFITS:

The potential benefits to you for taking part in this study are to gain a broader
understanding of the impact of accurate decision information systems and appropriate
amounts of supporting data in a supply chain setting.

Further, your participation in this study may contribute to the broader understanding of
how managers create knowledge from decision information systems and the ultimate
impact of uncertainty on operational performance.

4. POTENTIAL RISKS:

There are no foreseeable risks associated with participation in this study.

144

5. PRIVACY AND CONFIDENTIALITY:
I The data for this project are being collected anonymously. Neither the researchers nor
anyone else will be able to link data to you.
I The results of this study may be published or presented at professional meetings, but the
identities of all research participants will remain anonymous.

6. YOUR RIGHTS TO PARTICIPATE, SAY NO, OR WITHDRAW
I Participation in this research project is completely voluntary. You have the right to say
no.
I You may change your mind at any time and withdraw.
I You may choose not to answer specific questions or to stop participating at any time.
I Choosing not to participate or withdrawing from this study will not make any difference
m
o the quality of any services you may receive.
0 benefits to which you are otherwise entitled.
I Whether you choose to participate or not will have no affect on your grade or evaluation.
I You will be told of any significant findings that develop during the course of the study
that may influence your willingness to continue to participate in the research.

7. COSTS AND COMPENSATION FOR BEING IN THE STUDY:
I You will not receive money or any other form of compensation for participating in this
study.

8. CONTACT INFORMATION FOR QUESTIONS AND CONCERNS

If you have concems or questions about this study, such as scientific issues, how to do any part of
it, or to report an injury, please contact the researcher, Shawn Jones: 471 North Business
Complex, East Lansing, MI 48824, joncs@bus.msu.edu, 517-432-6457.

 

If you have any questions or concerns about your role and rights as a research participant, or
would like to register a complaint about this study, you may contact, anonymously if you wish,
the Director of MSU’s Human Research Protection Program at 517-355-2180, Fax 517-432-4503,
or e-mail irb@msu.edu or regular mail at 202 Olds Hall, MSU, East Lansing, MI 48824.

12. DOCUMENTATION OF INFORMED CONSENT.

Your signature below means that you voluntarily agree to participate in this research study.

 

Signature

You will be given a copy of this form to keep.

145

BIBLIOGRAPHY

American Heritage Dictionary Online, www.bartleby.com, accessed October 13, 2007.

Andrews, K. (1971). The Concept of Corporate Strategy. Homewood, 111: Dow Jones-
Irwin.

 

Badii, A. and A. Sharif (2003) Information Management and Knowledge Integration for
Enterprise Innovation, Logistics Information Management, 16, 145 — 155.

Bailey J ., and S. Pearson (1983) Development of a Tool for Measuring and Analyzing
Computer User Satisfaction, Management Science, 29, 530 - 545.

Barney, J. (1991) Firm Resources and Sustained Competitive Advantage, Journal of
Management, 17, 99 — 120.

Bendoly, E., E. Rosenzweig, and J. Stratman (2007) Performance Metric Portfolios: A

Framework and Empirical Analysis, Production and Operations Management, 16, 257 —
276.

Blasberg, J ., V. Vishwanath, and J. Allen (2008) Tools for Converting Consumers into
Advocates, Strategy and Leadership, 36, 16 — 23.

Center for Leadership of the Digital Enterprise, www.bus.msu.edu/clode/, accessed
November 13, 2007.

Challener, C. (2000) Taking the VMI Step to Collaborative Commerce, Chemical Market
Reporter, 258, 11 — 12.

Choi, B., S. Poon, and J. Davis (2006) Effects of Knowledge Management Strategy on
Organizational Performance: A Complementarity Theory-Based Approach, Omega, 36,
235 — 251.

Chow, H., K. Choy, and W. Lee (2006) Design of a Knowledge-Based Logistics
Management System: A Case-Based RFID Approach, International Journal of
Enterprise Network Management, 1, 5 — 28.

Cohen, J. (1988) Statistical Power Analysis for the Behavioral Sciences, 2d Edition.
Hillsdale, NJ: Erlbaum Associates.

Cohen, W. and D. Levinthal (1990) A New Perspective on Learning and Innovation,
Administrative Science Quarterly, 35, 128 — 152.

Conner, K. (1991) A Historical Comparison of Resource-based Theory and Five Schools
of Thought within Industrial Organization Economics: Do We Have a New Theory of
the F irm?, Journal of Management, 17, 121 — 154.

146

 

Conner, K. and C. Prahalad (1996) A Resource-Based Theory of the Firm: Knowledge
versus Opportunism, Organization Science, 7, 477 — 501.

Croson, R. and K. Donohue (2006) Behavioral Causes of the Bullwhip Effect and the
Observed Value of Inventory Information, Management Science, 52, 323 — 336.

 

Davenport, T. and L. Prusak (1998) Working Knowledge: How OLganizations Manage
What They Know. Boston, MA: Harvard Business School Press.

Davis, J ., K. Eisenhardt, C. Bingham (2008) Developing Theory Through Simulation
Methods, The Academy of Management Review, 32, 480 — 99.

Dennis, H. (2006) Create a System to Monitor Performance, Small Business Times, April
14, 2006.

Department of Defense (DoD) Dictionary (2003)
www.dtic.mil/doctrine/jel/doddict/data/d/O1486.html. Accessed October 10, 2007.

Dierickx, I, K. Cool (1989) Asset Stock Accumulation and Sustainability of Competitive
Advantage, Management Science, 35, 1504 — 1511.

Disney, S. and D. Towill (2003) The Effect of Vendor Managed Inventory (VMI)
Dynamics on the Bullwhip Effect in Supply Chains, International Journal of Production
Economics, 85, 199 — 215.

Druker, P. (1988) The Coming of the New Organization, Harvard Business Review, 66,
45 - 53.

F iala, P. (2004) Information Sharing in Supply Chains, Omega, 33, 419 — 423.

Grant, R. (1996) Toward a Knowledge-Based Theory of the Firm, Strategic Management
Journal, 17, 109 — 122.

Gunasekaran, A., C. Patel, and E. Tirtiroglu (2001) Performance Measures and Metrics in
a Supply Chain Environment, International Journal of Operations and Production
Management, 21, 71 — 87.

Hair, J ., R. Anderson, R. Tatham, and W. Black (1998). Multivariate Data Analysis.
Upper Saddle River, NJ: Prentice Hall Publishing.

Helo, P. (2000) Dunamic Modelling of Surge Effect and Capacity Limitation in Supply
Chains, International Journal of Production Research, 38, 4521 — 33.

Holmes, E. (2008) Air Force Launches Plan for Centralized Purchasing, Federal Times,
March 24, 2008.

I47

Huber, G (1990) A Theory of the Effects of Advanced Information Technologies on
Organizational Design, Intelligence, and Decision Making, The Academy of Management
Review, 15, 47 - 71.

Imber, J. and B Toffler (2000) Dictionary of Marketing Terms. San Francisco, CA:
Barron’s Educational Publishing.

 

Jones, R, C. Clarke-Hill, D. Comfort, D. Hillier, and P. Shears (2005) Radio Frequency
Identification and Food Retailing in the UK, British Food Journal, 107, 356 - 360.

Kang, Y. and S. Gershwin (2005) Inventory Inaccuracy in Inventory Systems: Stock
Loss and Stockout, IIE Transactions, 37, 843 — 859.

Kahn, K., E. Maltz, and J. Mentzer (2006) Demand Collaboration: Effects on
Knowledge Creation, Relationships, and Supply Chain Performance, Journal of Business
Logistics, 27, 191 — 223.

Kanawattanachai, P. and Y. Yoo (2007) The Impact of Knowledge Coordination on
Virtual Team Performance Over Time, MIS Quarterly, 31, 783 — 808.

Kaplan, R., and D. Norton (2000) Using the Balanced Scorecard as a Strategic
Management System, Harvard Business Review, January — February 2000.

Kelton, W., R. Sadowski, and D. Sturrock (2003) Simulation with Arena. Hightstown,
NJ: McGraw Hill Publishing.

Law A. and D. Kelton (2000) Simulation Modeling and Analysis. New York: McGraw-
Hill Publishing. ,

Lee, H. and B. Choi (2003) Knowledge Management Enablers, Processes, and
Organizational Performance: An Integrative View and Empirical Examination, Journal
of Management Information Systems, 20, 179 - 228.

Lee, H., V. Padmanabhan, and S. Whang (1997) Information Distortion in a Supply
Chain: The Bullwhip Effect, Management Science, 43, 546 — 558.

Mahoney, J. and J. Pandian (1992) The Resource-Based View within the Conversation of
Strategic Management, Strategic Management Journal, 15, 363 — 380.

Martin, W. and R. Stanford (2007) A Methodology for Estimating the Maximum
Profitable Turns for an ABC Inventory Classification System, IMA Journal of
Management Mathematics, 18, 223 — 233.

Michigan State University Operations Management Summit Proceedings (2007).

Min, 8., A. Roath, P. Daugherty, S. Genchev, H. Chen, A. Amdt, R. Richey (2005)
Supply Chain Collaboration: What’s Happening?, The International Journal of Logistics
Management, 16, 237 — 256.

148

Nonaka, I. (1994) A Dynamic Theory of Organizational Knowledge Creation,
Organization Science, 5, 14 — 37.

Pathirage, C ., D. Amaratunga, and R. Haigh (2007) Tacit Knowledge and Organisational

Performance: Construction Industry Perspective, Journal of Knowledge Management,
11, 115 — 126.

Peteraf, M. (1993) The Comerstones of Competitive Advantage: A Resource-Based
View, Strategic Management Journal, 14, 179 -— 191.

Raman, A., N. Deoratius, and Z. Ton (2001) Execution: The Missing Link in Retail
Operations, California Management Review, 43, 136 — 152.

Ridalls, C., S. Bennett, and N. Tipi (2000) Modelling the Dynamics of Supply Chains,
International Journal of System Science, 31, 979 — 76.

Robinson, E., F. Sahin, and L. Gao (2008) Master Production Schedule Time Interval
Strategies in Make-to-Order Supply Chains, International Journal of Production
Research, 46, 1933 — 1954.

Rodrigues, A. (2004) Simulation of the Impact of Forecasting Accuracy on Supply Chain
Performance: The Bias Effect, Ph.D. Dissertation, Michigan State University.

Rohleder, T. and G. Scudder (1995) Information Inaccuracy: Implications for Economic
Production Scheduling, International Journal of Production Economics, 42, 51 — 66.

Sargent R. (2000) Verification, Validation, and Accreditation of Simulation Models,
Proceedings of the 2000 Winter Simulation Conference.

Sharma, A., M. Levy, and H. Evanschitzky (2007) The Variance in Sales Performance
Explained by the Knowledge Structures of Salespeople, Journal of Personal Selling and
Sales Management, 27, 169 — 181.

Shin, N. (1999) Does Information Technology Improve Coordination? An Empirical
Analysis, Logistics Information Management, 12, 138 — 144.

Simon, H.(1957) Models of Man. New York: Wiley and Sons.

Slater, S., E. Olson, and V. Reddy (1997) Strategy-Based Performance Measurement,
Business Horizons, 40, 37 — 44.

Speier, C. (1996) The Effects of Task Interruption and Information Presentation on
Individual Decision Making, Ph.D. Dissertation, Indiana University.

Spurrier, G., H. Topi, and J. Valacich (1994) The Effects of Information Compression
and Time Pressure on Individual Decision-making Performance in A Task Using Textual
Information, Proceedings from the 25th Conference of the Decision Sciences Institute.

149

Sugimori, Y. K. Kusunoki, F. Cho, and S. Uchikawa (1977) Toyota Production System
and Kanban System: Materialisation of Just-in-Time and Respect-for-Human System,
International Journal of Production Research, 15, 553 — 564.

Szulanski, G. and R. Jensen (2004) Overcoming Stickiness: An Empirical Investigation
of the Role of the Template in the Replication of Organizational Routines, Managerial
and Decision Economics, 25, 347 — 363.

Van Buren, M. (1999) A Yardstick for Knowledge Management, Training and
Development, 53, 71 — 78.

Wacker, J. (1996) A Theoretical Model of Manufacturing Lead Times and Their
Relationship to a Manufacturing Goal Hierarchy, Decision Sciences, 27, 483 — 513.

Wemerfelt, B. (1984) A Resource-Based View of the Firm, Stratgegic Management
Journal, 5, 272 — 280.

White, G. (1996) A Meta-analysis of Manufacturing Capabilities, Journal of Operations
Management, 14, 315 — 331.

White House Budget www.whitehouse.gov/omb/budget/fy2008, accessed December 4,
2007.

Womack, J ., D. Jones, and D. Roos (1990). The Machine that Changed the World. New
York: Rowson Associates.

Xu, K., Y Dong, and P. Evers (2001) Towards Better Coordination of the Supply Chain,
Logistics and Transportation Review, 37, 35 — 54.

Zack, M (1999) Managing Codified Knowledge, Sloan Management Review, 40, 45 — 58.

150