THE IMPACTS OF AGILE DEVELOPMENT METHODOLOGY USE ON PROJECT
SUCCESS: A CONTINGENCY VIEW
By
John F. Tripp

A DISSERTATION
Submitted to the
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Business Information Systems
2012

ABSTRACT
THE IMPACTS OF AGILE DEVELOPMENT METHODOLOGY USE ON PROJECT
SUCCESS: A CONTINGENCY VIEW
By
John F. Tripp
Agile Information Systems Development Methodologies have emerged in the
past decade as an alternative manner of managing the work and delivery of information
systems development teams, with a large number of organizations reporting the
adoption & use of agile methodologies. The practitioners of these methodologies make
broad claims as to the benefits of their use. However, to date, only a few of these claims
have been tested in the research literature.
Agile methodologies, including Extreme Programming, Scrum, and others,
prescribe very different practices, some of which are contradictory. Additionally, the use
of the practices of agile methodologies is not restricted to agile development projects,
and has been observed in non-agile methodologies environments. Even so, the
previous research literature has usually focused on practices prescribed by a particular
agile method. So what is different about agile methodologies, and what is the
appropriate lens through which to study them?
This dissertation finds that the most distinctive element of agile methodologies is
their strong emphasis on obtaining and processing feedback from the environment. This
dissertation evaluates the impacts of agile methodologies as indicated by the use of
these feedback processes.
In this study, the theoretical lenses of team adaptation, organizational learning,
and the prior literature on new product development are used to explain the importance

of a team’s ability to process repeated and continuous feedback from the environment.
We motivate hypotheses regarding the positive impact of agile methodology use on a
multi-dimensional construct representing project success. This construct encompasses
the quality of the delivered product, the benefits of the project to the organization, and
impacts on project management outcomes. In addition, the nature of moderating
influences of uncertainty on project success is discussed.
The research design for the study utilized a survey that collected responses from
83 agile development teams. Generalized linear modeling was used to test four
hypotheses regarding the impact of the extent of agile methodology use on project
success, and the moderating influences of uncertainty. It was found that agile
methodology use positively impacts project success, while structural complexity
negatively moderates the impact of agile use. It was also found that environmental
dynamism positively moderated the impacts of agile methodologies on project success.
Discussion of the results is provided.

Copyright by
John F. Tripp
2012

To my mother, Pamela Richman Tripp, and to my father and step-mother John F.
Tripp, Sr. and Asya Lapidus; you have always supported my endeavors, without ever
criticizing. To my in-laws, John & Susan McLennan, for the support and encouragement
that you show to your daughter’s husband and his crazy endeavors.
To my children, Cayley Moriah, Joshua Lawrence Chamberlain, Zoe Elizabeth,
Jonathan Edwards, Winston Churchill, Thomas Stonewall Jackson, and Normandy
Reagan. Forgive me for being too busy to hear you, for missing soccer games, for
missing too many dinners with you. Your love for your very imperfect father kept me
going on many a long night’s work.
Finally, to my wife, Molly Ann. No one is as incredible a help to her husband. No
one loves a husband with more faults. No one is more courageous and self-sacrificing
for her family. Without you, more than anyone else, my PhD could not have been
completed. Thank you for marrying me, and thank you for the endless hours of putting
the kids to bed alone, keeping the house alone, and teaching and instructing the
children alone, all for the sake of my work.
I dedicate this dissertation to you, my family, who I love very much.

Soli Deo Gloria.

v

ACKNOWLEDGEMENTS
I would like to acknowledge the patience and guidance of my advisor, Dr. Vallabh
Sambamurthy, who first implanted the idea of pursuing the PhD in my head, and who
was forced to listen to me discuss the idea of this dissertation for far too long before it
became a “good” idea. Without his guidance and inspiration, I would not have
completed this dissertation.
Next, I would like to acknowledge the various faculty members who have,
formally or informally, contributed to my completing the PhD process. First, I wish to
thank the members of my committee, Dr. Roger Calantone, Dr. John Wagner, and Dr.
Brian

Pentland, I

thank

you for your guidance, feedback,

correction,

and

encouragement. I could not have asked for a better set of faculty to assist me in my
thinking and direction for this dissertation. I would also like to acknowledge the
assistance of Dr. D. Harrison McKnight, who spent significant time helping me craft the
survey instrument, and gave me help with this project whenever I asked, even though
he was not a member of my committee.
I would also like to acknowledge the support and assistance given to me by
Team Detroit, Inc., especially CFO Andy Weil and CIO James Weekley. Without the
incredible flexibility afforded to me, and the support given to me during my first three
years in the program, I could not have supported my family while also pursuing the PhD
degree. I also thank many of my co-workers at Team Detroit who provided me with
much of the inspiration for this project, and who contributed advice and feedback. I
thank you all for your flexibility, enthusiasm, tolerance and support.

vi

Finally, I would like to acknowledge the contribution and support of the agile
development community, especially Ken Faw and John Birgbauer of Xede, Rich
Sheridan of Menlo Innovations, Gary Gentry and Jay Aho of Pillar Technology, Eugene
Keil and Chris Beale of Cengage Learning, Pat Reed of The Gap, and their many coworkers who helped me refine the ideas of this dissertation, and to identify avenues for
data gathering. Also, I would like to thank the membership of the Mid-Michigan Agile
Meetup, the Michigan Agile Enthusiasts, and the Chicago Agile Methodology Group for
their encouragement and support.
Finally, I would like to acknowledge the many other people who lent support
during this process, including the members of Providence PCA Church and Oakland
Hills Community Church. Your encouragement meant a great deal.

vii

TABLE OF CONTENTS
LIST OF TABLES ........................................................................................................... XI
LIST OF FIGURES ...................................................................................................... XIV
CHAPTER ONE: INTRODUCTION AND OVERVIEW .................................................... 1
INTRODUCTION............................................................................................................... 1
BACKGROUND ................................................................................................................ 7
PURPOSE OF THIS STUDY ............................................................................................... 9
ORGANIZATION OF PRESENTATION ................................................................................ 10
CHAPTER TWO: THEORETICAL MODEL AND SUPPORTING LITERATURE........... 11
INTRODUCTION............................................................................................................. 11
THE EVOLUTION OF METHODOLOGY - A QUEST FOR FIT .................................................. 13
WHAT IS A SOFTWARE DEVELOPMENT METHODOLOGY? .................................................. 19
Waterfall – Big Delivery .......................................................................................... 21
Agile Methodologies – Continuous Delivery ........................................................... 25
Definable vs. Empirical Processes ......................................................................... 27
Agile Development – a Description ........................................................................ 30
THE PRACTITIONER LITERATURE ON AGILE METHODOLOGIES .......................................... 34
UNCERTAINTY: COMPLEXITY AND CHANGE ..................................................................... 42
TEAM ADAPTABILITY & ORGANIZATIONAL LEARNING ........................................................ 45
Cue Recognition..................................................................................................... 47
Team Situational Awareness ................................................................................. 49
Mutual Monitoring................................................................................................... 50
SOFTWARE DEVELOPMENT AS NEW PRODUCT DEVELOPMENT ........................................ 52
ADAPTABILITY, AGILE AND FEEDBACK CAPABILITIES ........................................................ 54
PROJECT SUCCESS ...................................................................................................... 56
Project Management Metrics ................................................................................. 57
Definitions of IS Success ....................................................................................... 58
CONTINGENCIES IN THE PROJECT ENVIRONMENT ........................................................... 59
THE COMPONENTS OF UNCERTAINTY AS CONTINGENT FACTORS ..................................... 60
THE RESEARCH LITERATURE ON THE IMPACTS OF AGILE DEVELOPMENT ........................... 63
AGILE METHODOLOGIES: STRUCTURE AND IMPACTS ....................................................... 76
Agile Philosophy Adoption ..................................................................................... 77
Agile Management Control..................................................................................... 78
Supporting Infrastructure Adoption......................................................................... 78
CHAPTER SUMMARY ..................................................................................................... 81
CHAPTER THREE: RESEARCH MODEL .................................................................... 82
INTRODUCTION............................................................................................................. 82
RESEARCH MODEL AND VARIABLES ............................................................................... 82
PROJECT SUCCESS ...................................................................................................... 86
EXTENT OF AGILE METHOD USE .................................................................................... 88
Reduced Up Front Planning ................................................................................... 88
viii

Environmental Feedback ....................................................................................... 89
Iterative Delivery .................................................................................................... 90
Technical Feedback ............................................................................................... 91
UNCERTAINTY AND PROJECT SUCCESS .......................................................................... 93
Structural Complexity ............................................................................................. 93
CONTROL VARIABLES ................................................................................................... 99
Project Size ............................................................................................................ 99
Organization Size ................................................................................................... 99
Time since Agile Adoption ...................................................................................... 99
Team Member Skill .............................................................................................. 100
Team Member Agile Experience .......................................................................... 100
CHAPTER SUMMARY ................................................................................................... 100
CHAPTER FOUR: RESEARCH METHODOLOGY ..................................................... 102
INTRODUCTION........................................................................................................... 102
CONSTRUCT OPERATIONALIZATION.............................................................................. 102
Extent of Agile Method Use .................................................................................. 103
Project Success ................................................................................................... 105
Control Variables.................................................................................................. 108
PRE-TESTING THE SURVEY ......................................................................................... 109
SAMPLE DESIGN AND DATA COLLECTION ..................................................................... 110
CHAPTER SUMMARY ................................................................................................... 115
CHAPTER 5: ANALYSIS AND RESULTS................................................................... 116
INTRODUCTION........................................................................................................... 116
Analysis Approach ............................................................................................... 116
EXTENT OF AGILE METHOD USE .................................................................................. 116
OUTCOME VARIABLES ................................................................................................ 118
UNCERTAINTY ............................................................................................................ 121
Dynamism ............................................................................................................ 121
Complexity ........................................................................................................... 122
ANALYSIS OF DATA CHARACTERISTICS......................................................................... 127
Tests for Normality ............................................................................................... 127
Tests for Unusual and Influential Data ................................................................. 140
Tests for Linearity................................................................................................. 142
Transformation of Data ........................................................................................ 143
ANALYSIS RESULTS .................................................................................................... 145
TESTING THE RESEARCH MODEL ................................................................................. 147
STRUCTURAL COMPLEXITY ......................................................................................... 147
Summary of results: Hypothesis 2a...................................................................... 154
TECHNICAL COMPLEXITY ............................................................................................ 154
Summary of results: Hypothesis 2b...................................................................... 160
DYNAMISM ................................................................................................................. 160
Summary of results: Hypothesis 2c ...................................................................... 166
Summary of results: Hypothesis 1........................................................................ 166
CHAPTER SUMMARY ................................................................................................... 169
ix

CHAPTER 6: DISCUSSION AND IMPLICATIONS OF THE RESULTS...................... 170
INTRODUCTION........................................................................................................... 170
THE EFFECTIVENESS OF AGILE METHODOLOGIES IN SOFTWARE DEVELOPMENT ............. 170
THE MODERATING EFFECTS OF UNCERTAINTY ............................................................. 172
LIMITATIONS OF THIS STUDY ....................................................................................... 174
IMPLICATIONS FOR PRACTICE ...................................................................................... 175
IMPLICATIONS FOR FUTURE RESEARCH ........................................................................ 177
CONCLUSION ............................................................................................................. 179
APPENDIX A: QUESTIONNARIE ITEMS ................................................................... 181
ITEMS FOR STRUCTURAL COMPLEXITY ......................................................................... 181
ITEMS FOR PROJECT CRITICALITY ................................................................................ 182
ITEMS FOR TECHNICAL COMPLEXITY ............................................................................ 182
ITEMS FOR DYNAMISM ................................................................................................ 183
ITEMS FOR AGILE METHOD USE .................................................................................. 184
ITEMS FOR PRODUCT QUALITY .................................................................................... 185
ITEMS FOR ORGANIZATIONAL BENEFITS ....................................................................... 186
ITEMS FOR PROJECT MANAGEMENT OUTCOMES ........................................................... 186
APPENDIX B: SUPPLEMENTAL DATA TABLES ....................................................... 187
REFERENCES ............................................................................................................ 191

x

LIST OF TABLES

Table 1.1. Anticipated Benefits of Agile Methodology Adoption (Version One 2010). ..... 5
Table 1.2. Significant Agile Methodologies and Proponents. .......................................... 8
Table 2.1. Summary of Generations of Development Methodologies ........................... 14
Table 2.2. Selected Definitions of Software Development Method[ology] ..................... 20
Table 2.3. Selected Definitions of “Agile/Agility” ............................................................ 28
Table 2.4. The 12 Principles of the Agile Manifesto ...................................................... 37
Table 2.5. Defined Practices of Major Agile Methodologies .......................................... 41
Table 2.6. Standish CHAOS results .............................................................................. 58
Table 2.7. Five Factors Determining Fit of Agile vs. Plan-Based Methodologies .......... 62
Table 2.8. Recent Empirical Literature on Agile Methodologies .................................... 67
Table 2.9. Output control vs. emergent outcome control (Harris, 2009) ........................ 72
Table 3.1. Research Variable Definitions ...................................................................... 85
Table 3.2. Summary of Hypotheses ............................................................................ 101
Table 4.1. Sections of Questionnaire Presented to Each Respondent Type ............... 103
Table 4.2. Items Indicating the Extent of Agile Method Use ........................................ 104
Table 4.3. Items Indicating Project Management Success .......................................... 105
Table 4.4. Items Indicating Product Quality ................................................................. 107
Table 4.5. Items Indicating Organization Benefits ....................................................... 108
Table 4.6. Demographics Breakdown of Respondents ............................................... 112
Table 4.7. Missing Data Summary for 373 Cases ....................................................... 114
Table 5.1. Factor Loadings for the Dimensions of Agile Method Use .......................... 117
Table 5.2. Descriptive Statistics for Agile Use Items ................................................... 118
Table 5.3. Factor Loading for Product Quality and Benefits Items .............................. 119
xi

Table 5.4. Descriptive Statistics for Product Quality and Benefits Items ..................... 120
Table 5.5. Factor Loadings for the Dimensions of Dynamism ..................................... 122
Table 5.6. Descriptive Statistics for Uncertainty Items ................................................ 122
Table 5.7. Factor Loadings for the Dimensions of Technical Complexity .................... 123
Table 5.8. Descriptive Statistics for Technical and Structural Complexity Items ......... 124
Table 5.9a. Second Order Factor Loadings: Quality ................................................... 124
Table 5.9b. Second Order Factor Loadings: Technical Complexity ............................ 124
Table 5.9c. Second Order Factor Loadings: Agile Method Use .................................. 125
Table 5.9d. Second Order Factor Loadings: Dynamism.............................................. 125
Table 5.10. Correlations and Variance Inflation Factors.............................................. 126
Table 5.11. Shapiro-Wilk Test for Normality of Distribution ......................................... 128
Table 5.12. Generalized Linear Models Used to Test Moderating Effects of Uncertainty
............................................................................................................................. 148
Table 5.13. Goodness of Fit of Model 3 the Moderation Effects of Structural Complexity.
............................................................................................................................. 149
Table 5.14. Parameter Estimates for Model 3, the Moderation Effects of Structural
Complexity. .......................................................................................................... 150
Table 5.15. Cases by Category – Structural Complexity ............................................. 151
Table 5.16. Summary of Interaction Effects - Structural Complexity ........................... 151
Table 5.17. Nested Model Interaction Coefficients – Structural Complexity ................ 153
Table 5.18. Goodness of Fit of Model 4. Moderation Effects of Technical Complexity.
............................................................................................................................. 156
Table 5.19. Parameter Estimates for Model 4, the Moderation Effects of Technical
Complexity. .......................................................................................................... 157
Table 5.20. Cases by Category – Technical Complexity ............................................. 158
Table 5.21. Summary of Interaction Effects - Technical Complexity ........................... 158
Table 5.22. Nested Model Interaction Coefficients – Technical Complexity ................ 159

xii

Table 5.23. Goodness of Fit of Model 5, the Moderation Effects of Dynamism. .......... 162
Table 5.24. Parameter Estimates for Model 5, the Moderation Effects of Dynamism. 163
Table 5.25. Cases by Category – Dynamism .............................................................. 164
Table 5.26. Summary of Interaction Effects - Dynamism ............................................ 164
Table 5.27. Nested Model Interaction Coefficients – Dynamism ................................. 165
Table 5.28. Summary of Hypothesis Findings ............................................................. 168
Table B.1. Response Breakdown by Team and Response Type ................................ 187

xiii

LIST OF FIGURES

Figure 2.1. The “Waterfall” Method ............................................................................... 22
Figure 2.2. The Cost Curve of Change Under The Waterfall Method ............................ 24
Figure 2.3. Agile Method Feedback Loop...................................................................... 30
Figure 2.4. The Cost of Change Curve Under Agile Method (Beck 1999) ..................... 34
Figure 2.5. The Nomological Network of Agile Method Adoption and Use .................... 77
Figure 2.6. The Theoretical Model of The Impacts of Agile Method Adoption and Use on
Project Success. .................................................................................................... 80
Figure 3.1: The Theoretical Model for Analyzing The Impacts of Agile Method Adoption
on Project Success ................................................................................................ 84
Figure 5.1a: Diagnostic Outputs for Quality - Histogram ............................................. 129
Figure 5.1b: Diagnostic Outputs for Quality – Box Plot ............................................... 130
Figure 5.1c: Diagnostic Outputs for Quality – Q-Normal Plot ...................................... 131
Figure 5.1d: Diagnostic Outputs for Quality – Symmetry Plot ..................................... 132
Figure 5.2a: Diagnostic Outputs for Agile Method Use - Histogram ............................ 133
Figure 5.2b: Diagnostic Outputs for Agile Method Use – Box Plot .............................. 134
Figure 5.2c: Diagnostic Outputs for Agile Method Use – Q-Normal Plot ..................... 135
Figure 5.2d: Diagnostic Outputs for Agile Method Use – Symmetry Plot .................... 136
Figure 5.3a: Diagnostic Outputs for Dynamism - Histogram ....................................... 137
Figure 5.3b: Diagnostic Outputs for Dynamism – Box Plot.......................................... 138
Figure 5.3c: Diagnostic Outputs for Dynamism – Q-Normal Plot ................................ 139
Figure 5.3d: Diagnostic Outputs for Dynamism – Symmetry Plot ................................ 140
Figure 5.4. Matrix Plot of Factor Relationships............................................................ 141
Figure 5.5. Spline Transformation Plots ...................................................................... 144

xiv

Chapter One: Introduction And Overview
“We are uncovering better ways of developing software by
doing it and helping others do it." – The Agile Manifesto
Introduction
Information system development methodologies (henceforth, IS development
methodologies) have been used over the past 50 years with the intent to both increase
control and reduce the uncertainty and risk of IS development projects (Avison and
Fitzgerald 1998; Hirschheim et al. 1995). IS projects have long been recognized as
being difficult to define, manage, and deliver (e.g. - Barki et al. 2001; Keil 1995; Keil et
al. 2000; Marakas and Elam 1998; Zmud et al. 1993). During its early years, software
development was primarily an engineering concern. However, the systems development
process context has increasingly become focused on the automation and support of
core business processes. Therefore, development teams are being asked to adapt
rapidly to changing technical, market, and user requirements (Conboy 2009), deliver
business value more quickly and to respond to change more (Lyytinen and Rose 2006).
However, even with the benefit of 50 years of method evolution, projects often exceed
budget and time estimates, escalate for years, and regularly deliver far less than
originally promised (Keil 1995).
IS development projects historically have been long-term efforts, delivering
systems after months or even years. However, with the increasing pace of business
competition, traditional development project timelines have been criticized as taking too
long. Time pressures have escalated due to heightened competition and the escalating
speed of technical innovation. This requires the software development process to
deliver business value as quickly as possible. At the same time, software development
1

teams are being asked to adapt to rapidly changing technical, market, and user
requirements. This pace of change must be managed without adversely impacting team
performance.
Multiple software development methodologies have emerged in the last four
decades of the twentieth century. At first, these methodologies emerged in response to
the need to control and standardize the development process. Development
methodologies continued to evolve to address issues related to growing systems and
environmental

complexity

(Hirschheim

et

al.

1995).

However,

while

these

methodologies have evolved for more than 50 years, project failure rates continue to be
extremely high (Standish Group, 2009). Project timetables continue to be long, flexibility
is low, and the delivery of business value is usually achieved at the end of the project, if
at all.
The latest generation of development methodologies, known as “agile”
methodologies, have emerged over the past two decades (Beck and Andres 2004;
Conboy 2009). These methodologies claim to be able to better handle the dynamic
nature of the business environment (Boehm and Turner 2004; Cockburn 2001;
Highsmith 2002; Schwaber 1996). Agile development methodologies share key
philosophical underpinnings. These include the early and continuous delivery of
software, the embracing and creating of change, the cultivation of empowered teams,
and the delivery of simple solutions.
Agile development methodologies are different from traditional methodologies in
several key ways. First, traditional methodologies usually planned to use 33% or more
of the project duration in the planning phase (PMBOK 2000). In contrast, agile

2

methodologies recommend a maximum of 10% of anticipated project time to be spent
on up front planning (Anderson 2004; Coad et al. 1999; Highsmith 2002), and most
recommend much less. Second, most traditional methodologies usually planned to
delivered the project system at the end of the project. In contrast, agile methodologies
call for early delivery of business value through an iterative, evolutionary process.
Because of this focus on early delivery, agile methodologies claim to increase the
potential for business value to be delivered, while at the same time reducing the risk
that changes in the environment will reduce the usefulness of the system before it is
delivered.
Additionally, it has long been recognized that it is often difficult to properly define
the requirements of the system up front (Brooks 1987). A great deal of research has
been

performed

on

various

methodologies

for

better

creating

requirements

specifications (Brooks 1987; Marakas and Elam 1998; Zmud et al. 1993). However,
users often cannot clearly enumerate their requirements until they can compare working
software with their expectations and task context (Brooks 1987). Traditional
methodologies such as the waterfall methodology were designed to deliver software at
the end of the project. Because of this, users often did not interact with the software
until it was completed at the end of the project. Agile methodologies, because of their
focus on early deliver of working software allow for incongruity between the developed
software and users needs to be detected earlier in the project. Because of this, agile
methodologies may be able to better adjust to meet the requirements that only emerge
as users interact with the real system.

3

Because of these reasons, agile methodologies have been proposed by the
practitioner community to deliver significant positive impacts to project performance.
Claims regarding the impacts of the use of agile development methodologies fall into a
few common categories. First, agile methodologies claim to boost productivity by
creating a sustainable pace of development (Beck 1999), which creates an
organizational environment in which people wish to work (Highsmith 2002). Second,
agile methodologies claim to build trust between the software development team and
their stakeholders due to the integration of stakeholders into the project team, and the
regular and repeated demonstration of working software (Fowler and Highsmith 2001).
Third, agile methodologies claim to better manage turbulent environments, due to their
focus on interactive and adaptive design and delivery. Because of this iterative delivery
paradigm, agile methodologies practitioners claim that agile methodologies deliver
better and more useful software. This increased delivery success is claimed to reduce
risk, and provide a better return on investment (Moran 2010) (Highsmith 2002).
These claims have spurred widespread agile methodology adoption, (Ambler
2009; Schwaber and Fichera 2005), with more than 50% of organizations reporting the
adoption of agile methodologies (West and Grant 2010). Job postings referencing agile
practitioners have increased more than 400% between 2008 and 2011 (Indeed.com
2011). Amongst practitioners, there is near unanimity as to the anticipated diverse set of
benefits of agile development (see Table 1.1).
However, the normative claims of agile methodology impacts have been
accompanied by little empirical evidence. The majority of reports of agile project
success have been anecdotal (e.g., Abrahamsson et al. 2002; Lindvall et al. 2004;

4

Nerur et al. 2005), or focused on particular agile practice adoption (Cockburn and
Williams 2001; Parrish et al. 2004; Williams et al. 2000). The few large sample empirical
studies that have been reported focused on only some of the claims. For instance, the
use of extreme programming practices has been found to be associated with higher
code quality, as measured by a reduction in bugs (Maruping et al. 2009a).
Table 1.1. Anticipated Benefits of Agile Methodology Adoption (Version One 2010).
Anticipated Benefits
% Respondents
Anticipating
Benefit
Improved Project Visibility
Enhanced ability to manage changing priorities
Increased productivity
Accelerated time to market
Enhanced software quality
Reduced risk
Reduced cost
Better management of distributed teams
Simplified development process
Improved alignment between IT and Business Objectives
Improved Engineering Discipline
Enhanced Software Maintainability
Improved Team Morale

88.7%
97.6%
97.7%
96.3%
94.5%
93.5%
85.3%
54.2%
89.6%
88.6%
84.9%
87.4%
85.1%

Additionally, there are conflicting viewpoints as to how and when agile
methodologies should be applied to the highly diverse contexts of software development
projects. Some have claimed that agile development is only suited for small, non-critical
projects (Agerfalk and Fitzgerald 2006; Boehm 2002). Others have reported successful
use of agile development methodologies on projects that are of large scale, high
technical complexity, and high coordination complexity (Anderson 2004; Beck and
Andres 2004; Highsmith 2002; Schwaber and Beedle 2002). Since there is
disagreement regarding the appropriateness and applicability of agile methodologies, it
5

is necessary that the research community develop a better understanding of how agile
methodologies impact project performance.
Though there is a long tradition of research on the merits of alternative
development methodologies, the research literature has only recently begun to seriously
investigate the agile methodology phenomenon (Conboy 2009; Conboy and Fitzgerald
2004). Even as the research community has begun to devote significant resources
toward the investigation of agile methodologies, this research is still at an early phase,
and trails behind practice (Conboy 2009). Several issues have hampered the research
literature. These include the inability to consistently define what “agility” means (Conboy
2009), and the limited conceptualization of agile development as being proxied by a
particular agile methodology such as Extreme Programming, an individual engineering
practice such as pair programming, or being viewed as an homogenous “black box”
(Lee and Xia 2010).
Agile development methodologies cannot be treated as a “black box”. While agile
methodologies share a common philosophical foundation, each stresses a different set
of engineering and process practices. Research on agile development should not focus
upon particular engineering practices for two reasons. First, these practices were used
before the emergence of agile development methodologies, and can be used outside of
agile methodologies. Second, the practices within agile methodologies are sometimes
inconsistent and can even be contradictory. For example, collective code ownership is a
practice that states that any member of the team can modify any portion of the code as
necessary. However, while collective code ownership is a foundational practice in
Extreme Programming (XP), it is specifically rejected in Feature Driven Development

6

(FDD). Collective code ownership has been used as one indicator of the level of agile
methodology utilization (Maruping et al. 2009a), but as both FDD and XP are
considered agile methodologies use of this indicator as a measure of agile methodology
adoption may be inappropriate. Because these practices are contradictory, but both
appear in agile methodologies, there is a need for a richer conceptualization of agile
development methodology adoption and use: one that extends beyond the use of
particular agile methodology practices.
Background
Agile development methodologies are a subcategory of iterative development
methodologies. These methodologies focus on shorter timetables for delivery and the
ability to rapidly respond to change. They stress the importance of good coding
techniques, and good coders, but also implement a broad set of practices meant to
manage work production, team norms and interaction, team boundary management,
and the response to external influences. An overarching argument behind adopting
agile development is that, because software development is an uncertain process, and
therefore cannot be predicted, change cannot be avoided and in fact should be
expected (Schwaber 1996). Therefore teams must be unencumbered by unnecessary
process and planning, and organized to adapt well to change in order to succeed (Beck
& Andres, 2004; Fowler & Highsmith, 2001).
Practitioners have proposed a set of methodologies that they consider to be agile
methodologies (Table 1.2 gives a partial listing), yet these methodologies prescribe a
wide variety of conflicting and sometimes contradictory practices to deliver agility. As
noted above, Extreme Programming (XP) and Feature Driven Development (FDD) are

7

both considered examples of successful agile methodologies, yet they prescribe
contradictory practices.
Table 1.2. Significant Agile Methodologies and Proponents.
Methodology

Proponent Agile Manifesto Author(s)

Extreme Programming

Kent Beck, Martin Fowler, Ward
Cunningham, Robert Martin

Feature Driven Development

Jon Kern

Scrum

Mike Beedle, Ken Schwaber, Jeff
Sutherland

Adaptive Software Development

James Highsmith

Crystal

Alistair Cockburn

DSDM

Arie van Bennekum

Adding to the confusion about agile development is the fact that most of the
practices proposed in agile are not novel, and can be practiced in non-agile as well as
agile settings (Beck 1999; Highsmith 2002; McConnell 1996). While the practices are
important, practitioner literature also stresses the importance of the adoption of an agile
“philosophy” within the team, management, and stakeholders as being at least as
important to the development of an agile “ecosystem” (Highsmith 2002). Unfortunately,
both practitioner and academic research has focused primarily on engineering practices
when measuring agile methodology use. This has led to the criticism of agile
methodologies despite their wide adoption, with many criticizing them for lack of rigor
(Boehm and Turner 2004; Stephens and Rosenberg 2003), and others calling for their
use to be limited to small, non-critical projects (Boehm 2002), or not used at all
(Agerfalk and Fitzgerald 2006). However, multiple agile methodology practitioners report

8

that agile methodology practices are being used successfully on projects of all sizes and
complexities (Anderson 2004; Beck and Andres 2004; Highsmith 2002).
For these reasons, there are important reasons to continue to research agile
development. Since most agile practices are not novel, developing a richer
conceptualization of the nature of agile software development methodologies is
required. Since there is disagreement regarding the appropriateness and applicability of
the use of agile methodologies, it is critical that the research community develop a
better understanding of the how and the why of agile development methodologies, and
when they are most impactful.
Purpose of this Study
As described above, the adoption of agile methodologies is reported to be
widespread, but many of the normative claims have yet to be tested. Additionally, the
methodologies that rest under the umbrella of “agile” methodologies are extremely
diverse and sometimes contradictory in their prescribed practices. Further, the specific
practices of most agile methodologies are not unique to agile environments. For these
reasons, building a greater understanding of the salient components of agile
methodologies is crucial.
Second, even if agile methodologies are adopted, there are still disagreements of
whether or not the use of agile methodologies is appropriate in various contexts. It is
reasonable to assume that influences from the environment, whether customer
influences, management influences, corporate culture, or project constraints may
influence the ability for agile methodologies to impact performance. Because of this,

9

developing an understanding of the key environmental and project characteristics that
influence the success of agile projects is an important pursuit.
Therefore, the purpose of this research is to identify the broad components of
agile methodology adoption and use, and to study when and how the use of agile
methodologies impacts project success. To do this, a conceptual and operational
definition of the use of agile methodologies is developed that captures multiple
dimensions of the practice of agile methodologies, and integrate that new construct into
a theoretical model of agile methodology impacts. Formally stated, the research
question is:
How does agile methodology adoption and use impact project success?

Organization of Presentation
The remainder of the dissertation is organized as follows. Chapter Two presents
a conceptual model that provides the theoretical underpinnings of the study with a
review of the supporting literature in information technology, organization theory, and
organizational behavior. The specific portion of this theory and the relationships that
were investigated are discussed in Chapter three via the exposition of the research
model, research questions and propositions. Chapter four describes the details of the
research methodology that was used to investigate and answer the research questions.
Chapter five provides an analysis of the results of the research study. Finally, Chapter
six discusses the major findings of the research, implications for research and practice,
and directions for future research, and concludes the dissertation.

10

Chapter Two: Theoretical Model and Supporting Literature
“All of us, as far as I can remember, thought waterfalling of a
huge project was rather stupid.” – Gerald W. Weinberg,
Project Mercury Software Project Member

Introduction
The creation and use of development methodologies has been a core focus of
the IS community for more than 50 years. During that time, a number of methodological
philosophies and goals have emerged (Hirschheim et al. 1995), the latest of which is socalled “agile” development. A short recapitulation of the evolution of software
methodologies will help to inform the discussion of agile development methodologies.
Hirschheim, Klein & Lyytinen (1995) described the evolution of seven
generations of software development methodologies, and focus on the philosophy
behind the methodologies and the environmental factors present at each stage.
Development methodologies emerged out of a “Pre-Methodology Era” in the 1950s (p
29). As Hirschheim et al. (1995) note:
“At that time [the 1950s] the only conceivable system design task
was programming and specifying computer room operations (Somogyi
and Galliers 1987). To accomplish these complex tasks systems
developers often followed a variety of systematic practices. New practices
were invented as needed, and they were usually very technology
oriented. Those practices which seemed to work…became the
developer’s ‘rules-of-thumb’ and, in a sense, his/her ‘methodology’”
(Hirschheim et al. 1995p. 29)
In the “ad-hoc practices” of the 1950s and 1960s, development methodologies
involved users speaking to programmers, after which programmers would build
systems. From this pre-methodology era emerged more structured development
methodologies in the 1970s. This transition was caused by the shift of computing
11

systems from being primarily designed for specialized and definable scientific
applications, to being designed to support complex business and other real world
processes (Avison and Fitzgerald 1998). However, even as methodologies evolved to
manage the emerging issues of software development and engineering, software
development projects continued to be marked by significant issues such as being
considered incomplete, unstable, inflexible, and generally failing to meet the
expectations of users and management (Avison and Fitzgerald 1998).
By the 1970’s the state of software engineering was such that a “software crisis”
was said to exist (Brooks 1987). Project estimates were grossly inaccurate, and projects
were often delivered at as much as 100% over initial time and budget estimates or
more. Methodologies and technologies were repeatedly proposed to be the magic bullet
to “solving” the crisis. If teams would simply use a particular methodology, language, or
programming paradigm, software projects would succeed. However, it was recognized
that these methodologies, development standards, and other practices that had
emerged did not assist in improving systems’ essential functionality (Brooks 1987). In
short, the “magic bullet” of software development methodologies was never found.
Table 2.1 lists seven generations of software development methodologies as
described by Hirschheim et al. (1995). While the details of these generations are
somewhat diverse, the repeated theme in the evolution is that methodologies emerged,
were tried, and found to be lacking in some important way, which affected the
underlying philosophy of development, and a new methodology was developed that was
both responded to previous observation, and was in congruence with the emergent
philosophy. Even so, after generations of software development methodologies, there is

12

a continued perception of a crisis in software engineering. U.S. Government audits, and
private research agree that software projects are widely problematic, with late delivery,
budget overruns and feature incompleteness being significant continued issues
(Standish Group 2009).
The Evolution of Methodology - a Quest for Fit
As described above, the history development methodologies can be described as
one of evolution and a quest for fit with environmental contingencies. In this dissertation,
the impacts of agile methodologies on project success through the lenses of
contingency theory, organizational learning, and team adaptation is described. The
application of contingency theory has a long history within the IS field.
Contingency theory has been recognized as an important theoretical lens, and a
building block for theoretical development in a variety of disciplines (Venkatraman
1989). Contingency theory in its simplest form states that the relationship between a
cause (X) and effect (Y) is impacted by another variable, a contingency (C). Thus, the
effect of X on Y will be different when C is low versus when C is high (Donaldson 2001).
Contingency theory states that organizations recursively organize around creating a fit
between internal structures, practices, and procedures, and the perceived contingent
factors of the environment.

13

Table 2.1. Summary of Generations of Development Methodologies
Generation

Principal Management or Organizational Issue

View of Environment

Perception of
Process
Technical

1. Formal Life-Cycle
Approaches

Control of SDLC; guidance of analysts/
programmers through standardization

In equilibrium

2. Structured Approaches

Productivity (information requirements quality
assurance to meet the ‘5Cs’ – clear, concise, costeffective, comprehensive and complete
specifications); better maintainable systems; control
of analysts/programmers (division of labor, e.g.
Kraft, 1977)

In equilibrium

Technical, but with
social consequences

3. Prototyping and
Evolutionary Approaches

Speed and flexibility of development (SDLC
methodologies take too long and are too rigid);
overcoming analysts/use communication gap with
technical specifications; emphasis is on getting the
right kind of system vs. getting the system right

In equilibrium

Technical, but with
social consequences

4. Socio-Technical,
Participatory Approaches

Control of development by users through
participation; conflict management in development;
joint optimization: cost-effectiveness and better
QWL through technology

In conflict, and in
need of compromise

Joint technical and
social process

14

Table 2.1. (cont’d)
5. Sense-Making and
Problem Formulation
Approaches

Dealing with multiple perspectives in problem
framing; software development as social reality
construction.

In conflict, and in
need of compromise

Mostly a social
process

6. Trade-Union Led
Approaches

Labor/management conflict; workers’ rights;
industrial democracy.

In conflict due to
class differences

Almost completely
social, collaborative

7. Emancipatory
Approaches

Overcoming barriers to effective communication due In conflict, in need of
rational compromise
to power and social differentiation (e.g. blockage,
bias, jargon, ambiguity); eliminating repression and
furthering emancipatory effects of development
(development as social learning and therapy, e.g.
questioning dominant forms of thinking, improving
access to facts and arguments, removing
unwarranted uses of power, etc.)

15

Political

In the IS literature, the concept of fit has been applied at many levels, spanning
from the organization strategy level (e.g. – IT strategy alignment) to the individual level
(e,g, - task-technology fit). In this study, the view is utilized that contingencies impact at
the business process level (Barki et al. 2001), and that particular environmental
characteristics indicate that a particular set of process characteristics will deliver a
better “fit”, and potentially better outcomes.
Fit is an evolutionary process. Organizations attempt to fit their processes to the
perceived environment, then modify their structure and practices when the fit between
the structure and practice is less than satisficing (Donaldson 2001). This recursive
structuring, feedback, and re-structuring is the hallmark of organizational learning
(Argyris and Schoen 1978). Organizational learning requires the continual testing of the
organization’s “theory-in-use” by its members.
“They detect an error in organizational theory-in-use, and they
correct it. This fundamental learning loop is one in which individuals act
from organizational theory-in-use, which leads to match or mismatch of
expectations with outcome, and thence to confirmation or disconfirmation
of organizational theory-in-use” (Argyris and Schoen 1978). In this
manner, as individuals detect a lack of fit between the environment and
task, they make adaptation moves in order to regain fit with the new
understanding of the environment.”
The concept of teams and adaptation will become a key lens for my analysis.
One theory of team adaptation, team adaptability theory defines team adaptability as:
“a change in team performance, in response to a salient cue or
cue stream, that leads to a functional outcome for the entire team. Team
adaptation is manifested in the innovation of new or modification of
existing structures, capacities, and/or behavioral or cognitive goaldirected actions.” (Burke et al. 2006)
In this definition three particular items regarding team adaptability should be
noted: (1) adaptation is a response to feedback, (2) adaptation manifests in the
16

modification of team structure, routine and motivations, (3) adaption is goal-directed.
Importantly, Adaptability is achieved by a recurrent cycle of planning, trial, learning, and
feedback.
Finally, it should be noted that teams must adapt when the situations in which
they are acting are uncertain, or unpredictable. Uncertainty has been characterized as
the fundamental problem that faces any organization (Thompson 2003). The concept of
uncertainty has been utilized widely in information and decision theory, in economic
theory, and in organizational theory. While definitions of uncertainty in the information
and decision theory approach the concept of uncertainty mathematically, this study
adopts the definition of uncertainty simply as “the degree to which future states of the
world cannot be anticipated and accurately predicted” (Pfeffer and Salancik 2003, p.
67). In the organization theory literature, uncertainty has been posited to arise from a
number of sources. Lawrence & Lorsch (1967) described uncertainty as consisting of
(1) the lack of clarity of information regarding the situation, (2) the length of time before
receiving definitive feedback, and (3) the uncertainty of causality. Thompson (2003, p.
159) also viewed uncertainty as emerging from three sources, (1) the lack of ability to
determine cause and effect in general, (2) the contingent nature of action, in which
outcomes

are

partially

dependent

upon

environmental

elements

and

(3),

interdependence of organizational components. Duncan (1972) described a similar
three-component view of uncertainty as emerging from (1), the lack of clarity regarding
how environmental factors influence the situation, (2) inability to predict risks of the
outcomes of actions, and (3), the inability to confidently predict the likelihood of success
of actions. Based upon these viewpoints, Duncan developed a model of uncertainty that

17

views uncertainty as a function of the complexity and dynamism within an environment.
For this study Duncan’s viewpoint is adopted, specifically that the complexity and
dynamism of an environment are directly related to uncertainty - the inability to
anticipate and accurately predict future states of the environment and organization.
The remainder of this chapter will develop these lenses, and explain how they
provide a conceptual underpinning for the impacts of agile methodologies.
The remainder of the chapter is structured as follows. First, the phenomenon of
agile development methodologies is described, including a discussion of software
methodologies in general and a comparison of agile methodologies with traditional planbased methodologies in particular is developed. Next, the theoretical lenses described
above, are discussed, the role of team adaptation and learning is considered, and the
specifics of agile philosophy and practices are presented within this discussion. After
this, the concept of project success is addressed and the role of external contingencies
is considered. Finally, a review of the literature on the impacts of agile development is
presented.

18

What is a Software Development Methodology?
There has been long debate about whether “methodology” or “method” is the
right term, what constitutes a methodology, and how practices and methodologies differ
(Wynekoop and Russo 1995). Table 2.1 displays several selected definitions of method
1

(or methodology) .
These definitions are consistent in describing that a development method
emerges out of a philosophical view, defines certain practices to be carried out in the
process of developing the system, and provides guidance for the organization and
management of work. Taking the understanding that any method is more than the sum
of its practices is key to understanding why agile methodologies are different from
traditional methodologies.
Because of the importance of underlying philosophy in the motivation of any
method’s practices, a comparison is made between the philosophy of agile development
the philosophy of the earlier, traditional method, the structured, or waterfall method.
Many have focused on engineering practices of specific agile development method. In
contrast, the focus of this dissertation is on several key philosophical differentiators
between the traditional and agile methodologies that are shared across all agile
methodologies. The first of these is the concept of delivery – in other words, when does
the project deliver a product, and how does that delivery enable teams to adapt to
change.

1

I will utilize the term “methodology” for the remainder of this dissertation.

19

Table 2.2. Selected Definitions of Software Development Method[ology]
Source
Definition
Avison et al.(1998)
"...a collection of procedures, techniques,
tools, and documentation aids which will
help the systems developers in their
efforts to implement a new information
system. A methodology will consist of
phases, themselves consisting of subphases, which will guide the systems
developers in their choice of the
techniques that might be appropriate at
each stage of the project and also help
them plan, manage, control, and evaluate
information systems projects...It is usually
based upon some philosophical view,
otherwise it is merely a method, like a
recipe." (p. 8)
Hirschheim et al. (1995)

"An information systems development
methodology is an organized collection of
concepts, methods, beliefs, values, and
normative principles supported by
material resources." (p. 22)

Conboy (2009)

An development method encompasses
the complete range of practices involved
in the process of designing, building,
implementing, and maintaining an
information system, how these activities
are accomplished and managed, the
sequence and frequency of these
activities, as well as the values and goals
of all of the above. (p. 329)

Wynekoop & Russo(1995)

A systematic approach to conducting at
least one complete phase (e.g. design or
requirements analysis) of software
production, consisting of a set of
guidelines, activities techniques and tools,
based on a particular philosophy of
system development and the target
system (p. 66)

20

Waterfall – Big Delivery
The Software Engineering Institute’s Capability Maturity Model framework is
currently the most accepted standard for measuring software engineering process
excellence. It adopts the philosophy of the waterfall method, namely that the software
development process is a definable process (Paulk et al. 1995; Schwaber 1996), and
that software project performance can be predicted through the establishment of
repeatable processes (Paulk et al. 1995). The classic waterfall model is represented in
Figure 2.2. The waterfall method, characterized by the gate system familiar in physical
engineering, had a variety of stage numbers and names, but all possessed aspects of
the following:
•

Analysis

•

Design

•

Construction

•

Testing

•

Deployment

The analysis stage consists of the identification of the business problem, the
nature of the process being addressed by the project. The goal of the analysis phase is
to identify the “facts” (Avison and Fitzgerald 1998) as they are understood, the details of
the kind of data that had to be included in the system, requirements, constraints,
exceptions, and the feasibility of the system. The outputs of the analysis phase of the
lifecycle are detailed documents that describe the problem space, the requirements of
the system, and the plan for completing the project. It is commonly understood that the

21

stages of the waterfall method do not overlap, and that backtracking is generally limited
to moving one step back up the process (Boehm 1988).
The design stage(s) consist of taking the products of the analysis phase and
designing all of the relevant portions of the system, the inputs and outputs, and the
boundaries of the system (Avison and Fitzgerald 1998). While it was understood that
the design & requirements phases might be somewhat iterative, it was accepted that
after this phase, the cost of changes to the system based upon changing requirements
would increase exponentially (see Figure 2.1). This led to the establishment of
“requirements freezes” and sign offs from customers on requirements at this point in the
process.

Figure 2.1. The “Waterfall” Method
22

Once the design stage(s) are completed, the Coding and Debugging stage
proceeds. During this stage, the development team builds the system, performs unit
testing and debugging, and integration testing of the system. Finally, the completed
system is delivered for testing in the System Testing stage.
The structured method had definite strengths over the previous, ad-hoc model of
development. First, it established formal requirements and design procedures. This
made for better code quality, and higher probability of acceptance by users. Second, it
recognized explicitly the need for formal phases of testing and ongoing maintenance
(Boehm 1988).
The method relied on the creation of full documentation of requirements at the
very early stages of the project. As Boehm articulates, this is very effective for some
classes of software, but not for complex and interactive end-user focused systems,
which describes the majority of software projects undertaken (Boehm 1988). These
limitations led to the continued evolution of development methodologies.

23

Figure 2.2. The Cost Curve of Change Under The Waterfall Method
The philosophy of the waterfall method assumes the “facts” about the system can
be elicited up front with relative accuracy and certainty. This implies that the
organization is generally in equilibrium, and that the environment will create little change
of the requirements of the system during development. Second, the method assumes
that the software development staff can build software from the documented
descriptions of analysts. Third, the method assumes that the cost of change (Figure 2.2)
becomes significantly higher as time passes, and that changes later in the project
should be avoided, hence the focus on early definition and codification of requirements.
Finally, the method assumes a linear delivery model, with successive completion of
each phase, although later views of the waterfall model allowed for feedback loops
between adjacent phases (Boehm 1988). Because of this, the pure waterfall method is
24

not good at handling changing or uncertain requirements (Boehm 1988). While projects
often spent up to 40% of their budgets on requirements elicitation, analysis, and design,
this did not prevent changing requirements, which often lead to costly system
modifications. Programmers believed that the users continually changed their minds,
users believed that the programmers did not give them an accurate understanding of
what the system would deliver, and analysts were accused of poorly eliciting
requirements (Hirschheim et al. 1995).
Additionally, the timescale of project delivery under the waterfall method is
usually months, and often, years. Projects in general have a tendency to escalate due
to the need for managers to justify their previous decisions, the psychological impact of
sunk costs, and the desire to avoid the negative professional impacts to the project
team members (Keil et al. 2000). Because the traditional waterfall method delivers most
of the project value at the end of the project causes there to be a long delay before
value is delivered, and before there is the ability to measure the progress and success
of the project, leading to more pressure to escalate to deliver value from the project.
As can be seen from Figure 2.2 and the description, delivery is conceived of as
an event that occurs usually once, usually at the end of a project. This philosophy has
dominated the practice of software development. However, the delivery philosophy of
agile methodologies is significantly different.
Agile Methodologies – Continuous Delivery
In contrast to the practice of the waterfall method, agile methodologies claim a
philosophy to deliver value early iteratively throughout a project. While the assumption
of waterfall methodologies was that requirements could be elicited accurately up front

25

as a “definable” process, agile methodologies approach software development as an
“empirical” process. Schwaber (Schwaber 1996) describes these two process types:
“If a process can be fully defined, with all things known about it so
that it can be designed and run repeatably [sic] with predictable results, it
is known as a defined process, and it can be subjected to automation. If
all things about a process aren't fully known-only what generally happens
when you mix these inputs and what to measure and control to get the
desired output-these are called empirical processes….a defined process
is predictable; it performs the same every time. An empirical process
requires close watching and control, with frequent intervention. It is
chaotic and unrepeatable, requiring constant measurement and control
through intelligent monitoring.” (Schwaber 1996)
Hence, agile methodologies are based on the philosophical assumption that a
software method must control an uncertain and ambiguous process. However, neither
the practitioner nor research literature has universally adopted this viewpoint. Table 2.4
provides several definitions of agility that have emerged from the practitioner and
academic literature. While not completely aligned with each other, there are multiple
concepts that are common to all of them.
First, these definitions assume the presence of an interaction between the
software development team and its organization (environment). Furthermore, these
definitions stress the continuous nature of this interaction. Second is the notion of being
able to sense, to learn from, and to respond to change. Third is the concept that through
this sensing and responding, swift adjustments that increase customer value can be
achieved. This customer value can be achieved via the early delivery of functionality.
Additionally, both Highsmith and Conboy refer to the agile development team
possessing the ability to “create” change as well as respond to change. By using this
terminology, they indicate that the product of the work of the team – the finished
software – provides a cue to the environment. These cues are not available in the
26

traditional development structure, as working code is rarely delivered and deployed
before the end of the project. However, in agile development, the regular delivery of
working software is a cue that assists the organization in its ability to understand how
well the system will fit into its work processes. Because of this, the system that emerges
from the project itself becomes an agent of change (DeSanctis and Poole 1994;
Orlikowski 1992).
The definitions of “agility” provided in Table 2.3 are all consistent in their
recognition of the need of a method to be structured in a way to receive and respond to
environment changes. This indicates the wide recognition of development as being an
empirical, ambiguous process. However, while the general philosophy of agile is
consistent across agile methodologies, the specifics of the methodologies exhibit
significant variance. The next section describes the similarities and differences between
the agile methodologies.
Definable vs. Empirical Processes
The ability to forecast the outcome of the development process is at the core of
the differences between the philosophies of the traditional and agile methodologies.
These methodological perspectives hold conflicting views of the “definability” of the
systems development process. If a process is “definable”, its inputs, process tasks, and
outputs are understood well enough that, given a specific set of inputs, a predictable
and consistent set of outputs can be expected. If a process is “empirical” (or nondefinable), control of the process cannot be completed via the planning and
standardization of process steps. Rather, an empirical process in controlled by the
repeated inspection of its output, associated feedback, and adaptations made based

27

upon the results of the inspection, which leads to regularity of process (Ogunnaike and
Ray 1994). Without the necessity to manage the non-routine, adaptation is
unnecessary.
In a group process situation, there are several kinds of inputs. First the resources
provided to the team, and the skills of the team members at the beginning of the
process can be considered inputs. Additionally, the requirements, or the understanding
of the nature and properties of the group’s goal are also inputs to the process. The
steps necessary to transform the inputs to the process into the goal are the process.
Finally, the final product of the group, whether it is a decision, a document, or a software
system is the output.
Table 2.3. Selected Definitions of “Agile/Agility”
Source
Definition
IBM Rational (Ambler 2009)
“Agile is the use of continuous stakeholder feedback
to produce high-quality consumable code through
user stories (or use cases) and a series of short
time- boxed iterations.”
(Lyytinen and Rose 2006)

“In the context of information system development
(development), agility can be defined as an
development organization’s ability to sense and
respond swiftly to technical changes and new
business opportunities.”

(Conboy 2009)

“The continual readiness of an development method
to rapidly or inherently create change, proactively or
reactively embrace change, and learn from change
while contributing to perceived customer value
(economy, quality, and simplicity), through its
collective components and relationships with its
environment.”

(Highsmith 2002)

“Agility is the ability to both create and respond to
change in order to profit in a turbulent business
environment.”

28

When all the components of this process – inputs, process, and output – are well
understood, unambiguous, and stable, the process is considered to be “definable” and,
therefore,

predictable.

Most

traditional,

plan-based

software

development

methodologies are built on the concept of being able to, at the beginning of a project,
describe the process needed to complete the project including necessary tasks, their
sequence, and inputs and outputs. This assumes a definable process, with steps that
can be predicted. This quest for predictability leads to methodologies such as IBM’s
Worldwide Project Management Method that defines thousands of potential
development process tasks. Even so, as Schwaber (2002) points out, “for the defined
control mechanisms to work, these methodologies must define each process with
enough accuracy that the resultant noise does not interfere with its repeatability, or the
predictability of the outcome.” In short, if a process is truly repeatable, its steps should
be defined to such an extent that a team could follow the process like a recipe and,
without variation, deliver the anticipated outputs.
In contrast, agile methodologies either explicitly or implicitly adopt the philosophy
that the development process is empirical. For this reason, most have adopted the
“iterative” development approach discussed above, which attempts to balance the
needs for flexibility and adaptation with the need for stability for execution. While agile
teams generally develop plans for delivery, they do so in short bursts (iterations), which
allow them to commit to delivery, while leaving room to adapt repeatedly based upon
feedback. These differences in assumptions regarding definability lead to significantly
different planning, management and control (Eisenhardt and Tabrizi 1995).

29

Agile Development – a Description
Agile Development methodologies are, based on the concept of iterative delivery.
Iterations consist of short, time-boxed deadlines, usually no longer than 4-6 weeks
(Highsmith 2002; Schwaber 1996; Schwaber and Beedle 2002). Prototyping and other
evolutionary methodologies emerged before agile methodologies. However, agile
development is differentiated in that in most cases each iteration includes all or most of
the steps of the waterfall method. This means that the code necessary to deliver the
requirements selected for the iteration is designed, built, tested, and delivered (Beck
and Andres 2004; Highsmith 2002; Schwaber and Beedle 2002). This iterative delivery
loop is illustrated in Figure 2.3.

Figure 2.3. Agile Method Feedback Loop
In this delivery model, high-level concept and requirements lists may be
developed for the entire system up front. However, only those requirements agreed

30

upon for the current iteration need to be completed in order for the iteration to begin.
Once the iteration begins, the features and associated requirements to be delivered in
the iteration are frozen. The team then completes design, development, testing and
delivery for all features for the iteration. While this does not necessarily mean that the
code is deployed to production, the understanding that the code should be deployable
at the end of any iteration, at the discretion of the user is explicit in several
methodologies, including Extreme Programming and Scrum (Beck 1999; Highsmith
2002; Schwaber and Beedle 2002).
Teams require some stability in order to take action (Argyris and Schoen 1978;
Gersick and Hackman 1990; Okhuysen 2001). By organizing in the fashion described
above, the agile development team attempts to create a balance between stability and
flexibility. The team provides flexibility and adaptability by receiving feedback cues from
the environment at least at the end of each iteration. These feedback cues are of high
quality because, rather than being based upon representations of the system as would
occur during the analysis and design phase of a waterfall project, they are based upon a
review of the working software (Lave 1991). This working software is a far richer
medium for users to evaluate whether the system that is being built is suitable for the
task.
Additionally, because the team chooses a small portion of the system’s features
to build during an iteration and then freezes the requirements for those features during
the execution of the iteration, they provide stability for action. Additionally, this inprocess requirements freeze provides the efficiency of buffering the team from the
environment, and allowing the team to execute the process as efficiently as possible

31

(Thompson 2003). In a real sense, this process enables the team to behave as if the
software development process is definable during the iteration, while embedding these
episodes of definability within a structure of empirical reflection and adaptation.
To illustrate an episode of definability, in each iteration, a small subset of the
system features is chosen to be addressed. These features are usually the most
valuable (as defined by the user) remaining to be completed (Schwaber and Beedle
2002). During the iteration, while these features are being completed, requirements for
other parts of the system can be changed. Changes in requirements for the portion of
the system under development are postponed until a later iteration. The timeframe of
the iteration is only a maximum of four to six weeks (Schwaber and Beedle 2002) (Beck
1999), and often substantially less (Highsmith 2002). Therefore, very few requirements
that emerge during a cycle will rise to the level of importance as to merit an exception to
this rule.
However, at the end of this episode of definability, the team intentionally emerges
from behind the buffer and displays the results of the iteration. This review of the
functional code allows for a determination to be made regarding the fit of the developed
software to the requirements of the environment as it stands at the end of the iteration.
Incongruities spawn new feature requests, and are prioritized as are any other features.
After each iteration is completed, all agile methodologies prescribe a
retrospective assessment. During this time, the team reflects upon the process
performed during the priori iteration, and the source of any incongruities between
requirements and delivered software. Additionally, they may adapt by implementing
process changes speculated to generate higher performance in the next iteration.

32

Finally, agile development practitioners propose that the time-boxed iteration
allows the team to eliminate several traditional software development artifacts. While
traditional methodologies focus on up front planning and documentation, the iterative
time-box approach allows for learning during the project. Rather than working to provide
a full enumeration of requirements as a control mechanism up front, control is
established via the time-box (Highsmith 2002). Additionally, because the system is
delivered each iteration, the system itself acts as documentation, removing the need to
build and maintain abstract documentation artifacts (Consortium 2002-2011).
Agile practitioners (Beck 1999; Highsmith 2002) claim that this structure of
adaptive iterations, enabled by agile method engineering and process practices,
provides a lower cost curve of change. Because the system is built in small steps,
delivered in short durations, the agile methodologists claim that the cost of change
curve that impacted waterfall projects to a far higher degree in later stages can be
significantly flattened (See Figure 2.4).

33

Figure 2.4. The Cost of Change Curve Under Agile Method (Beck 1999)

The Practitioner Literature on Agile Methodologies
One of the first formal articulations of the agile philosophy came with the
publication of the Manifesto for Agile Software Development, a document outlining socalled “agile” development as an alternative mode of software production. The
signatories of the manifesto had already been practicing various “agile” methodologies
for several years (Beck 1999; Schwaber 1996). As such, the Manifesto is a statement of
shared philosophical beliefs held by the proponents of common agile methodologies.
The Agile Manifesto is structured into a value statement and twelve foundational
principles. The value statement is reproduced here:
“We are uncovering better ways of developing software by doing it
and helping others do it. We value:
•

Individuals and interactions over processes and tools.
34

•

Working software over comprehensive documentation.

•

Customer collaboration over contract negotiation.

•

Responding to change over following a plan.

That is, while there is value in the items on the right, we value the
items on the left more.” (Beck et al. 2001)
As explained by Martin Fowler and James Highsmith in their exposition of the
Agile Manifesto (Fowler and Highsmith 2001), this purpose statement is a description of
the foundational philosophy of agile development. In each of the value statements, two
valued items are contrasted. Fowler and Highsmith are clear, both items on each bullet
point are valued, but the item on the left is of more value. Fowler & Highsmith’s
commentary throughout their article indicates that the value statements drive decisions
that must be made when change arises:
“Contract negotiation, whether through an internal project charter
or external legal contract, isn't a bad practice, just an insufficient one.
Contracts and project charters may provide some boundary conditions
within which the parties can work, but only through ongoing collaboration
can a development team hope to understand and deliver what the client
wants.
No one can argue that following a plan is a good idea—right? Well,
yes and no. In the turbulent world of business and technology,
scrupulously following a plan can have dire consequences, even if it's
executed faithfully. However carefully a plan is crafted, it becomes
dangerous if it blinds you to change. We've examined plenty of
successful projects and few, if any, delivered what was planned in the
beginning, yet they succeeded because the development team was agile
enough to respond again and again to external changes.” (Fowler and
Highsmith 2001)
Therefore, according to the agile manifesto, an agile philosophy is one in which
individuals are valued, working software is the key indicator of progress, customer
interaction is rich and continuous, and change is an accepted part of the development
35

process. The agile manifesto further defines its philosophy via the 12 Principles of Agile
Development (see Table 2.4).
The principles can be loosely categorized into five categories. (1) Iterative
Delivery – all agile methodologies stress this principle, namely the importance of
delivering incrementally. (2) Product Focus – the product – working software – should
be the measure of success and project value delivery. (3) Self-directed team – the
team should have the power to organize itself, and should be responsible for
maintaining its process. (4) – Feedback – teams should have an involved user, and
should communicate regularly. (5) Responsiveness – The team should receive
feedback, but also be willing to act in accordance with the information contained in the
feedback. At a high level, these five categories of principles relate to both the
philosophy of agile methodologies, and the actions or practices of agile methodologies.

36

Table 2.4. The 12 Principles of the Agile Manifesto
Principle
Clarification (Fowler and Highsmith 2001)
Our highest priority is to satisfy
the customer through early and
continuous delivery of valuable
software.

“…we need to understand that customers don’t
care about documents, UML diagrams or legacy
integration. Customers care about whether or not
you’re delivering working software to them every
development cycle—some piece of business
functionality that proves to them that the evolving
software application serves their business
needs.”

Welcome changing requirements,
even late in development. Agile
processes harness change for the
customer’s competitive
advantage.

“Rather than resist change, the agile approach
strives to accommodate it as easily and efficiently
as possible, while maintaining an awareness of
its consequences. Although most people agree
that feedback is important, they often ignore the
fact that the result of accepted feedback is
change. Agile methodologies harness this result,
because their proponents understand that
facilitating change is more effective than
attempting to prevent it.”

Deliver working software
frequently, from a couple of weeks
to a couple of months, with a
preference for the shorter
timescale.

“For many years, process gurus have been
telling everyone to use an incremental
or iterative style of software development, with
multiple deliveries of ever-growing functionality.
While the practice has grown in use, it’s still not
predominant; however, it’s essential for agile
projects. Furthermore, we push hard to reduce
delivery cycle time.”

Business people and developers
work together daily throughout the
project.

“Many folks want to buy software the way they
buy a car. They have a list of features in mind,
they negotiate a price, and they pay for what they
asked for. This simple buying model is appealing,
but for most software projects, it doesn’t work.”

37

Table 2.4. (cont’d)
Principle

Clarification (Fowler and Highsmith 2001)

Build projects around motivated
individuals, give them the
environment and support they need
and trust them to get the job done.

“Deploy all the tools, technologies and processes you
like, even our agile processes, but in the end, it's
people who make the difference between success
and failure.”

The most efficient and effective
method of conveying information with
and within a development team is
face-to-face conversation.

“Inevitably, when discussing agile methodologies, the
topic of documentation
arises. Our opponents appear apoplectic at times,
deriding our "lack" of documentation. It's enough to
make us scream, "the issue is not documentation—
the issue is understanding!" Yes, physical
documentation has heft and substance, but the real
measure of success is abstract: Will the people
involved gain the understanding they need?”

Working software is the primary
measure of progress.

Too often, we've seen project teams who don't realize
they're in trouble until a short time before delivery.
They did the requirements on time, the design on
time, maybe even the code on time, but testing and
integration took much longer than they thought.

Agile processes promote sustainable
development. The sponsors,
developers, and users should be able
to maintain a constant pace
indefinitely.

Agility relies upon people who are alert and creative,
and can maintain that alertness and creativity for the
full length of a software development project.
Sustainable development means finding a working
pace (40 or so hours a week) that the team can
sustain over time and remain healthy.

Continuous attention to technical
excellence and good design
enhances agility.

When many people look at agile development, they
see reminders of the "quick and dirty" RAD (Rapid
Application Development) efforts of the last decade.
But, while agile development is similar to RAD in
terms of speed and flexibility, there's a big difference
when it comes to technical cleanliness. Agile
approaches emphasize quality of design, because
design quality is essential to maintaining agility.

38

Table 2.4. (cont’d)
Principle

Clarification (Fowler and Highsmith 2001)

Simplicity--the art of maximizing the
amount of work not done-- is
essential.

It's easier to add something to a process that's too
simple than it is to take something away from a
process that's too complicated. Hence, there's a
strong taste of minimalism in all the agile methods.
Include only what everybody needs rather than
what anybody needs, to make it easier for teams to
add something that addresses their own particular
needs.

The best architectures,
requirements, and designs emerge
from self- organizing teams.

that the best designs (architectures, requirements)
emerge from iterative development and use rather
than from early plans. The second point of the
principle is that emergent properties (emergence, a
key property of complex systems, roughly
translates to innovation and creativity in human
organizations) are best generated from selforganizing teams in which the interactions are high
and the process rules are few.

At regular intervals, the team
reflects on how to become more
effective, then tunes and adjusts its
behavior accordingly.

Agile methods are not something you pick and
follow slavishly. You may start with one of these
processes, but we all recognize that we can't come
up with the right process for every situation. So any
agile team must refine and reflect as it goes along,
constantly improving its practices in its local
circumstances.

A full comparison of major methodologies is beyond the scope of this
2
dissertation . For a full

Table 2.5 provides a short summary of the practices

encouraged by each method. In brief, the majority of the methodologies focus on
coordination and communication practices, rather than development and coding
practices. The significant exception to this statement is Extreme Programming (XP). As
illustrated in Table 2.5, most methodologies focus on coordination and communication
practices, indicating the shared focus on developing the ability to both receive and

2 For an excellent exposition of the various agile methods, see Highsmith (2002)

39

process feedback. Because of the various methodologies’ unique evolution, they define
differing enabling processes and practices that are specific to their focuses. This is why
some methodologies have developed enabling practices that are in conflict. The
concept of enabling processes in agile methodologies will be discussed more fully
below.
Finally, the importance of the “tools” of agile development – source code control,
continuous integration software, and automated unit testing has been stressed since the
beginning of the agile movement (Beck 1999; Highsmith 2002). The practitioners stress
that the agile practices are enabled by a complex system of people, processes and
technology, and that these entities reinforce and enable one another.
In summary, the philosophy of agile development stresses the value of
completed code that meets customers’ needs. In order to delver software based upon
that philosophy, agile methodologies each propose a set of direct and supporting
practices, processes, and technologies that they claim will enable higher levels of
success in software delivery. However, we argue that the defining practices of agile
methodologies, and those most likely to directly impact project success, are those
practices that enable feedback.
We now turn to the theoretical lenses through which we will build my theoretical
model of the impacts of agile software development on project success. My analysis
relies on three theoretical perspectives for grounding. First, uncertainty theory describes
when agile methodologies are most likely to be impactful. Second, dynamic capability
theory describes how organizations develop the ability to sense and respond to
environmental cues under uncertainty. Finally, team adaptability theory describes a

40

complex network of team norms, practices and processes that enhance performance
under team-level uncertainty.
Table 2.5. Defined Practices of Major Agile Methodologies
Practice
XP
XP
SCRUM FDD Crystal
Clear
(Beck
(Beck
99)
04)
Pair Programming
Yes
Yes
Collective Code Ownership
Yes
Yes
Continuous Refactoring
Yes
Continuous Integration
Yes
Yes
Yes
Unit Testing
Yes
Yes
Yes
Yes
Coding Standards
Yes
Planning Game
Yes
Yes
Small Releases
Yes
Yes
Yes
Yes
Metaphor
Yes
Simple Design
Yes
Sustainable Pace
Yes
On-site customer
Yes
Daily Short Meeting
Yes
Yes
Iteration Retrospective
Yes
Yes
meeting
Formal Iteration Review
Yes
Yes
(with customer)
Sit Together
Yes
Yes
Informative Workspace
Yes
Weekly Cycle
Yes
Slack
Yes
Incremental Design
Yes
Whole Team
Yes
Energized Work
Yes
Stories
Yes
Yes
Yes
Yes
Quarterly Cycle
Yes
Ten-minute build
Yes
Test-first programming
Yes
Real Customer Involvement
Yes
Yes
Team Continuity
Yes
Root-cause analysis
Yes
Code & Tests
Yes

41

ASD

Yes

Yes

Practice

Daily Deployment
Pay-per-use
Shrinking Teams
Single Code Base
Negotiated Scope Contract
Develop an overall model
Build a Features List
Plan by Feature
Design by Feature
Build by Feature
Chief Programmer
Class Owner
Feature Teams
Project Documentation
Develop a Mission Statement
Time boxed cycles
Burndown

Table 2.5. (cont’d)
XP
XP
SCRUM
(Beck
(Beck
99)
04)
Yes
Yes
Yes
Yes
Yes
Yes

Yes

Yes

FDD

Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes

Crystal ASD
Clear

Yes
Yes

Yes
Yes
Yes

Yes

Uncertainty: Complexity and Change
As noted above, uncertainty has been characterized as the fundamental problem
that faces any organization (Thompson 2003). Uncertainty means “the degree to which
future states of the world cannot be anticipated and accurately predicted” (Pfeffer and
Salancik 2003, p. 67). In the organization theory literature, uncertainty has been posited
to arise from a number of sources. The complexity and dynamism of an environment
are directly related to the inability to anticipate and accurately predict future states of the
environment and organization.
Complexity of various forms has been shown to impact the success of software
development projects (Xia and Lee 2005). Software projects are often highly complex
by nature, because they must manage both the technical complexity of the project
environment, and the organizational complexity of the environment in which the project
42

takes place (Baccarini 1996; Brooks 1995; Kirsch 1996; Kirsch et al. 2002; Xia and Lee
2005). IS Project uncertainty therefore is based upon both the structural and technical
complexity of the project, as well as the complexity that comes from organizational and
requirements dynamism (Williams 1999; Xia and Lee 2005).
Structural complexity refers to the complexity of the organizational environment
in which the project takes place. Project teams are often temporary in nature, and utilize
members of multiple departments and other functions. Structural complexity refers
specifically to the size of the project team, the number of departments for which the
team is attempting to deliver a solution, the number of departments to which the team
members report, and the geographic distribution of the project team.
As the number of departments that are represented in the project, whether as
stakeholders providing requirements for the team, or as members of the team
increases, leads to an increase in the difficulty of establishing well-defined and stable
project goals. The lack of these defined and stable goals has been posited as a key
driver of project complexity and reduction of project success (Turner and Cochrane
1993).
Project teams require regular feedback to ensure that they are progressing
toward completion. This feedback is obtained through formal status reporting,
serendipitous discussions, ambient awareness within the team area, and other forms of
direct contact. Structural complexity increases when the mechanisms that occur
naturally in team environments are impeded due to the lack of direct communication.
Finally, structural complexity is associated with the criticality of the project and its
associated deliverables. Traditionally, project tasks are either defined as “critical” or “not

43

critical”. Critical tasks are those that, if delayed, cause delay in the entire project
(Kuchta 2001; PMBOK 2000). However whole projects are recognized as having levels
of criticality. When projects are critical, their potential positive and/or negative impact to
the organization is greater. However, in project portfolio management and software
method literatures, the criticality of a project is generally measured in terms of its
potential negative impact.
Technical complexity arises from the combination of scope and structure of the
project system, and the ambiguity regarding the use of various technologies to complete
the project (Shenhar and Dvir 1996).
The dynamic nature of the systems development environment is reflected in the
extent to which the project team must respond to changing requirements throughout the
duration of the project. Changing requirements are caused by ambiguity in initial
requirements definition, as well as actual volatility in the system requirements (McKeen
et al. 1994; Meyer and Curley 1991; Ribbers and Schoo 2002).
Uncertainty in systems development environments is defined as the extent to
which the development team cannot accurately identify the necessary inputs, tasks, and
sequencing necessary to ensure the desired outcome of the systems development
process, due to the impacts of the complexity of the environment and deliverable, and
the dynamic nature of the environment. In an uncertain environment, organizations must
adapt, rather than rely on pre-defined plans to be successful. The next section contains
a discussion of team adaptability theory, which describes mechanisms through which
teams may react to uncertain environments.

44

Team Adaptability & Organizational Learning
A software development team is a specialized form of small group. Small group
researchers in social psychology and organizational behavior have investigated the
interaction between teams and their environments at length. Groups originally were
viewed as a unit with firm boundaries, conceived as simply the sum of the properties of
the individual team members. Teams were conceived of as black boxes that received
inputs, processed them, and produced output (McGrath 1964). Group performance
research focused on what characteristics of team members and team processes might
encourage or hinder efficiency, and group success was thought to be indicated by this
efficiency (Ancona 1990; Ancona and Caldwell 1992). However, more recently,
researchers have focused less on the how of team process and more on why some
teams perform at high levels while others don’t (Ilgen et al. 2005). This literature has
moved beyond the borders of the team and has focused on the complex interplay
between the team and its environment (Burke et al. 2006; Kozlowski et al. 1999;
McGrath et al. 2000).
Team adaptation refers to the process through which teams develop a shared
vision of their goals, enact plans, norms, processes and procedures to achieve those
goals, and proceed to execute those plans to achieve the goals. As the team executes
the plans, it receives feedback from the environment and, if a “salient cue” is identified,
the team adapts to the cue, and re-formulates its plan in reaction to the cue (Burke et al.
2006). “Team adaptation is manifested in the innovation of new or modification of
existing structures, capacities, and/or behavioral or cognitive goal-directed actions”
(Burke et al. 2006). Adaptive teams are indicated by their ability to sense barriers to

45

performance, and to modify their practices and goals to effective manage these barriers.
This adaptation can take the form of barrier removal or barrier avoidance. (Tesluk and
Mathieu 1999). Whether emerging from organization, environmental or technical forces,
adaptive teams develop the ability to first sense the barrier (or potential barrier), and
then take action to remove the barrier’s impact on team performance. In method
environments, these barriers may stem from the policies of the customer, the impact of
external system constraints, such as enterprise architecture requirements, or the ability
for the technology platform to accommodate changes.
Teams are required to adapt to non-routine situations. Agile development views
software development as unpredictable, non-routine and novel. When responding to
novel conditions, teams whose members can recognize that the environment had
changed were shown to be more adaptable (Waller 1999). Additionally, teams that
interrupted their processes to “stop and think” about the process and its performance
were shown to be able to recognize the need to respond to environmental change more
quickly (Okhuysen and Waller 2002).
Team adaptability theory proposes several team processes and emergent states
that impact their adaptation capability. Burke et al. (2006) propose an integrated
process model of adaptation that bears strong resemblance to the agile methodologies
in practice. They describe an iterative, feedback intensive learning process
characterized by receiving repeated cues from the environment, ascribing meaning to
and learning from those cues, and adapting team behavior based upon the new
learning.

46

Team adaptation is a cyclical process. At each stage of the cycle, teams either
receive or react to cues from the environment. These stages lead to the creation of, and
modification of shared mental models. Teams build situational awareness and a plan to
execute the team tasks in Phase 2. As the plan is executed, the teams coordinate their
activities via mutual monitoring, and coordination, which lead to learning and feedback.
Agile development requires the ability to sense and respond to change (Conboy 2009;
Lyytinen and Rose 2006) Therefore agile development methodologies propose
practices that map closely to the constructs developed by Burke et al. (2006). These
characteristics are described below.
Cue Recognition
The presence of cues is not equivalent to recognition and action due to those
cues. Gersick & Hackman (Gersick and Hackman 1990) found that when the responses
to pre-flight checklists became habituated, even the presence of cues that would require
a modification to the pre-flight checklist procedure did not cause a change in action. In
this case, the flight crew (in other words the team) had become so familiar with the
responses acceptable in one situation (flying in warm, southern climates), that when the
situation changed (taking off in an ice storm), the crew could not recognize the need for
the process to change. So, the fact that environmental cues exist does not indicate the
ability for the team to assess that cue as being relevant to taking action.
A team must be able to identify which cues are triggers for adaptation. Louis and
Sutton (Louis and Sutton 1991) describe the cognitive “shifting of gears” which is
required to pull thinking from the automatic mode to the conscious mode. They identify
several situations that can cause teams to recognize cues in either a planned or

47

unplanned manner. When teams experience unusual or novel events, an incongruity
between expectation and experience, or the specific initiative to draw attention to cues,
and opportunity for cue identification arises. Okhuysen (2001) found that groups who
were given a mechanism for formally diagnosing process discrepancies generated more
work interruptions were able to adapt
Since Agile methodologies are based upon the philosophy of delivering in small
pieces and in short cycles called iterations, they provide the opportunity for repeatedly
receiving cues from users as to whether the shared mental model begin developed
regarding the problem and the solution is correct. Further, because planning occurs at
the iteration level, external environmental changes can be detected and incorporated
into the plan for the next iteration.
Cues are not limited to information crossing the team boundary, but can also be
associated with temporal structures within a team’s execution plan. Gersick (1988)
found that teams reevaluate and revise their execution plan based upon the cue of the
project schedule mid-point. Additionally, she found that the mid-point of a project acted
as a catalyst for adaptation. Interestingly, this study showed that the length of time
between the deadlines did not matter; rather the mid-point of the schedule was the
catalyst for change, not the duration of time passing. This suggests that by taking
advantage of this seemingly built-in trigger for adaptation moves, teams who plan
shorter duration milestone deadlines should receive more cues to adapt than those who
schedule longer duration milestones.

48

Team Situational Awareness
“Team situation awareness refers to a shared understanding of the current
situation at a given point in time” (Salas et al. 1995) Situational awareness is the
product of interaction of individual situational awareness of the team members, and the
degree of the team members’ shared understanding of the environment and task, in
other words, the team’s shared mental models (Salas et al. 2001). Endsley’s model of
situational awareness (Endsley 1995) describes the existence of three levels of
increasing complexity of situational awareness. Level one awareness describes the
ability to simply perceive cues from the environment. Level two awareness exists when
a team is able to comprehend the relevance of its environmental cues, as to their
relationship to the current task. Finally, by developing the capacity to evaluate the
relevant environmental cues and extrapolate likely outcomes of the current team task,
based upon the new cues, a team attains level three awareness.
Developing and maintaining team situational awareness is a focus of several
agile practices. As will be discussed below, Agile development methodologies stress the
importance of two situational awareness-building processes. First, the team receives
feedback about the fit of the output of the team task (the system) each iteration. Based
upon this feedback (cue), the team evaluates both the output of the process and the
process itself. This process provides the team the ability to receive and process cues,
and to evaluate and forecast performance. Second, because most agile methodologies
call for the establishment of a daily short review meeting, situational changes are
disseminated to the team repeatedly throughout the iteration itself, allowing for the

49

perception and reaction to cues more often, and more quickly than other methodologies
that call for longer periodic status update reporting.
Mutual Monitoring
Mutual performance monitoring is a process through which team members
observe the actions of their teammates, and attempt to identify mistakes and other
performance issues. By making this identification, teams can expeditiously correct
performance issues. Mutual performance monitoring implies that team members still
can complete their own work while keeping track of the work of other team members
and ensuring compliance with existing team procedures. (McIntyre and Salas 1995)
Mutual performance monitoring contributes to a team’s ability to adaptively
execute a plan in several ways. Marks & Panzer (2004) found that mutual performance
monitoring enables the recognition of when team members need assistance, increases
coordination within the team, and increases team performance. Mutual performance
monitoring provides intra-team adaptive cues, in that it provides a mechanism to alert
the team that the team process may be off track (Dickinson and McIntyre 1997).
Additionally, mutual performance monitoring helps to build team situational awareness,
leading to a greater ability to sense environmental cues (Salas et al. 1995).
Agile development methodologies prescribe mutual monitoring via several
practices, including pair programming, chief programmer reviews, code standards and
daily stand up meetings.
The process of team adaptation described above is an organizational learning
process. As teams build shared mental models in order to understand their tasks,
processes, and goals. Based upon this shared understanding, the teams build a plan of

50

action, and execute the plan. They team establishes the plan based upon their
theorized understanding of the goals of the task.
As teams act on their expectations of outcomes, the learning process occurs
when there is incongruity between the goals of the team, and the perceived outcomes
(Argyris and Schoen 1978). However, there are several sources of potential incongruity
between the established goals of the team, and the delivered output of the process.
First, the team may have developed an inaccurate view of the goal based upon
misunderstanding of cues from the environment. Second, the team’s established plan
may not deliver the outputs it was expected to. Third, it is possible that the team
delivered the output desired, accurately representing the previous cues from the
environment, and yet the output is not acceptable, either due to changes in the
environment during the execution process, or due to the fact that the source of the cue
was incorrect in establishing the team’s goal. In short, learning occurs when signals are
received that are incongruent with the team’s expectations (Pich et al. 2002).
However, for learning to occur within the team, “learning agents’ discoveries,
inventions and evaluations must be embedded in organizational memory. They must be
encoded in the individual images and the shared maps of the organization theory-in-use
from which individuals will subsequently act. If this encoding does not occur, individuals
will have learned but the organization will not have done so” (Argyris and Schoen 1978).
Adaptive teams reform their shared mental models based upon the detection of
incongruity (via situational awareness and cue sensitivity). After detection and
reformation, adaptive teams encode their learning into the theory-in-use during each
adaptive cycle.

51

Software Development As New Product Development
In the early history of building software, software creation efforts were often
characterized as software engineering. As such, early software development
methodologies bear similarity to plan-based engineering project methodologies (Boehm
1988). However, agile practitioners argue that the process of developing software bears
a much greater resemblance to new product development efforts (Highsmith 2002). A
great deal of literature exists that describes the issues of product development under
uncertainty. This literature presents several congruent perspectives about the impact of
uncertainty on planning and new product development.
MacCormack & Verganti (2003) investigated 29 internet companies and the
impacts of two specific types of uncertainty – platform uncertainty and market
uncertainty – on the impacts of the choice of development process on project
performance. The authors suggest that various types of uncertainty and other
contingencies might lead to different responses by the team in the design of their
development process. They look at the concept of flexibility of development process as
being a core response to uncertainty.
Bhattacharya, Krishnan & Mahajan (1998) defined a concept of “Real-time
definition” of product requirements. They found that in new product development,
contingencies drive trade-offs between early and late product specification locking. In
highly uncertain environments, either due to marketing uncertainty or the lack of full
understanding of customer’s requirements, iterative design with multiple rounds of
feedback was found to lead to higher performance.

52

Agile development teams are intended to rapidly respond with new solutions.
Eisenhardt and Tabrizi (1995) compared two competing strategies for making rapid
innovations, the compression strategy and the experiential strategy. The assumptions of
these two strategies are analogues to the assumptions of plan-based and agile
development processes. Compression based strategies are based upon the assumption
that the product development process can be defined with a series of pre-defined steps.
Because the steps are definable, the compression strategy states that the steps can be
performed in an overlapping fashion, by shortening the steps themselves, and by
rewarding developers for schedule completion.
On the other hand, experiential strategies are based on the assumption that the
product development process is marked by uncertainty in the environment; both
requirements-based uncertainty and technology-based uncertainty. They note, “the key
to fast product development is, then, rapidly building intuition and flexible options in
order to learn quickly about and shift with uncertain environments. At the same time, it is
also important to create structure and motivate pace in these settings” (p 91). This
suggests that in uncertain environments, building situational awareness and a shared
team mental model is a key to both team adaptation in general, and to the speed of task
completion.
In experiential environments, the authors found that the number of iterations
increased development speed through shared understanding. Additionally they find that
product development team is shortened due to early feedback (via testing), due to the
fact that it gives developers early feedback regarding failures. Finally, they find that the
reduction in time between milestones reduces project development time. Gersick (1988)

53

found that teams reorganize task efforts and develop urgency to complete assigned
tasks at the midpoint of the task schedule, regardless of the length of time schedules.
Eisenhardt & Tabrizi (1995) found that, in general, projects that were more
definable could be delivered more quickly using a compression strategy, and that they
were not able to be delivered more quickly via an experiential strategy. For uncertain
environments, they found the reverse to be true. Importantly, their analysis results
showed that when each strategy was applied to each
Similarly MacCormack & Verganti (2003) found that early feedback and
adaptation that continued throughout the length of the project was important to project
performance. They found that under uncertainty significant differences in investment in
architectural design expenditures, early technical feedback and late changes to
requirements all predicted performance. MacCormack (2001) also found that rapid and
early feedback positively impacts product quality and performance.
As the discussion above illustrates, under conditions of uncertainty, the new
product development literature illustrates the impact on project outcomes of greater
depth and frequency of feedback.
Adaptability, Agile and Feedback Capabilities
The theories and literature presented above share significant commonality in
their definition of necessary conditions for develop team adaptation capabilities.
According to Team Adaptability Theory, Organizational Learning Theory, and the new
product development literature, the central capability necessary to respond to
uncertainty is the ability to obtain feedback.

54

It was noted above that agile methodologies’ internal practices are quite diverse
and have diverse focuses. However, what is universally accepted in agile
methodologies is the need for continuous feedback. Every agile method except
Lean/Kanban argues for an iterative delivery cycle. Lean/Kanban takes the iterative
cycle to the extreme, considering each development action as an iteration.
The iterative development cycle is an excellent illustration of team adaptability
theory in action. At the beginning of each iteration, the team and stakeholders assess
the current situation, including processing feedback from the previous iteration. They
evaluate the current needs of the organization, and develop a shared mental model for
the environmental status, organizational needs, and the deliverables for the coming
iteration. Next, they formulate the plan to deliver the features of the iteration. After this,
they execute the iteration, process cues and coordinate while delivering. Finally, they
process the results of the iteration and determine necessary modifications to process for
the next delivery cycle.
By performing the entire product development lifecycle for the given
requirements during every iteration, agile methodologies “freeze” the requirements at
the beginning of every cycle. This makes the environment and task conceptually
definable, and the team can take advantage of the compression based speed
enhancement mechanisms. However, at the end of each iteration, the agile
methodologies dictate the use of several feedback mechanisms, including system
reviews with the stakeholders and users, and retrospective analyses of the iteration
processes. This rapid feedback loop provides the three characteristics of the

55

experiential process. Iterations are short, are of high number and constitute the
repeated tests of the product necessary to guide its speedy development.
In a sense, the agile development team treats the macro process (the project) as
an empirical process, and the micro process (the iteration) as definable. Because
building a shared understanding and a theory in practice is necessary for action, the
agile development team must believe that what they are to act upon immediately is
knowable, even if what is to be defined later is not.
This discussion suggests that rapid, early feedback should impact agile
performance, due to the experiential nature of agile project management. Additionally,
the research of Gersick (1988) suggests that shorter, iterative development milestones
should result in more rapid response to change, and faster delivery due to the
motivation of pace (Eisenhardt and Tabrizi 1995).
Project Success
A common way to contrast agile development with traditional project
methodologies is to label projects as “agile” vs. “plan based”. Traditional project
management is based upon the assumption that accurate plans can be created for the
entire project, up front. Because of this, project management performance metrics are
usually based on the conformance of team process and outcomes to the plan. However,
in the traditional project management field, measuring project performance by the “Iron
Triangle” (Scope, Time & Budget) has begun to fall out of favor. Rather, a broader
definition of project success has been proposed (Atkinson 1999; Phillips et al. 2002).
These calls for broader definitions of project success call for the evaluation of project
process and outcomes across four categories: (1) Project management metrics, (2) the

56

quality of the project output, (3) individual benefits, and (4) organizational benefits
(Atkinson 1999).
Project Management Metrics
The “Iron Triangle” of project management metrics, Cost, Quality & Scope, has
been the traditional measure of project “success”. Consistent with this tradition is the
CHAOS report from the Standish Group, one of the most widely disseminated sources
of IT project result data, which utilizes only the Iron Triangle metrics to define project
success. Standish Group places projects into three categories, Success, Challenged,
and Failed, based upon these criteria:
“A project is successful if it is completed on time, within budget,
and with all features and functions as initially specified. A project is
challenged in case it is completed and operational but over budget, over
the initial time forecast, and if it offers fewer features and functions than
originally specified. A project is impaired or failed if it is canceled at some
point during the development cycle” (Eveleens and Verhoef 2009)
Based upon these criteria, the CHAOS report has shown that project success is
different when measured in project management metrics. Table 2.6 shows the Chaos
categorized results for selected years of the report.
However, when one adopts the view of IT projects as empirical processes, it is
predictable that a report that relies wholly on project management metrics would show
significant project failure rates. Project Management metrics are based upon the ability
to forecast accurately, at the beginning of the project, what the required development
effort is, how much time that will take, and as a function of those factors, how much will
the project cost. In short, the Chaos report measures the success of the project forecast
to a much greater extent than it does the success of the project.

57

Eveleens and Verhoef (2009) found that organizational politics played a
substantial role in their forecasting procedures, and that project management metrics
are easy to manipulate. They found that an organization that had adopted the Standish
definitions of project success produced significantly better results than the Standish
report itself, but upon further review they found that this organization’s forecasts were
based upon the worst-case scenarios. Further, another organization that they reviewed
scored very poorly based upon the Standish definition of success, yet was found to be
the most accurate with initial forecasts.
Table 2.6. Standish CHAOS results
Year
Success Challenged
Failure
1994

16%

53%

31%

1996

27%

33%

40%

1998

26%

46%

28%

2000

28%

49%

23%

2004

29%

53%

18%

Even so, project management forecasting quality remains important, as it plays a
significant part in the decision as to which projects are to be undertaken, even in agile
environments. However, because agile claims of impacts include assertions about
higher quality, higher satisfaction, and additional organizational benefits, this study is
also concerned with additional, alternative measures of project success.
Definitions of IS Success
Delone & McLean (1992) defined a model of IS success with the dimensions of
system quality, information quality, use, user satisfaction, individual impact, and
58

organizational impact. This Delone & McLean model has been tested numerous times
(Delone and McLean 2003; Petter et al. 2008; Rai et al. 2002), and has been found to
have good explanatory power (Rai et al. 2002). Seddon (1997) proposed a modified
model which removes IS Use from the causal nomological network, and Rai et al.
(2002) showed that use occurred independent of user satisfaction, due to the fact that
IS systems often become the only source for specific information, and access to that
information requires use. However, Rai et al (2002) found that both the Delone &
McLean and Seddon models showed significant explanatory power.
Both Rai et al. (2002) and Petter et al. (2008) state that the evidence regarding
IS Success models validates the viewpoint that an integrated multi-construct measure of
IS success is appropriate. Further, this measurement of IS success should measure
beliefs, attitudes and behaviors and their interdependencies.
Thus project success in this study is conceived as a combination of project
management success and the success of the IS project output in terms of system
quality, customer satisfaction, and perceived benefits.
Contingencies In the Project Environment
Contingency theory was originally developed as a way to explain the relationship
between organizational structure and performance. Contingency theory proposes that
there is a fit between organizational structure and various external contingencies, and
that the level of fit between the two is a driver of performance (Donaldson 2001;
Lawrence and Lorsch 1967). Two core environmental contingencies that have been
hypothesized as requiring adjustments for fit are environmental uncertainty and
technology.

59

Burns and Stalker (1994) contrasted mechanistic structures, which were typified
by hierarchical structures, and where task related knowledge and decision-making is
centralized. In contrast, an “organic” structure is characterized by a network of
empowered employees, who build a shared understanding of the task, and accept
decentralized responsibility for delivery. They argue that high rates of change in
technology or market environments require the use of organic structure.
Early versions of contingency theory focused on the impacts of contingencies on
organizational structure and corporate strategy (Venkatraman 1989), but the theory has
been applied to the contingent effects of new product design and software development,
in which the impact of contingent factors on the performance of product development
process performance is well documented (Barki et al. 2001). The theoretical model of
this dissertation adopts the view that the development process chosen by a team may fit
more or less well with the environment. While agile methodologies may be used in any
development situation, it is proposed that because of its stress on multiple levels of
feedback, and the repeated acquisition of feedback, agile method use impacts may be
seen most in environments that are more uncertain. Further, uncertainty as provides
key contingent factors that moderate the impacts of agile method use on project
success.
The Components of Uncertainty as Contingent Factors
Projects are undertaken to deliver a unique product or service (PMBOK 2000).
The practice of project management establishes specific processes to plan, execute,
monitor and control projects, based upon numerous contingent factors. Major contingent
factors in projects include the proposed size or scope of the project, the availability of

60

resources, financial, human, and material, and the time available to complete the
project.
Based upon these factors, projects are more or less complex to manage. The
appropriateness of any software method will be judged on its ability to complete the
project to which it has been applied. As described above, most projects are challenged
or fail, according to traditional project management metrics. As such, traditional
software methodologies may be considered a poor fit for the projects to which they have
been applied. However, several particular project & product characteristics have been
proposed as being specifically important when choosing software methodologies.
“Home grounds” have been proposed for agile and plan-based approaches
(Boehm and Turner 2004). Boehm & Turner contrast the primary goals of agile vs. plan
based approaches, and state that there are specific project, product and environmental
characteristics that indicate more appropriate fit for the use of one type of method vs.
the other. Their five factors are project size, project criticality, environmental dynamism,
personnel, and culture and are listed here as Table 2.7. This table provides a list of the
kinds of contingencies that may affect the fit between the environment and development
method.
Size: As a project becomes larger, due to greater scope, team size, or both,
communication becomes more difficult. Methodologies that rely on building a shared
mental model of a problem and its associated solution will face obstacles as the size of
the problem to be modeled becomes larger.
Criticality: Boehm and Turner (2004) claim that as a system’s criticality rises,
plan-driven approaches are a better fit. This is likely caused by the growing objective

61

nature of requirements on life-critical systems vs. process-automation systems of lower
criticality (see Table 2.7).
Dynamism: In highly dynamic environments, up front planning does not provide
opportunities to learn from feedback, to make sense of the environment, and to
generate high hit potential (MacCormack and Verganti 2003).
Table 2.7. Five Factors Determining Fit of Agile vs. Plan-Based Methodologies
Factor
Agility Discriminators
Plan-Driven Discriminators
Size
Well-matched to small
Methodologies evolved to handle
products and teams.
large products and teams. Hard to
Reliance on tacit knowledge
tailor down to small projects.
limit scalability
Criticality
Untested on safety-critical
Methodologies evolved to handle
products. Potential difficulties highly critical products. Hard to
with simple design and lack
tailor down to low-criticality
of documentation.
products.
Dynamism
Simple design and
Detailed plans and Big Design up
continuous refactoring are
Front excellent for highly stable
excellent for highly dynamic
environment, but a source of
environments, but a source
expensive rework for highly
of potentially expensive
dynamic environments.
rework for highly stable
environments.
Needs a critical mass of scarce
Personnel
Requires continuous
presence of a critical mass of expert resources during project
definition, but can work with fewer
expert resources.
later in the project – unless the
environment is highly dynamic.
Culture
Thrives in a culture where
Thrives in a culture where people
people feel comfortable and
feel comfortable and empowered
empowered by having many
by having their roles clearly
degrees of freedom (Thriving defined by clear policies and
on chaos)
procedures. (Thriving on order)

Personnel: Alistair Cockburn’s (2001) “levels” of developer skills define three
skill categories of developer. Level 1 developers can only, with training and experience,
act in well defined situations. Level 2 & Level 3 developers have the understanding of
methodologies, practices, and corresponding rules to either modify or break the rules to
62

align the method with emergent, unexpected and unprecedented situations. As agile
methodologies are intended to be flexible and reactive to unexpected situations, Boehm
& Turner propose that development methodologies must retain a higher level of
resource throughout a longer portion of the project than plan-based methodologies.
Culture: Finally, the adoption of a cultural philosophy that values the flexibility
and potential value of allowing change is key to successful adaptive product
development (McAvoy and Butler 2009).
The Research Literature on the Impacts of Agile development
The research literature on the impacts of agile development is still relatively new.
Even as web searches find literally hundreds of presentations and papers presented at
conferences, as of 2008, only 35 empirical studies of rigor had been published, and of
those, only 6 had appeared in journals (Dyba and Dingsoyr 2008). Further, a number of
those studies that had been completed were descriptive, focused on why organizations
might adopt agile development, or compared agile to traditional development
processes. Since then, several special issues have focused on agile development.
However, even as research on agile development has continued to grow, the majority of
the published work does not address empirical observations of agile development
impacts.
Table 2.8 lists a selection of recent scholarly literature on the impacts of agile
development. There has been a great deal of literature that has focused on agile
development practices, a number of pieces that have applied the lens of agile to certain
aspects of organizations, and a few articles that have attempted to tie agile
development into the previous literature on control and organizational learning.

63

However, to my knowledge, relatively few examples exist of research that attempts to
create a holistic theory of agile development adoption and impacts.
Maruping et al. (Maruping et al. 2009a; Maruping et al. 2009b) produced two
recent articles that integrated agile method with the control and expertise coordination
literature. They measured XP practices including pair programming, collective code
ownership, refactoring, and continuous integration, and hypothesized that these
practices would increase software quality, as indicated by the number of software bugs
identified. They identified two contingencies that inform the theoretical model adopted .
First, they treated requirements change as a moderator, in the presence of which
the impacts of agile method use increase. Second, they investigate two control
phenomena, outcome control and self control. They found that there is a significant
interaction in how these factors affect agile development impacts. Outcome control
increases the impact of agile methodologies when requirements volatility is high, but not
when it is low. This is intuitive, based upon the discussion of team adaptability above.
Outcome control is more appropriate in highly volatile environments since outcome
controls give teams a general goal toward which to work, while providing the team with
the authority and autonomy to determine the best way to achieve the goal. When a
process output is not novel, or can be well defined in advance, the steps necessary to
complete the process are likely to be definable, leading to the likelihood of the
applicability of behavioral control.
Maruping et al. (2009a) found that the construct of output control moderated the
impact of agile method use on a set of software quality measures. However, their
measurement of output control involved relatively vague outcome indicators such as

64

“Significant weight will be placed upon timely completion”, and “Project goals were
outlined at the beginning of the project”. As output control specifies the need for wellunderstood and clear indicators of task outputs, it is unclear if these indicators
measured outcome control.
Maruping et al. (2009b) measure the use of 2 XP practices, collective code
ownership and coding standards on software quality, as measured by number of bugs.
They find that these relationships are significant, and additionally that the impacts of
expertise coordination are lowered on teams that use these practices.
These two studies are constrained by their narrow conceptualization of project
success, and their treatment of agile method use as indicated solely by the level of
adoption of particular XP practices. Additionally, the teams being measured had
implemented a wider set of practices than were included in the model in Maruping
2009b. Further the practices are proposed to have heavy interaction effects when used
together. Therefore the impact of the particular processes on the dependent variable is
difficult to extract from the impact of the use of XP in general.
Harris et al. (2009) used case study research to extend the control literature with
a modification to the concept of outcome control. As described previously, they
recognized that outcome control occurs at the end of a project. Since it is based upon
criteria developed at the beginning of the project, it is not aligned with agile
development philosophy. They illustrate that turbulent environments require the
application of flexible capabilities, and that a priori understanding of both the goal of a
project, and the evaluation of project results is made more difficult based on uncertainty.

65

As they explained, output control is not effective in an iterative and emergent
development context because it is seen as an evaluation measure of an entire process
or project, is performed one time, at the completion of task or process performance, and
the evaluation is performed based upon an a priori definition of process requirements.
Table 2.7 compares output and emergent outcome control.

66

Study

Salo and
Abrahamsson
(Salo and
Abrahamsson
2008)
Mann (Mann
and Maurer
2005)

Maruping, et
al. (2009a)

Table 2.8. Recent Empirical Literature on Agile Methodologies
Description & Research
Key
Agile
Agile
Method
Constructs
Method
Practices
Tested (if Measured
specified)
Descriptive survey of
N/A
XP and
XP (12)
European XP and Scrum.
Scrum
Scrum (5)
Survey, 35 projects over 18
organizations
Studied the impacts of
N/A
deploying Scrum in a single
organization.
Case study – single
organization over two
years.
Studied the impact of agile
practices on teams’ ability
to respond to requirements
change.
Longitudinal Survey of 862
members of software
development teams within
a large U.S. Consulting
organization

Contributions

Found that agile
methodologies are
generally perceived by
the development team
as helpful.

Scrum

XP
Agile Method
Use
hypothesized to
positively impact
Software Project
Quality
Moderators:
Requirements
Change
Outcome
Control
Self Control

67

Pair
Programming
Refactoring
Coding
Standards
Collective
Code
Ownership
Continuous
Integration

Found that agile method
use is most important in
improving project quality
when outcome control
and requirements
change are high.

Study

Description & Research
Method

Harris, et al.
(2009)

Studied the impact of XP in
turbulent environments.
Introduced the concept of
emergent outcome control.

Three case studies.
Austin & Devin Theory of agile as post(2009)
industrial production.
Inductive Argument

Table 2.8. (cont’d)
Key
Agile
Constructs
Method
Tested (if
specified)
Emergent
XP
Outcome
Control

Post-Industrial
Production is
theorized to be
possible due to
the fact that
technology now
exists that
allows the cost
of novelty to be
reduced to a
level below the
benefits of
novelty

68

N/A

Agile
Practices
Measured

Contributions

None

The concept of
emergent outcome
control as an extension
of control theory.

None

Made a theoretical
argument that enabling
technologies might allow
for novel development to
be inexpensive enough
to be feasible.

Study

Description & Research
Method

Lee & Xia
(2010)

Conceptualized “agile” as
team response
extensiveness and
efficiency.
Focused on the idea that
software team autonomy &
diversity led to greater
response to change.

Fruhling & De
Vreede
(Fruhling and
Vreede 2006)

Survey of 565 project
managers
Studied XP in an
environment building an
emergency response
application.
Used action research to
view the 12 XP principles in
practice, with the intention
of operationalizing the
practices.

Table 2.8. (cont’d)
Key
Agile
Constructs
Method
Tested (if
specified)
On-Time
N/A
Completion
On-Budget
Did not
Completion
measure
Software
any
Functionality
particular
agile
practices
or culture.

Utilized the four
agile values as
a lens to
describe the
level of agile
adoption.

69

XP

Agile
Practices
Measured

Contributions

None

Re-conceptualized Agile
as the level of autonomy
enabling response to
changes.
Team diversity is not
mentioned by agile
alliance. Team
autonomy may be
interpreted as Principle 5

All XP
practices

Descriptive study Illustrated the use of XP
in practice, how they
could be
operationalized.

Study

Description & Research
Method

Lytinnen &
Rose, 2006

Viewed agility via an
organizational learning,
exportation/exploitation
lens.
Longitudinal Case Study

Maruping, et
al. (Maruping
et al. 2009b)

Studied the impacts of two
XP practices (collective
code ownership and coding
standards) on software
technical quality.
Survey of 56 project teams
consisting of 509 software
developers.

Table 2.8. (cont’d)
Key Constructs Agile
Method
Tested (if
specified)
Type 0
N/A
innovation (by
others)
Type 1
innovation (new
ways of building
IS)
Type 2
innovations(
innovative new
IS)
Outcome
Variables:
Innovative
Content, Speed
Cost
Expertise
XP
Coordination
Collective Code
Ownership
Coding
Standards
Software
Technical Quality
(lack of bugs)

70

Agile
Practices
Measured

Contributions

None

Gave agility a more
multi-faceted definition.
Described the
importance of adoption
of innovative
technologies as an
enabler of agile
innovation.

Collective
Code
Ownership
Coding
Standards

Found that collective
code ownership and
coding standards
moderated the impact of
expertise coordination.

Study

Description & Research
Method

Cao, et al.
(Cao et al.
2009)

Looked at the adoption of
agile through the lens of
adaptive structuration.
Found that structure
emerged that both enabled
and constrained the
adoption of agile practices.
Case Study.

Mangalaraj, et
al.
(Mangalaraj et
al. 2009)

Port, et al
(Port and Bui
2009)

Using the lens of software
process innovation
diffusion, looked at factors
that impacted the adoption
of agile practices.
Case Study.
Using simulation, the
authors test the ability of
plan-based methodologies
and agile methodologies to
respond to requirements
change.

Table 2.8. (cont’d)
Key Constructs Agile
Method
Tested (if
specified)
Challenges to
XP
the adoption of
agile
Customer
Challenges
Developer
Challenges
Management
Challenges
Acceptance of
XP Practice

Requirements
volatility
Requirements
prioritization
process (agile or
plan-based)

71

Agile
Practices
Measured

Contributions

All XP
practices

Illustrated the impacts of
social structure on
success of agile
projects.

XP

All XP
practices

Found individual, team,
technological, task, and
environmental factors to
influence the acceptance
of XP practices in an
organization.

None
(General
concept of
prioritizati
on)

Not specified

They found that,
consistent with Boehm &
Turner (2004), that
requiremetns volatility
was a key driver of
success. When volatility
was high, agile
methodologies
performed better than
plan based, and viceversa.

The key differences between traditional output control and emergent outcome
control are shown in Table 2.9. Emergent outcome control is executed repeatedly
throughout the project, is used as a method for correcting the trajectory of the project
deliverables, and the “definition of done” (in scrum terminology) evolves and emerges
during the project.
Austin & Devin (2009) describe the cost of novelty in “post-industrial” making.
They contrast the making processes of pre-industrial vs. industrial making, and illustrate
how the pre-industrial (craft) making process allowed for the delivery of high levels of
uniqueness, or novelty, but at a high cost. Contrasting with this, industrial making is best
at producing low-cost products of low variance. They make the assertion that novel
outputs will only be sought when the cost of producing them is lower than the benefits.

Output

Table 2.9. Output control vs. emergent outcome control (Harris, 2009)
Purpose
Frequency
Evaluator
Construction Comparison
of Standard
Evaluation
Once
Supervisor
A priori
Completed
project
versus
specification

Emergent
Outcome

Corrective
Action

Continuous

Multiple
Evolving by Emergent
stakeholders stakeholder coutcomes
versus tacit
specifications

However, they make the claim that technology may enable the emergence of
post-industrial making, just as it enabled the production processes of the industrial
revolution. Because digital technologies can reduce the reconfiguration and exploration
costs of software development, they propose that enabling technologies may drive the
cost curve of novelty lower, allowing for the more frequent delivery of novel outputs.

72

Although they make this claim, they do not develop it, nor do they provide illustrations of
which technologies might lower this cost curve. However, as described above, agile
development practitioners make several claims about the ways that technologies enable
refactoring and other processes that are central to agile development.
Cao et al. (2009) and Mangalaraj et al. (2009) both investigate impacts of the
environment on agile development adoption and impacts. Cao et al. adopt the lens of
adaptive structuration theory to describe the effects of XP implementations, and the
social structuration that took place. They found that there were significant structural
barriers to the use of agile methodologies like XP. For example, they found that upper
management was unwilling to proceed with development without a strong up front
estimate of the project cost. This led to the team not adopting the planning game
practice, and instead adopting a plan-based estimation process. At the same time,
management did agree that based upon the experience during a pilot XP project, that
cost estimation was likely to change during the project, and that cost estimate changes
would not be used as an indication of project failure.
Mangalaraj et al. (2009) used the lens of innovation diffusion to investigate the
acceptance and adoption of various XP practices. Within a single organization that was
adopting XP, they observed two projects, one a new development effort, and one an
ongoing development team for an existing system. They observed that the two teams (X
& Y) had significantly different reactions to the decision to adopt XP. Team X exhibited
significant adoption of XP, while team Y resisted the implementation. They identified five
categories of barriers to and enablers of adoption of new development practices,
individual, team, technology, task, and environmental. While this study investigated

73

agile development innovation adoptions, the contingent impact of factors such as these
on project outcomes can be interpolated.
At the individual level, they found that team X’s Knowledge of XP processes gave
Team X an understanding of the purpose and interplay of the XP processes. Team Y,
even though they were trained on XP in the same fashion as team X did not retain
knowledge of XP practices, how to perform them, or what purpose they served. Team X
had a very positive attitude toward XP, while team Y had a very negative attitude. At the
team level, Team X adopted an XP egalitarian and empowered perspective, while Team
Y maintained their previous, hierarchical roles.
Consistent with Austin & Devin’s (2009) proposition, and the practitioner literature
on agile development, Mangalaraj et al. (2009) found a significant impact of technology
factors on both agile development adoption and impacts. These fell into two main
categories, compatibility, and tools support.
Compatibility - Team X utilized Java as the sole language for development. This
single technology made utilizing collective code ownership possible, and contributed to
velocity, while Team Y was constrained by the legacy technology that hindered their
ability to collectively manage the code. Rather, the team members were more or less
expert in their use of the various legacy coding platforms and tools, and ownership &
specialization developed because of this. Additionally, the ability to pair program was
constrained by the same issue.
Tools Support – Team X utilized tools that supported continuous integration,
refactoring and test-first programming, while Team Y could not, due to their legacy
technology. “These tools greatly helped members of project X in adhering to XP

74

practices. This notion was echoed by several members of project X. According to one
member, ‘[tool] is a big factor in being successful. It greatly helps in refactoring the code
quickly and safely.’ Another member said, ‘Tool support is definitely important and it
varies according to the language. Without [tool]-like tool it may be hard to do
refactoring.’ In contrast, the tools and Integrated Development Environments employed
in project Y offered little or no support for core XP practices. Because of this, members
of project Y faced great difficulty in implementing some of the practices and had to
impro- vise. According to a team member, ‘Test first in C++ platform is difficult to
implement; for Java they have [tool] and here we tried it but we could not’” (Mangalaraj
et al., 2009 p 350).
Task level factors also emerged in this study. Team X built a green-field system,
consisting of 30,000 lines of code in less than 16 months. Team Y on the other hand
had two significant constraining task factors. First, they had a large legacy code base to
support. This code base was built on three disparate technologies, had no automated
regression test code, and had been developed over ten years. While Team X could
make refactoring moves at will, Team Y could not, for fear of breaking the system.
Additionally, while Team X’s system required little integration with external systems,
Team Y’s system required extensive integration, limiting its ability to act in an agile
fashion.
Finally, the authors identified three environmental factors, budget constraints,
time constraints and customer participation. The team’s customers and management
had significantly different methodologies of control. While team X’s management seems
to have used primarily output control, Team Y’s external control was performed by

75

“demanding” product managers, who “dictated” their time and budget constraints.
Because of this, pairing and refactoring (the modification of code to make it of higher
quality, while leaving functionality unchanged) were viewed as wasteful.
Lee & Xia’s (2010) study views the impacts of team agility as indicated by team
response efficiency and extensiveness (Lee and Xia 2010). They conceptualized the
use of agile method as the level of autonomy granted the team. This conceptualization
was used as an independent variable to test its impact on team’s response to changing
requirements. They found that team autonomy led to the ability to respond effectively to
changing requirements, which increased project performance.
Finally, Port & Bui (2009) used simulation to test the agile vs. plan-based
processes for requirements prioritization. They found that agile development
requirements prioritization processes outperform plan-based processes in situations of
high requirements volatility.
Agile Methodologies: Structure and Impacts
Based upon the discussion of the principles of the agile manifesto, the
importance of feedback, the discrepancies between agile methodologies, and the
previous research described above, it is clear that agile methodologies are a complex
phenomenon. While much of the previous literature has focused method-specific
practices, it is encouraging to see that more broad based constructs are emerging.
However, it is believed that, based upon the previous literature and prior theory, a case
can be made for opening the black box of agile to a degree, and theorizing regarding
the relationships between agile constructs and the nature of their impacts.

76

Figure 2.5 illustrates a theoretical perspective that integrates the prior literature
into a holistic model of agile method use and impacts. As can be seen, the constructs
method
that are likely to directly impact project success are separated from those likely to
enable the direct effects.

Figure 2.5. The Nomological Network of Agile Method Adoption and Use
Agile Philosophy Adoption
option
Because all methodologies are motivated by an underlying philosophy, adoption
of an agile philosophy by the delivery organization will motivate the use of agile
processes and practices, and the adoption of supporting technologies. However,
philosophy adoption is not theorized to directly impact project performance. Rather, the
philosophy motivates a congruent set of practices and processes that are hypothesized
by the organization to reinforce each other toward a desired goal. In agile development,
.
77

this goal is to develop a dynamic capability to sense and respond to salient
environmental cues, and to learn over time. This desire to quickly respond while
learning requires both the support of management, and suitable technology support.
Agile Management Control
As discussed above, traditional control modes, particularly managerial control
and outcome control, are incongruent with an agile method approach. Managerial
control stresses the empowerment of management over the team to direct the team on
actions to take, and to reward them for compliance. Outcome control is based upon the
concept that the result of the project is well-definable by the team at the outset, and that
proper incentives and rewards can be defined based upon an up front plan. Agile
methodologies on the other had, due to their philosophical belief in an empowered team
and in the empirical nature of the development cycle reject these modes of control.
Instead, agile teams have adopted new control modes that are congruent with the
concepts of agile. methodologies Emergent outcome control, based upon “mid-course”
correction, supports the team’s goal of adaptive goal orienting throughout the project.
Supporting Infrastructure Adoption
Austin & Devin proposed that the agile methodologies indicate a break from
industrial making, into an era of post-industrial making. They illustrated how the
emergence of certain technologies enabled the industrial revolution and the principles
(philosophy) of industrial making. They also proposed that certain technologies most
likely enable post-industrial makers to make the cost of novelty lower. Further, they
conjectured that these supporting technologies might shift the cost curve of the making
of unique deliverables (Austin and Devin 2009). The agile movement has been enabled

78

by the serendipitous emergence of multiple, complementary supporting technologies,
and that these technologies, described below, are a key component of agile teams’
agility. Several technologies that directly impact a team’s ability to adaptively sense and
respond are Source Code Control, automated testing software, and continuous
integration software. We propose that the three constructs described above are
enabling processes for the feedback processes that directly impact the team’s ability to
deliver projects successfully.
Figure 2.8 presents the full theoretical model adopted for this study. Although the
full theoretical model of agile methodologies and their structure was presented here, the
research study that was undertaken in this dissertation focuses solely on testing the four
feedback-related constructs for which direct impacts are proposed. Those constructs
are described and elaborated in Chapter Three.
This model proposes a richer conceptualization of agile method adoption than
has appeared previously in the literature. This conceptualization proposes that agile
method adoption consists of more than the use of practices defined by agile practitioner.
Three additional components, the adoption of an agile, innovative philosophy which
informs and drives the use of the practices, the use of enabling technologies, which
reduce the cost of operating in an iterative, adaptive manner, and congruent, agile
management controls are key components of agile adoption.

79

Figure 2.6. The Theoretical Model of The Impacts of Agile Method Adoption and Use on Project Success.

80

Chapter Summary
This chapter presented relevant literature from the organization theory and
information technology literatures to support a theoretical model that is useful for
conducting research on the impacts of agile method adoption on project success. This
chapter illustrated that treating agile methodologies as a “black box” is inappropriate.
We have presented a nomological network that describes the structure of agile method
adoption and use as consisting of a set of reinforcing constructs. These constructs fall
into two categories, one that impact project outcomes directly, and one that supports or
enables these direct impacts.
Chapter three describes the theoretical model of the direct impacts of agile
methodologies, which will drive the research study performed in this dissertation.

81

Chapter Three: Research Model
Introduction
Chapter two developed a theoretical structure that described a framework for
understanding the adoption and use of agile methodologies as being a two-level
structure, with the adoption of agile philosophy, agile management control, and agile
supporting technology supporting the ability to perform practices that would directly
impact project success. Further, a conceptualization of agile practices was described
that is based upon the feedback mechanisms built into agile methodologies, rather than
the specific engineering and process practices of particular agile methodologies. The
full testing of this theoretical model is beyond the scope of this dissertation. Instead, this
study focuses on the ability to test the impacts of agile practices posited as being most
likely to directly impact project outcomes. In this chapter, a research model for testing
the direct impacts of agile practices on project success is presented. The empirical
portion of this dissertation addresses the following research question:
How do agile methodologies impact project success?
Research Model and Variables
Prior IS research has identified that the nature of development method impacts
are contingent. This means that when there is higher fit between the method itself and
the class of problem being solved, higher performance is predicted (Avison and Taylor
1997; Barki et al. 2001). As uncertainty rises, method flexibility becomes a required
component of fit, and of potential positive project outcomes (Barki et al. 2001;
MacCormack and Verganti 2003; Miller 1992). Further, agile methodologies are claimed
to be most appropriate when used under uncertainty.

82

The research model proposed in this chapter is essentially an operationalization
of components of the theoretical framework exposited in Chapter Two. Figure 3.1 shows
the research model.
The unit of analysis of this dissertation is the individual software project. The key
measure of success is the delivery of a particular project deliverable. A summary of the
variables presented in this model is provided in Table 3.1.
In the following sections, the constructs of the research model are described,
including the conceptualization of agile method and use and project success, and utilize
the theoretical lenses previously discussed to motivate hypotheses of the impact of
agile method use on project success.

83

Figure 3.1: The Theoretical Model for Analyzing The Impacts of Agile Method Adoption on Project Success

84

Table 3.1. Research Variable Definitions
Dependent Variables
Project Management Success

Product Quality

Project Impacts

The level to which project results match
defined goals of scope, schedule &
budget.
The perceived quality of the system as
defined by perceived quality,
effectiveness, completeness, reliability,
suitability, and accuracy.
The level to which the project is perceived
to have positively impacted the
organization, as indicated by perceived
benefits and perceived satisfaction.

Extent of Use of Agile Methodologies
Reduced Up Front Planning
Iterative Delivery
Environmental Feedback

Technical Feedback

Moderating Variables - Uncertainty
Structural Complexity

Technical Complexity

Environmental Dynamism

The level to which the team reduces the
time spent before beginning work.
The level to which the team delivers
functional code each iteration.
The level to which the team utilizes
mechanisms to obtain feedback from
customers and stakeholders.
The level to which the team utilizes
mechanisms to ensure that the system is
functioning properly.

Index of size of team, criticality of project,
number of stakeholders, and team
distribution (organizational and
geographical)
The level of complexity of the system,
and complexity of integration with other
systems.
The level of requirements, technical, and
organizational change during the project.

85

Table 3.1. (cont’d)
Control Variables
Project Size

Organization Size
Management Support

Team Member Experience
Team Member Agile Experience
Time since Agile development adopted

The size of the agile development project,
as indicated by the number of people on
the agile development team.
The size of the organization served by the
agile team.
Willingness of top management to modify
practices and provide support to enable
agile adoption.
Number of years development
experience.
Number of years experience in agile
teams.
Number of months since organization
adopted Agile development.

Project Success
Delivering the promised scope of a project, within time and budget constraints, is
the core of project management. Agile methodologies claim to better manage the
impacts of uncertain environments, and to deliver projects more successfully. The
negative impact of uncertainty on project management success is well documented
(e.g., Keil et al. 1998; Nidumolu 1995; Schmidt et al. 2001; Wallace and Keil 2004; Xia
and Lee 2005). IS project failure is particularly common, due to the delivery of programs
that do not deliver useful features, are delivered late or not at all, or escalate out of
control (Keil 1995; Lyytinen and Hirschheim 1988).
Project success has been viewed as a multi-dimensional construct, including the
performance of both the software development process and the performance of the
product itself. Project performance means the level of performance against (1) project
management metrics, (2) product quality, and (3) perceived impacts of the product. It is
appropriate that they are measured as separate constructs, because they are not
86

expected to correlate highly. For example, while a project might be delivered
significantly over budget, that particular project may at the same time deliver a product
of very high quality.
Most projects, whether utilizing agile methodologies or not, are justified based
upon some set of cost and duration and feature set (scope) estimates established
before the project begins. These project estimates assume the ability to deliver a
suitable system within the constraints of this business-driven timeline. Even though
agile methodologies attempt to minimize up-front planning, agile methodologies have
been reported to deliver projects within high level estimates generated at the beginning
of projects (Coad et al. 1999; Highsmith 2002). Because agile methodologies make
claims of high rates of delivery within established project constraints, one dimension of
project success in this model is project management metric performance.
Agile methodologies also make claims that they delivery more useful software
that better reflects the needs of the customer. In short, they claim that they develop
high-quality software. Because agile development projects involve quick iterative
feedback, the progress of a project is communicated much more transparently than
within traditional development environments. Agile development teams strive to produce
working software as their progress metric. Because agile development teams develop
priorities with their users within the short iteration cycle, adaptation occurs rapidly and
repeatedly throughout the project. This adaptation enables the agile team to identify
required changes to plans, develop an emergent understanding of the real needs of the
customer, and sense and respond to incongruity between the emerging system and the
required system.

87

Finally, although it is not expected that formal organizational impacts such as
ROI will be identifiable for most projects, a successful project will most likely be
accompanied with perceived benefits to the user and organization.

Because, agile

development teams value frequent, direct interaction with users in order to develop a
shared understanding of the business problem and the associated solution
requirements. Because of this, the product of an agile development project should
exhibit a stronger congruence with organizational requirements, and should be
perceived as delivering a significant, positive impact to the organization.
Extent of Agile Method Use
As previously noted, very few of the particular practices advocated by agile
practitioners are truly novel, and most agile development practices are suitable to be
applied outside of an agile project setting. Because of this, the fact that a team performs
agile practices does not necessarily indicate that it has adopted an agile method (Beck
1999; Highsmith 2002). Therefore, to measure the extent of agile method use, the
generalized set of constructs defined in Chapter Two was used: Reduced Up-Front
Planning, Iterative Delivery, Environmental Feedback, and Technical Feedback.
Project feedback events occur naturally and intentionally. As described in
Chapter Two, four generalized processes are most likely to provide feedback to the
team, and drive project impacts.
Reduced Up Front Planning
Reduced up front planning is likely to directly impact project performance in two
ways, by reducing the time to initial feedback, and by reducing waste from planning
tasks too far in advance, leading to rework. The manifesto argues that it is more

88

important to deliver software than the deliverables that are generally created at the
beginning of projects. The agile methodologies surveyed all placed a premium on
reducing the amount of time spent on up front planning. Speed of initial feedback has
been shown to increase performance under uncertainty (MacCormack 2001), and
reducing the window of feedback triggers natural feedback assessments to occur more
quickly (Gersick 1988).
Further, agile practitioners argue that the software development process is highly
uncertain, and the process of completing documentation of system requirements up
front is likely to be wasted effort for tasks scheduled for completion far in the future.
Environmental Feedback
Agile methodologies are founded on the premise that environmental feedback is
key to a project’s success. The agile methodologies prescribe numerous practices
meant to generate environmental feedback, including the retrospective, the daily stand
up, the on-site customer, and the informative workspace among others. These practices
reflect the agile manifesto’s call for direct, face-to-face interaction and their philosophy
that recognizing the need for change, and facilitating change is more productive than
attempting to restrict change (Fowler and Highsmith 2001).
In uncertain environments, feedback is the source data that allows teams to
sense and respond to the environment (Burke et al. 2006). If, as described above, the
software development process is inherently uncertain, access to and processing of
environmental feedback should positively impact project performance (Eisenhardt and
Tabrizi 1995).

89

Iterative Delivery
The first principle of the agile manifesto is “Our highest priority is to satisfy the
customer through early and continuous delivery of valuable software.” This entails an
unstated corollary, which is that the software is deployable at an early stage, and
continues to be deployable throughout the delivery schedule. When software is
delivered iteratively, it can be deployed and utilized much earlier in the project.
Iterations consist of short, time-boxed deadlines, usually no longer than 4-6
weeks (Highsmith 2002; Schwaber 1996; Schwaber and Beedle 2002), and often much
shorter. While prototyping and other evolutionary methodologies emerged before agile
methodologies, agile methodologies are differentiated in that, in most cases, the agile
iteration structure include all or most of the steps of the waterfall method. For the
requirements selected for the iteration, the solution is designed, built, tested, and
delivered (Beck and Andres 2004; Highsmith 2002; Schwaber and Beedle 2002). While
this does not necessarily mean that the code is deployed to production, the
understanding is explicit in several methodologies that the code should be deployable at
the end of any iteration, at the discretion of the user (Beck 1999; Highsmith 2002;
Schwaber and Beedle 2002).
By using this this delivery model, agile teams reduce the timeframe of definitive
feedback, which allows them to respond more quickly to uncertainty that stems from
environmental dynamism (Duncan 1972; Lawrence and Lorsch 1967). By working on
small parts of the system in short timeframes, the potential impact of requirements
changes is limited to those requirements being worked on in the current cycle. By
delivering working software each cycle ambiguity in requirements is reduced, as users

90

provide feedback about a working system, not an abstract requirements document.
These feedback cues are of high quality because, instead of being based upon
representations of the system (as would occur during the analysis and design phase of
a waterfall project), they are based upon a review of the working software. This working
software is a far richer medium for users to evaluate whether the system that is being
built is suitable for the task (Brooks 1987).
Technical Feedback
Technical feedback is achieved in agile methodologies by technology-mediated
processes. The technology-mediated processes are source code control, automated
testing and continuous integration.
Source Code Control, or Software Configuration Management (SCM) provides
developers with the ability to manage the full repository of system components, via
versioning, file control, and configuration (Estublier 2000). In addition, SCM systems
provide build and rebuild support, reduce the need for developers to manage
dependencies between files, and allows for teams to collaboratively and cooperatively
build software via check in and rebase (download changes made by others)
functionality. Most importantly, SCM systems allow for the identification of groups of
changes made at the same time, and allow the rollbacks of single or groups of changes
easily.
Automated Testing software enables software teams to develop libraries of test
code and run them repeatedly. Developers might have previously executed tests on an
ad hoc basis. Automated testing frameworks allow teams to write suites of test code
that cover the unit, system, integration, and user interface levels of testing. These tests

91

are standardized via a framework, and can be run automatically with little effort or
special training. Teams can enforce routines that ensure that all the tests, covering the
entire system, must be run before code is checked into the repository. Previous to the
emergence of automated testing platforms, teams often relied on familiarity with the
software and technical environment to predict the impacts of a change. This familiarity
helped developers to be more efficient in testing and implementing larger changes
because they know what needs to be tested and how to implement the changes (Curtis
et al. 1988).
Eisenhardt and Tabrizi’s (1995) findings presented in Chapter Two suggest that
the speed of testing directly impacts the speed of project completion due to the early
feedback that the team gets as to product failure. As automated testing leads to near
continuous testing of the entire product each day, it should be expected that project
success will be enhanced.
Historically, one of the most time consuming and problematic activities of
software development teams is the integration of various code modules into a single,
working build. Continuous Integration (CI) is the process of integrating the entire code
base in an automated fashion, as often as possible, and successfully (Fowler 2006).
While automated testing can enable team routines that require testing before check-in,
issues can arise if a single developer doesn’t remember to re-base (refresh the code
base locally) before testing, or if it is not feasible to perform local unit testing cannot be
used to test integration to external test systems. In these cases, a centralized,
automated build process can be used to ensure that the full code base is functional at
all times. The presence of automated build software, enabled by the presence of

92

automated testing frameworks and software configuration management, enables agile
development teams to implement practices, such as CI that otherwise could not be
implemented. These practices directly impact the ability for the development team to
respond to the feedback cues that the environment provides, and provide specific
technical feedback to the team.
The previous discussion motivates the first hypothesis:
Hypothesis 1: The extent of agile method use will positively impact project
success.
Uncertainty and Project Success
Uncertainty, as indicated by complexity and dynamism, moderates the impacts of
agile methodology use on project success. We theorize that the moderation effects of
uncertainty will present themselves differently by the component of uncertainty.
Turning first to complexity, it is theorized that increased levels of complexity will
impede the ability for a team to both process and respond to salient environmental
cues. As described above, two dimensions of complexity salient to software
development are structural complexity and technical complexity.
Structural Complexity
Structural complexity is characterized in this study by (1) the size of the project
team, (2) the number of reporting units within the team, (3) the geographic distribution of
the team, and (4) the criticality of the project. As project organizations grow in both size,
and in the number of reporting units represented, the number of stakeholders increase
(Baccarini 1996). Further, structural complexity rises due to the geographical distribution
of the team (Hinds and Mortensen 2005), and the criticality of the project (Xia and Lee

93

2005). Together, high levels of these structural elements increase both the complexity
of requirements definition and coordination, and the cost of communication between
team members. The presence of multiple stakeholders reduces the ability for the project
team to clearly identify the most critical requirements, and higher coordination costs
impede the team’s ability to bring the full capabilities of the team to bear on the problem.
As the number of stakeholders and the cost of communication rise, agile teams
are likely to be particularly negatively impacted. Agile development teams respond to
continual feedback by making repeated updates to the shared mental model of the
team. This shared mental model contains the understanding of the current
environmental situation, current priorities, agreed upon standards of behavior, and
codified team response plans to specific issues that arise.
The development of a shared mental model requires the team to process the
meaning of cues together. Structural complexity in the form of multiple stakeholders,
multiple groups providing requirements, and the level of criticality of the project is likely
to result in the presence of both a greater number of environmental cues in general, and
a higher proportion of heterogeneous cues. The presence of more and potentially
conflicting cues will potentially reduce the ability of the team to respond to the
environment.
Additionally, the geographic dispersion of the team places an additional level of
communication overhead that affects the manner by which the environmental cues are
propagated and processed by the entire team. When teams are collocated in the same
room, face-to-face discussion and spontaneous and serendipitous communication
naturally occur (Kiesler and Cummings 2002). Teams that are separated by distance

94

require additional formal communications protocols to ensure that important information
is shared across all members of the team. In addition, geographic distribution has been
found to increase the level of intra-team conflict (Hinds and Mortensen 2005).
The discussion above motivates the following hypothesis:
Hypothesis 2a: Structural complexity will negatively moderate the impact of agile
methodologies on project success.
Technical complexity arises from the combination of scope and structure of the
project system, and the ambiguity regarding the use of various technologies to complete
the project (Shenhar and Dvir 1996). In this study, technical complexity is theorized as
the combination of external integration requirements, the number of technology
platforms, and the number of distinct modules that are contained within the project
system. Because agile teams focus on the ability to make rapid adjustments throughout
the project, the level of external integration required may hinder adaptation, either due
to limitations of the remote system, the team’s lack of authority to modify the external
system, or due to the delivery schedule of the remote system (Meyer and Curley 1991).
Integration with external systems is a form of reciprocal interdependence. Reciprocal
interdependence means that the outputs of one element become the inputs of another,
and is considered the highest level of interdependence-driven complexity (Thompson
2003). Finally, the number of platforms and number of functional modules of systems
have been shown to cause an increase in technical complexity (McKeen et al. 1994;
Meyer and Curley 1991).
High technical complexity impacts both the ability for agile teams to react to
change in the environment, and the ability for project teams to evaluate the

95

requirements of the system and process those requirements into a design (Tait and
Vessey 1988). As more features of the agile development project system require
interaction with external systems, the impact of integration on the agile development
process can grow more severe. Because most agile methodologies stress the practice
of incremental design, early phases of the project may not recognize the pervasive
nature of the system integration task. As more of the system is designed, a simple
integration architecture that was appropriate for an earlier version of the system may
need to be recreated to be robust enough for the full integration requirements of the
system. Further, agile methodologies manage the risk of incremental design via the
practice of continuous refactoring, and the supporting technologies of source code
control and automated unit testing. However, projects teams with higher integration
requirements may control less of the IT architecture that is relevant to the project,
restricting the ability to perform extensive and efficient refactoring (Fowler 1999).
Even in the case that the existing system changes are possible, the iterative work
process of agile development teams still may be impacted. Agile method philosophy
dictates that the most valuable feature should be addressed first. However, if a very
valuable feature depends upon a to-be enabled external integration, the agile team may
be forced to delay the more valuable feature, and work on other, less valuable features
first. Agile methodologies feedback loops provide important insight to the team about
the congruence of the developed software with business requirements. Delaying
feedback on important requirements may delay or reduce the ability to understand a
broader set of project issues, thus reducing the ability of agile development to impact
project success.

96

Finally, many legacy systems are production systems, and often do not have fully
featured, open availability, test environments. Additionally, since external systems were
often built before the adoption of automated testing methodologies, ad hoc and
continuous testing against legacy systems often cannot be accomplished. This leads to
the necessity to perform manual integration testing and may lead to the reduction in
speed of the CI process. As discussed in Chapter 2, the pace of the CI process dictates
the pace of deploying changes to the code base. A reduction in the pace of CI
negatively impacts the ability to obtain feedback from the environment.
Based upon this discussion, the following is hypothesized:
Hypothesis 2b: Technical complexity will negatively moderate the impact of agile
method use on project success.
The dynamic nature of the systems development environment is reflected in the
extent to which the project team must respond to changing requirements throughout the
duration of the project. Changing requirements are caused by ambiguity in initial
requirements definition, as well as actual volatility in the system requirements due to
emergent understanding of the business problem, or due to real organizational or
technical change (McKeen et al. 1994; Meyer and Curley 1991; Ribbers and Schoo
2002). Environmental dynamism impacts project success because it potentially reduces
the ability for the organization to possess the necessary information to make a decision
(Thompson 2003), or impacts the continued relevance of previously defined
requirements or courses of action. As changes occur within a project environment,
uncertainty rises due to the reduction in the team’s ability to clearly identify cause and
effect relationships as to how environmental factors influence the situation, the

97

reduction in the ability to predict the success of decisions, and to confidently predict the
likelihood of success of actions (Duncan 1972; Xia and Lee 2005). It is specifically
because of this uncertainty that agile methodologies have developed their “sense and
respond” nature (Lyytinen and Rose 2006).
Traditional plan-based development approaches generally perform poorly in
dynamic environments while iterative and adaptive approaches fare worse in stable
predictable environments (Eisenhardt and Tabrizi 1995; Port and Bui 2009). Although
requirements change is an expected part of every software development effort
(Schwaber 1996), some efforts are likely to have higher or lower levels of dynamism.
Because dynamism has been identified as a key driver of software project failures, as
defined by scope, time and budget constraints (Keil 1995), methodologies that better
manage requirements changes should positively impact project performance.
However, when dynamism is low, it should be expected that the impacts of
experiential or adaptive processes would be reduced. These processes add a measure
of overhead to acquire feedback, adapt and respond. When this dynamism is not
present, or when a problem is not novel, this overhead may simply be waste and add
inefficiency. Therefore, agile method impacts would be expected to be higher in
environments in which environmental dynamism is high:
Hypothesis 2c: Environmental dynamism will positively moderate the impact of
agile methodologies on project success.

98

Control Variables
Project Size
Project size is expected to have an impact on project success, and the literature
is inconsistent as to whether agile development methodologies are appropriate for
larger projects. As projects grow in size, complexity increases, as does the potential for
exceptions. Additionally, project size will have an impact on the length of time that it
takes to build the software system. As a project continues across time, additional
external factors of uncertainty such as changing business requirements and market
conditions, may impact the success of the project.
Organization Size
Organization size has long been used as a proxy for a number of organizational
characteristics, including structure, complexity, resource availability, among others.
However, organizational size has been specifically linked to impacts on IS project
implementation cost, time & deployment strategy (Mabert et al. 2003).
Time since Agile Adoption
Time is an important element in agile method use. First, agile practitioners
recommend that teams adopt agile practices gradually. Rather than adopting a number
of new practices, practitioners recommend that teams identify those practices that the
believe may deliver the most impact first then add additional practices as their routines
evolve. Over time, it is expected that teams will adopt more agile practices.
Additionally, as teams become more familiar with the practices, and integrate the
philosophy of agile methodologies into their motivation, they become familiar with the

99

interplay between the practices, and can more effectively bring these practices to bear
in specific situations.
Team Member Skill
Team member skill is measured as a control variable as it is likely to affect
project performance in general regardless of method. Boehm & Turner (2004) claimed
that team member skill was key to agile method performance, and more specifically that
high skill members of the team are more important to the success of agile projects than
traditional plan-based approaches. However, the various agile methodologies make
conflicting claims and prescriptions regarding team member skill. FDD defines two
levels of developer roles required in the team, chief programmer and class owner. While
this makes an assumption about skill, the method doesn’t clearly make statements
regarding skill level. Scrum’s three team roles do not mention skill level at all.
Team Member Agile Experience
Team member experience with agile most likely did not begin with the current
project team. Because of this, if an organization that recently adopted agile acquired
significantly experienced resources to implement its method and execute its project,
team member agile experience is potentially a separate impact on project success.
Chapter Summary
Agile methodologies make a number of normative claims regarding the impacts
of agile methodologies. In this chapter, a research model was described that tests the
theoretical model presented in Chapter Two. The research model included several
hypotheses about the relationships between the constructs, and are summarized in
Table 3.2.

100

Table 3.2. Summary of Hypotheses
Agile Method Use
H1:

The Extent of Agile Method Use will positively impact Project Success

Uncertainty
H2a: Structural Complexity will negatively moderate the impact of Agile Method
Use on Project Success
H2b: Technical Complexity will negatively moderate the impact of Agile Method
Use on Project Success
H2c: Environmental Dynamism will positively moderate the impact of Agile
Method Use on Project Success

In Chapter Four, describe the research study that was performed to test the
model presented in this chapter is described.

101

Chapter Four: Research Methodology
Introduction
The intent of this chapter is to develop and describe the research design that was
utilized to complete this study. According to Creswell, a research design should
illustrate how the sample, methodologies, and general study structure work together to
answer the central research questions posed (Creswell 2003).
Chapter three identified the research questions, propositions and units of
analysis for this study. This chapter will deal specifically with the logic linking the
sample, methodologies, and anticipated data to the propositions, and the criteria for the
evaluation of this data in respect to the propositions.
As described previously, the research on agile methodologies has been
characterized by the measurement of particular agile method practices, such as pair
programming and code ownership. As described in Chapter Three, agile method use is
conceptualized as a multi-dimensional construct. Because of this new conceptualization
of agile method use, this research was exploratory in nature. The research was
conducted by developing a new quantitative survey instrument, which was used to test
the propositions described in chapter three.
Construct Operationalization
Whenever possible and applicable, items were adapted from previously existing
questionnaires. However, because some of the constructs proposed in this study are
new, additional items have been created for this study. Additionally, several additional
dimensions were added to existing constructs.

102

Because only team members could reasonably be expected to accurately answer
all of the questions on the questionnaire, respondents were first asked what their role on
the project was. Based upon this initial response, the respondent would be presented
with the appropriate sections of the survey. A listing of the sections of the survey
presented to each class of respondent is included in Table 4.1
Table 4.1. Sections of Questionnaire Presented to Each Respondent Type
Section of Survey
Team
IT Mgmt
Stakeholder
Agile Method Use
X
Technical Complexity
X
Structural Complexity
X
X
Requirements Dynamism
X
X
X
External Dynamism
X
X
X
Project Success
X
X
X

Extent of Agile Method Use
Agile method use was measured by creating a set of items that indicated the
agile practices that are both universal to agile methodologies, and likely to directly
impact project success The agile practices measured were: Reduced Up Front
Planning, iterative delivery, customer feedback and technical feedback. This portion of
the questionnaire was answered only by the team members involved in the day to day
execution of the project.
For each of these processes, several items were developed utilizing both the
agile manifesto itself, as well as agile method practitioner explanatory literature (Fowler
and Highsmith 2001; Highsmith 2002). Each agile principle was measured with a sevenpoint scale ranging from Strongly Disagree to Strongly Agree. Respondents were given
the option to also select “Don’t Know”. The agile feedback processes were measured as
described in Table 4.2.

103

Table 4.2. Items Indicating the Extent of Agile Method Use
Reduced Up Front Planning
“Please indicate the extent to which you agree or disagree that the following statements
ACCURATELY REFLECT the project team's BEHAVIOR during the project"
1. The team spent less than 10% of the total project timeline on up-front planning
(planning that occurred before ANY coding began).
2. At the beginning of the project, the team tried to make only the decisions that
were necessary for coding to begin.
Iterative Delivery
“Please indicate the extent to which you agree or disagree that the following statements
ACCURATELY REFLECT the project team's BEHAVIOR during the project"
1. The team executed development using a series of short cycles or iterations.
2. At the beginning of each development cycle, the team and business owners
agreed on what would be delivered during the development cycle.
3. At the end of every development cycle, the code was deployable.
Environmental Feedback
“Please indicate the extent to which you agree or disagree that the following statements
ACCURATELY REFLECT the project team's BEHAVIOR during the project"
The team had a short meeting every day to discuss what was going on that day.
1. On a regular basis, the team demonstrated working software to the
customer/user.
2. The team had a review/verification meeting with stakeholders to demonstrate
when software features were complete.
3. On a regular basis, the team reflected on previous work, and looked for ways to
improve team performance.
4. The team had a short meeting every day to discuss what was going on that day.
Technical Feedback
“Please indicate the extent to which you agree or disagree that the following statements
ACCURATELY REFLECT the project team's BEHAVIOR during the project"
1. Members of the team integrated code changes as soon as possible.
2. Every programmer was responsible for writing automated tests for the code he or
she wrote.
3. Programmers ran a set of automated tests until they all ran successfully before
checking in changes.
Scale:
5-point Likert scale: Strongly Disagree to Strongly Agree [or Don’t Know]

104

Project Success
Because agile methodologies stress the development of shared understanding
amongst the team and stakeholders, the questions about the success of the project
were presented to all groups of respondents. Project success was measured by three
constructs: (a) Project Management Success, (b) Product Quality, and (c) Perceived
Project Impacts. Project management success will be measured using the “iron triangle”
measures of budget forecast, scope forecast and duration forecast. Previous literature
indicates that these measures are subject to gaming based upon political pressures.
Even so, most IS projects, whether utilizing agile method or not, are expected to be
justified based upon some set of up-front cost estimates, and the ability to deliver a
suitable system within the constraints of a business-driven timeline.
Table 4.3. Items Indicating Project Management Success
Project Management Metrics
1. In comparison to the initial budget estimate the final budget was:
2. In comparison to the initial time estimate the final project duration was:
3. In comparison to the initial feature set (scope) the final project scope was:
Scale
5-point likert scales:
1. “Very much lower” to “Very much higher”
2. “Very much shorter” to “Very much longer”
3. “Very much smaller” to “Very much larger”

Even though significant criticism has been leveled at agile method for its lack of
up front planning, agile method have been reported to deliver projects within the high
level estimates generated at the beginning of projects (Coad et al. 1999; Highsmith
2002). For these reasons, it is reasonable to measure the “iron triangle” as an indicator

105

of agile project success. The measures for project management success are listed in
Table 4.3.
Project management success was measured with three constructs: (a) budget
estimation, (b) timeline estimation, and (c) scope estimation. Each construct will be
measured with six items. The items for each construct are very similar, and utilize the a
five point likert scale ranging from Very Much Lower to Very Much Higher. In addition,
the respondents were able to indicate “don’t know”, or that there was not an initial
estimate made.
Product quality was measured with a set of scales adapted and modified from
Wixom and Todd (2005). Because these original scales were specific to particular
technology contexts, they were generalized. The product quality dimensions measured
were: (a) quality, (b) usefulness, (c) completeness, (d) reliability, (e) accuracy, and (f)
suitability. The items used for Product quality are presented in Table 4.4.

106

Table 4.4. Items Indicating Product Quality
Prompt
Please give your opinion about the following statements about the project system:
Quality
1. In terms of system quality, I would rate [the project system] highly.
2. Overall, [the project system] is of high quality.
3. I would give the quality of [the project system] a high rating.
Usefulness
1. [The project system] improves users abilities to perform their tasks.
2. The project system] allows users to get work done more effectively.
3. [The project system] allows users to get their tasks done more quickly.
Completeness
1. [The project system] provides users with a complete set features and information.
2. [The project system] is a comprehensive solution.
3. [The project system] provides users with all needed information to do their tasks
in the system.
Reliability
1. [The project system] operates reliably.
2. The company can rely on [the project system].
3. The operation of [the project system] is dependable.
Accuracy
1. [The project system] properly performs the tasks it was intended to perform.
2. There are few errors or bugs in [the project system].
3. The information provided by [the project system] is accurate.
Suitability
1. The [the project system] delivered the desired project outcome.
2. The [the project system] accomplishes what was needed.
3. The [the project system] does what was it is supposed to.
Scale
5-point likert scale: Definitely Not True to Definitely True

The perceived benefits of the system were measured with two constructs asking
respondent their perceptions of the organizational benefits associated with the project.
These measures are summarized in Table 4.5.

107

Table 4.5. Items Indicating Organization Benefits
Customer Satisfaction
Prompt
Please give your opinion about these statements related to YOUR PERCEPTION of
satisfaction with the project system:
1. The users are satisfied with [the project system].
2. The customer is satisfied with [the project system].
3. [The project system] satisfies the company’s needs.
Organizational Impacts
Prompt
Please give your opinion about these statements related to YOUR PERCEPTION of
satisfaction with the project system:
1.
2.
3.
4.
5.

Because of this project, our organization can better realize its goals.
This project helped our organization to perform better.
Our organization is more competitive because of this project.
This project was worth the investment.
This project delivered on its promise.

Scale
5-point likert scale: Definitely Not True to Definitely True

Control Variables
Six

control

variables

were

measured:

project

size,

organization

size,

management support, team member experience, team member agile experience, and
time since the organization adopted agile method. Management support was measured
through modifying items previously used in Purvis, et al. (Purvis et al. 2001), with slight
modifications.
Organization size was measured by asking the respondent to estimate the total
number of people in the organization, while team size was measured by asking the
respondent to estimate the number of people who participated on the team.
Experience level was measured by asking the respondent how long he had been
working in software development. Agile experience was measured by asking the
108

respondent how long she had been working on agile projects in general. Finally, each
respondent was asked to indicate how much time had elapsed since the organization
had adopted agile methodologies. Control variable items can be found in the full survey
instrument in Appendix A.

Pre-Testing the Survey
The lack of pre-testing has often been cited as a major source of failures of
otherwise well-conceived studies. Converse & Presser (Converse and Presser 1986)
argue that even well-established item scales should not be assumed to properly
translate into new contexts. Because of the novel nature of the items in this survey, and
because the items utilized from previous research were modified, a two-stage
confirmatory pre-test design was utilized.
The initial pre-test of the survey instruments was carried out using a convenient
sample of five volunteers who were experts in agile development. Each volunteer was
given the survey via the Qualtrics online survey tool which was used for the full survey
implementation. As each volunteer completed the survey, he or she was asked to
provide initial feedback as to the instrument, flow, item validity and applicability to agile
method. Then, the volunteer was asked specific questions about the meaning of the
indicator items on the survey. Some items were found to be ambiguous based upon the
respondent’s misidentification of the question’s meaning and/or scale. Based upon the
pretest, items were modified, flow was altered, and the final survey was created.
After the initial pre-test, a set of pilot tests with three groups of 8-15 respondents
(36 total responses) was performed to test initial construct reliability. After each pilot

109

test, the group of respondents was asked about what they were thinking during each
section of the survey, about sequencing and clarity of questions, and about areas of
importance that were excluded from the survey. After this phase, the final version of the
survey was produced.
Sample Design and Data Collection
A field survey of information systems development teams who self-identified as
using agile methodologies was performed using multiple respondent design. We utilized
a purposive sample design to recruit these teams. It is appropriate to administer this
survey to a purposive sample, because all project teams do not use agile method. As
the major constructs of interest involve the adoption of agile methodologies, sampling
from outside the population active agile development teams would be inappropriate.
The sample was obtained via a two-phase recruitment effort. In the first phase, a
passive invitation was addressed to two groups of self-reported agile practitioners. First,
an email and web posting announcing the survey and calling for volunteers was posted
to the membership list of a large agile affinity group in the Midwest. In addition, the
invitation was posted by several recognized agile authorities on a set of agile affinity
groups on the LinkedIn.com social networking platform. To attempt to generate a high
response rate, the invitation and survey instrument included strong assurances of
confidentiality and anonymity. Respondents were asked to identify a project that he or
she worked on in the past year that utilized agile method, and the role that was played
on the project. Additionally, the respondent was asked to give the time frame of when
this project was completed.

110

Because the study sought to receive at least two respondents per team,
respondents from these invitations were asked to provide additional contact information
for other team members, IT management, and stakeholders. This phase of the data
gathering generated about 75 responses of which only 22 were usable, and did not
generate additional usable additional team responses.
The second phase of data gathering involved direct recruitment of teams. The
researcher solicited organizations to volunteer by presenting the research project at
several agile affinity groups in the Midwest, and by directly contacting organizations
known to practice agile methodologies. Once an organization expressed interest, a
short explanation of the research project was shared, and any necessary corporate
approval for the research project was obtained. Two incentives were provided during
this phase. First, individuals were offered a $5 Starbucks gift card for fully completing
the survey. Additionally, respondents were given the opportunity to request a white
paper detailing results of the study. Finally, organizations that provided 30 or more
respondents from 5 or more teams were given the added incentive of a customized
report that contrasted the results of the organization’s teams with the full sample.
Because teams were identified by the organization, each team was given a unique
identifier which was sent to the organization to distribute. Based upon these unique
identifiers, responses were able to be tied to particular teams.
Table 4.6 describes the structure of the sample acquired. Because the
composition of agile teams is dependent upon organization, an “other” category was
added to the survey. Where it was clear that the respondent should have chosen one of

111

the defined options, the record was recoded. The recoded numbers are listed in the
table.
Table 4.6. Demographics Breakdown of Respondents
Role
Phase 1 Phase 2
Total
%
Cumulative %
Business Stakeholder
3
42
45
11.4%
12.2%
IT Management
4
27
31
8.4%
20.6%
Team Lead
4
35
39
8.9%
31.2%
Architect
2
4
6
1.9%
32.8%
Developer
3
126
129
34.1%
67.8%
Project Manager
5
14
19
3.8%
72.9%
QA/Testing
0
32
32
8.9%
81.6%
Business Analyst
0
23
23
6.2%
87.8%
Other
1
44
45
16.3%
100.0%
Totals
22
347
369

Based upon the role selected, each user was presented with the sections of the
survey noted in Table 4.7. Respondents who chose the “Business Stakeholder” role
were presented with the stakeholder section of the survey. Respondents who chose the
“IT Management” role received the IT Management sections of the survey. All other
roles received the full team member survey.
Response volume from the project teams varied, with project response rates
ranging from 1 to 20 respondents. The full breakdown of team responses is presented
in Appendix B. Responses were received from 83 teams, of which 57 teams provided
more than one response.
Because all of the survey items were not presented to each class of respondent,
there are several constructs with data missing for particular classes. Table 4.7 presents
a summary of the missing data by item. There are two classes of “missing” data.
Missing (not presented) indicates the number of respondents who did not have the
opportunity to provide data because the question was not presented. The missing

112

column describes true “missing” data, meaning that the item was presented to the
respondent, but no answer was provided.
Multiple imputation was performed on the data, using 10 imputations. After this
process, there were still four records for which data could not be imputed. Those four
cases were dropped, reducing the size of the sample to 369 respondents.

113

Table 4.7. Missing Data Summary for 373 Cases
Variable
Construct
NonMissing
Missing
(not
presented)
Team Size
Structural Complexity
325
42(11%)
Team Companies
Structural Complexity
326
42(11%)
Team Departments Structural Complexity
326
42(11%)
Project User
Structural Complexity
325
42(11%)
Groups
Geographic
Structural Complexity
326
42(11%)
Distance
Project Criticality 1
Structural Complexity
325
42(11%)
Project Criticality 2
Structural Complexity
325
42(11%)
Project Criticality 3
Structural Complexity
325
42(11%)
Tech Comp 1
Technical Complexity
293
73(19.6%)
Tech Comp 2
Technical Complexity
293
73(19.6%)
Tech Comp 3
Technical Complexity
293
73(19.6%)
Tech Comp 4
Technical Complexity
293
73(19.6%)
Req. Dynamism 1
Dynamism
362
0
Req. Dynamism 2
Dynamism
361
0
Req. Dynamism 3
Dynamism
362
0
Ext. Dynamism 1
Dynamism
362
0
Ext. Dynamism 2
Dynamism
361
0
Ext. Dynamism 3
Dynamism
361
0
Tech. Dynamism 1
Dynamism
362
0
Tech. Dynamism 2
Dynamism
361
0
Reduced Up Front 1 Agile Use
289
73(19.6%)
Reduced Up Front 2 Agile Use
289
73(19.6%)
Env Feedback 1
Agile Use
290
73(19.6%)
Env Feedback 2
Agile Use
290
73(19.6%)
Env. Feedback 3
Agile Use
290
73(19.6%)
Env. Feedback 4
Agile Use
290
73(19.6%)
Env. Feedback 5
Agile Use
290
73(19.6%)
Tech Feedback 1
Agile Use
290
73(19.6%)
Tech Feedback 2
Agile Use
290
73(19.6%)
Budget Outcome
Budget Outcome
359
0
Time Outcome
Time Outcome
360
0
Scope Outcome
Scope Outcome
359
0
Reliability 1
Quality
357
0
Reliability 2
Quality
357
0
Reliability 3
Quality
357
0
Usefulness 1
Quality
360
0
Usefulness 2
Quality
359
0
Usefulness 3
Quality
360
0

114

Missing

6(1.6%)
5(1.3%)
5(1.3%)
6(1.6%)
5(1.3%)
6(1.6%)
6(1.6%)
6(1.6%)
7(1.8%)
7(1.8%)
7(1.8%)
7(1.8%)
11(2.9%)
12(3.2%)
11(2.9%)
11(2.9%)
12(3.2%)
12(3.2%)
11(2.9%)
12(3.2%)
11(2.9%)
11(2.9%)
10(2.6%)
10(2.6%)
10(2.6%)
10(2.6%)
10(2.6%)
10(2.6%)
10(2.6%)
14(3.8%)
13(3.5%)
14(3.8%)
16(4.3%)
16(4.3%)
16(4.3%)
13(3.5%)
14(3.8%)
13(3.5%)

Variable

Table 4.7. (cont’d)
Construct
NonMissing

Completeness 1
Completeness 2
Completeness 3
Benefits 1
Benefits 2
Benefits 3

Quality
Quality
Quality
Org. Benefits
Org. Benefits
Org. Benefits

356
357
356
358
357
358

Missing
(not
presented)
0
0
0
0
0
0

Missing

17(4.6%)
16(4.3%)
17(4.6%)
15(4.0%)
16(4.3%)
15(4.0%)

Chapter Summary
This chapter described the method used for the measurement and collection of
data for this study on the use and impacts of agile methodologies. Chapter five
describes the results.

115

Chapter 5: Analysis and Results
Introduction
This chapter presents results of statistical analysis performed to test the specified
hypotheses on the data obtained from the research described earlier.
Analysis Approach
Most of the measures used in this study were either new or were modified from
their original sources, so the analysis performed was exploratory. First the instrument
was verified utilizing exploratory factor analysis. Next the nature of the relationships
between the factored constructs was investigated. Finally a test of the full model was
performed using a generalized linear model. The results of the analysis are described
below.
In Chapters 2 and 3, it was theorized that Agile Method Use would positively
impact project success. Further, it was proposed that this relationship was moderated
by three components of uncertainty: structural complexity, technical complexity, and
dynamism. The independent variable of Extent of Agile Method Use first.
Extent of Agile Method Use
Because the proposed theoretical model and the items developed to indicate the
constructs were new, a exploratory factor analysis was performed. When using a new
measurement instrument, it likely that some items will not perform as well as others, and
exploratory factor analysis can be used to identify poorly clustered items. All exploratory
factor analyses in this chapter were performed using Stata, utilizing principal
components factors, and oblique oblimin rotation. Using oblique rotation rather than an
orthogonal rotation is appropriate in this case as the factors we are analyzing are

116

dimensions of a higher-order construct, and would be expected to correlate with one
another. Agile method use was originally predicted to include four dimensions, however
the data factored into three factors, environmental feedback, technical feedback, and
reduced up front planning. The theorized factor of iterative delivery did not materialize in
this data set and was dropped from the analysis, although one of the items that was
anticipated to load on iterative delivery loaded instead on environmental feedback.
Factor loadings are presented in Table 5.1
As can be seen all with loadings are over .70, and no cross loadings over .30.
Descriptive statistics for the items included in the factors are presented in Table 5.2.
Table 5.1. Factor Loadings for the Dimensions of Agile Method Use
Reduced
Up Front
Env.
Technical
Planning
Item
Feedback
Feedback
Less than 10% Upfront planning
-.0054
.0593
.8617
.0144
-.0256
.8964
Minimum decisions before beginning
work
Short Cycles
Defined Scope for Cycles
Daily Feedback Meeting
Regular Demonstration
Verification Meetings

.7003
.8656
.7142
.7299
.8138

.0224
-.1686
.1511
.1980
.0120

.2028
.0384
-.0606
-.0054
-.0738

Tests Required
Tests Run Before Check In

.0145
.0141

.9175
.9547

.0590
-.0285

Environmental feedback is represented by those items that indicated the team’s
use of structures and processes that provided opportunities to collect and process
environmental data. As discussed in Chapter Two, the ability to regularly allow feedback
to cross the team buffer is an important contributing factor to team adaptability, and
performance in an uncertain environment (Burke et al. 2006; Eisenhardt and Tabrizi

117

1995; MacCormack 2001). Technical Feedback was indicated by the process of
automated testing – both the practice of requiring tests and requiring their successful
execution before checking in changes.
Reduced up front planning reduces waste regarding the definition of
requirements and actions that will be taken in the future, and are likely to change, and
decreases the time between project commencement and initial feedback. Reduced up
front planning was indicated by two items that related directly to minimizing the time and
decisions made before starting work.
Table 5.2. Descriptive Statistics for Agile Use Items
Item
Mean
Std. Dev.

Min

Max

Less than 10% Upfront planning
Minimum decisions before beginning work

3.13
2.98

1.10
1.06

1
1

5
5

Defined Scope for Cycles
Daily Feedback Meeting
Verification Meetings

3.74
4.23
4.01

.99
.97
.85

1
1
1

5
5
5

Tests Required
Tests Run Before Check In

3.47
3.56

1.30
1.23

1
1

5
5

Outcome Variables
The outcome variables for this study were quality, organizational benefits, and
project management outcomes. We measured quality with six dimensions, perceived
quality, reliability, usefulness, accuracy, completeness, and suitability. In initial
exploratory factor analysis, these dimensions loaded onto three factors, with usefulness
and completeness loading on their own, and the remaining four dimensions loading
together. Because Chapter Three proposed perceived quality to be indicated by the
other five factors, we dropped it to determine if it influenced the other factors to load
118

together. However, after dropping the perceived quality items, the data still loaded on
three factors.
Accuracy did not load on a single factor, and was dropped. Reliability and
perceived quality loaded together on a single factor. This combined factor was
compared with the quality and reliability factors separately, and was found to perform
nearly identically in correlation/covariance comparisons, and in simple regression
analyses. The combined factor performed worse than the reliability-only construct, and
thus, perceived quality was dropped.

Table 5.3. Factor Loading for Product Quality and Benefits Items
Item
Reliability
Usefulness
CompleteOrg.
ness
Benefits
System operates reliably
.9476
.0115
.0250
-.0587
Company can rely on it
.9255
-.0046
.0173
.0541
It is dependable
.9537
-.0070
-.0310
.0348
Improves users’ tasks
Helps users do tasks
Users get work done

.0516
.0268
-.0769

.8981
.9375
.9010

-.0246
.0298
-.0059

.0147
-.0496
.0807

Complete set of features
All needed information
Comprehensive solution

.0043
-.0398
.0839

-.0435
-.0353
.1205

.8800
.9120
.8115

.0550
.0652
-.1250

Org realize goals
Org performs better
Org more competitive

.0534
-.0064
.0018

.0386
.0730
-.0380

-.0007
.1104
-.0468

.8560
.7955
.8844

In the interest of parsimony, reliability items loaded at the highest level were
retained while accuracy and suitability items were dropped, leaving Quality with three
factors. Then, an additional EFA added the organizational benefits items. Three of the

119

organizational benefits loaded on one factor and were retained; the final EFA for the
product quality dimensions and org benefits factor is listed in Table 5.3.

Item

Table 5.4. Descriptive Statistics for Product Quality and Benefits Items
Mean
Std. Dev.
Min
Max

System operates reliably.
Company can rely on system
The system is dependable

4.20
4.30
4.26

.739
.661
.694

1
1
1

5
5
5

System improves users’ tasks
Helps users complete tasks
Helps users get work done

4.33
4.28
4.18

.709
.752
.782

1
1
1

5
5
5

Complete set of features
All needed information
Comprehensive solution

3.82
3.61
3.73

.918
1.09
.957

1
1
1

5
5
5

Org can better realize goals
Org performs better
Org more competitive

4.08
3.96
4.01

.779
.872
.860

1
1
1

5
5
5

The

three

retained

quality dimensions

are

reliability,

usefulness,

and

completeness. Reliability includes the items that relate to the sense that the system will
not let the user or the company down, in short, the level to which the organization can
depend on the system. Usefulness includes those items that relate to the system’s
ability to help uses improve the manner in which they complete their tasks, get their
tasks done at all, and complete their work. Completeness was indicated by the three
items that denoted the system’s features as being complete, information as being
complete, and solving a complete business problem. The descriptive statistics for the
product quality and benefits items are listed in Table 5.4.

120

Uncertainty
The discussion in Chapter Two theorized that the factors of Structural
Complexity, Technical Complexity, and Environmental Dynamism indicate uncertainty in
a software development project. This theorized factor model was tested with exploratory
factor analysis, and it was found that the data loaded into seven dimensions of
uncertainty, three that related to dynamism (rather than the proposed two), and two that
related to technical complexity.
Dynamism
It was initially proposed that dynamism would come from three sources, changes
in requirements, changes in the technical environment, and changes to the
organizational environment outside of the team. Dynamism did load into these three
factors, with a minimum loading of .6924, with no cross loadings over .23. Factor
loadings for the dimensions of dynamism are presented in Table 5.5, and descriptive
statistics are listed in Table 5.6. External dynamism is indicated by items that describe
changes that the project was associated with that occurred outside the project. These
changes included associated business processes, organizational structure and
information needs. Requirements dynamism relates to changes in actual system
requirements during the project. Finally, technical dynamism relates to changes in
infrastructure and software development environments during the project.

121

Table 5.5. Factor Loadings for the Dimensions of Dynamism
Item
Requirements
Technical
External
Dynamism
Dynamism
Dynamism
Early Req. Change
Late Req. Change
Final Req. Different

.7815
.8252
.8280

-.0220
-.0352
.0816

-.0033
.0596
-.0359

IT infrastructure change
SW Dev. tool change

-.0145
.0464

.8721
.8918

.0637
-.0534

Business Process Change
Org Structure Change
Information Needs Changed

-.0183
-.1110
.1999

-.1275
.2814
.0449

.8849
.6973
.7235

Item

Table 5.6. Descriptive Statistics for Uncertainty Items
Mean
Std. Dev.
Min

Max

Early Req. Change
Late Req. Change
Final Req. Different

3.68
3.20
3.08

.973
1.02
.996

1
1
1

5
5
5

IT infrastructure change
SW Dev. tool change

2.22
2.16

.907
.884

1
1

5
5

Business Process Change
Org Structure Change
Information Needs Changed

2.93
2.27
2.49

1.08
.865
.948

1
1
1

5
5
5

Complexity
In Chapter Two, it was theorized technical complexity to be indicated by two
factors, structural complexity and technical complexity. In the study data, the technical
complexity items loaded into two dimensions: Integration complexity and system
complexity. Integration complexity consists of the items related to the number of
software environments and platforms used within the product system, as well as
integration with external systems that was necessary in the project. System complexity

122

consists of the evaluation by the project team of the high-level project system
complexity and organizational systems environment complexity. Factor Loadings for
Technical Complexity are displayed in Table 5.7.
Table 5.7. Factor Loadings for the Dimensions of Technical Complexity
Item
Integration
System
Complexity
Complexity
Multiple SW Environments
.7043
.0102
Multiple Tech Platforms
.7408
-.0756
External Integration
.8133
.0083
Complexity of Integration
.7651
.0541
Org. System Environment Complex
Project System Complex

.0710
-.0604

.8673
.9031

Structural complexity was constructed as an index that averaged the items that
indicated the size of the team, the number of departments represented on the team, the
number of user groups providing requirements, the project criticality, and the geographic
distribution of the team. Descriptive statistics of all complexity items are listed in Table
5.8.

123

Table 5.8. Descriptive Statistics for Technical and Structural Complexity Items
Item
Mean
Std. Dev.
Min
Max
Project part of strategic plan
Project in response to competition
Project failure has financial impact

4.34
3.60
3.25

.676
1.09
1.03

1
1
1

5
5
5

Org. System Environment Complex
Project System Complex

2.61
2.57

1.11
1.10

1
1

5
5

Multiple SW Environments
Multiple Tech Platforms
External Integration
Level of Integration Complexity

3.77
3.77
3.77
3.34

1.06
1.03
1.03
1.06

1
1
1
1

5
5
5
5

Once the first order factors were generated, a second-order EFA was performed
by factor. The factor loadings for the theorized second order factors are presented in
Tables 5.9a-d.
As can be seen in Table 5.9a and Table 5.9b, second order loadings of Quality
and technical complexity were acceptable, with all loadings over .7.
Table 5.9a. Second Order Factor Loadings: Quality
Factor
Quality
Reliability
.7714
Usefulness
.7372
Completeness
.7783

Table 5.9b. Second Order Factor Loadings: Technical Complexity
Factor
Complexity
System Complexity
.7549
Integration Complexity
.7549

Table 5.9c presents the factor loadings for the Agile Method Use factor.
Environmental feedback and technical feedback meet the threshold of .7 that indicate
124

suitable fit. Reduced up front planning loads only at the .69 level. This indicates only
marginal fit, but due to fact that this scale is newly developed, .69 should be considered
acceptable, although further development and improvement of this scale is required in
future studies.
Table 5.9c. Second Order Factor Loadings: Agile Method Use
Factor
Agile
Environmental Feedback
.6963
Technical Feedback
.7289
Reduced Up-Front Planning
.6913

Table 5.9d presents the second order factor loadings for Dynamism. External
and Technical dynamism each load above .7, but requirements dynamism loads only at
.63. While this loading is low, this factor is conceptually a component of dynamism.
Further, while a loading of .7 is generally considered the cutoff for suitable fit, when
correcting for sample size, .6 may be considered acceptable in this case, with the
sample size being higher than 300 cases.
Table 5.9d. Second Order Factor Loadings: Dynamism
Factor
Dynamism
Requirements Dynamism
.6252
External Dynamism
.7235
Technical Dynamism
.7253

125

Factor
Quality

Table 5.10. Correlations and Variance Inflation Factors
Quality
Agile
Structural
Technical
Complexity
Complexity
1.05

.1632
1.27
(.0003)
Structural Complexity
-.0364
.2381
1.16
(.4858)
(.0000)
Technical Complexity
.0270
.2343
.3003
1.18
(.6057)
(.0000)
(.0000)
Dynamism
-.1178
.0038
.2012
.1594
(.0238)
(.9416)
(.0001)
(.0022)
Management Support
.1339
.3810
.1017
.1631
(.0101)
(.0000)
(.0513)
(.0017)
(Numbers in parentheses indicate p-value, Numbers on the diagonal represent VIF)

Dynamism

Management
Support

Agile

126

1.07
-.0345
(.5091)

1.19

The correlations between the proposed model factors are shown in Table 5.10.
Because of the highly correlated nature of several of the predictors a diagnostic test for
multicollinearity was run, and the variance inflation factors are included on the diagonal
of Table 5.10. The presence of a VIF over 10 is indicative of multicollinearity. As the
highest VIF for the factors is 1.27, this indicates that multicollinearity is not an issue.
However, as can be seen the correlations between the theorized predictors and
outcome variables are generally non-significant. This indicates that the data factors are
either truly not correlated, or are being influenced by violations of the assumptions of
correlation analysis. Correlation analysis depends upon the assumptions of linearity in
the relationship between the two variables, normality of the distribution of each of the
variables, and homoscedascity across the range of the relationship.
In the next section the analysis of the data as to whether it meets these and other
assumptions of statistical analysis.

Analysis of Data Characteristics
Several tests were run to evaluate the conformance of the data to statistical
assumptions. The description of the tests, results, and the implications of these tests
are described below.
Tests for Normality
First, Shapiro-Wilk tests were performed on all of the model variables. Results
are shown in Table 5.11. In this test, the null hypothesis that the data is normally
distributed is rejected if the p value is less than .05. As shown in the table, all of the
outcome variables, and the agile method use variable are below the .05 level, which

127

means the null hypothesis that they are distributed normally can be rejected. The three
moderators in the model, Structural Complexity, Technical Complexity, and Dynamism
all are above the .05 level, indicating that we cannot reject that they are distributed
normally.
Table 5.11. Shapiro-Wilk Test for Normality of Distribution
Factor
p-value
Quality
.0007
Organization Benefits
.0000
Budget Outcome
.0012
Time Outcome
.0000
Scope Outcome
.0000
Agile Method Use
.0000
Structural Complexity
.4800
Technical Complexity
.8105
Dynamism
.0569

While the Shapiro-Wilk test can indicate whether or not the distribution may be
normal, it does not indicate the nature of the non-normality. In order to understand the
nature of the data distributions, a series of inspections of the data were performed,
using histograms, box plots, symmetry plots and Q-Q normal plots. The output of these
tests is shown in Figures 5.1-5.5

128

Figure 5.1a: Diagnostic Outputs for Quality - Histogram

129

Figure 5.1b: Diagnostic Outputs for Quality – Box Plot

130

Figure 5.1c: Diagnostic Outputs for Quality – Q-Normal Plot

131

Figure 5.1d: Diagnostic Outputs for Quality – Symmetry Plot
As can be seen from Figures 5.1a and b, while no outliers are evident in the box
plot, quality has a slight left skew. These tests confirm the results of the Shapiro-Wilk
test, and give the additional insight that the right side of the distribution is heavy. The
symmetry and q-normal plots (figures 5.1c and 5.1d) indicate that the right tail is the
most problematic departure from normality.
Figures 5.2a-d display the output from the tests for normality for agile method
use.

132

Figure 5.2a: Diagnostic Outputs for Agile Method Use - Histogram

133

Figure 5.2b: Diagnostic Outputs for Agile Method Use – Box Plot

134

Figure 5.2c: Diagnostic Outputs for Agile Method Use – Q-Normal Plot

135

Figure 5.2d: Diagnostic Outputs for Agile Method Use – Symmetry Plot
Agile Method use is also not normal. Based upon the plots, it can be seen that
both tails are heavy, and the distribution may have multiple peaks. One outlier may be
indicated. Tests for outliers will be discussed below.
Figures 5.3a-d display the diagnostic output for the Dynamism factor. This output
indicates the presence of outliers, but besides the outliers, the rest of the plots indicate
that Dynamism approximates normality.

136

Figure 5.3a: Diagnostic Outputs for Dynamism - Histogram

137

Figure 5.3b: Diagnostic Outputs for Dynamism – Box Plot

138

Figure 5.3c: Diagnostic Outputs for Dynamism – Q-Normal Plot

139

Figure 5.3d: Diagnostic Outputs for Dynamism – Symmetry Plot
Diagnostics for the structural and technical complexity were also performed,
which indicated that these distributions were approximately normal, as was expected
from the results of the Shapiro-Wilk test.
Tests for Unusual and Influential Data
Next tests for the influence of outliers in the dynamism, agile, and quality data
were performed. Figure 5.4 presents a matrix scatter plot for the relationships between
each pair of variables. There are several potential outliers indicated in the agile,
dynamism, and quality variables. We calculated cooks d for each of the relationships
and dropped those cases that exceeded the cutoff for any of the relationships as set by
Bollen and Jackman (1990).

140

Figure 5.4. Matrix Plot of Factor Relationships
141

Based upon this analysis, 42 highly influential cases were dropped, leaving the
sample with 326 observations. Another Shapiro-Wilk test was performed. No change in
results was observed, indicating that influential outliers did not drive the non-normality of
the data.
Tests for Linearity
When performing the correlation analysis above, the lack of significant
relationships between the factors was noted. One assumption of correlation analysis is
the linearity of the relationships between the variables. We performed several analyses
of the residuals when performing univariate regressions between the Quality dependent
variable and the other factors in the model. Analysis of the homscedascity of residuals
and review of augmented component-plus-residual plots indicated that the assumption
of linear relationships was unlikely to hold between the predictors and the dependent
variables.
In order to investigate the nonlinearity of the relationships between the variables,
PROC TRANSREG in SAS was used. PROC TRANSREG allows for regression
analysis to be performed with transformed variables. In this analysis, the spline function
was chosen, as TRANSREG will fit a spline function transformation to the observed
data. Usually, use of a spline function requires the designation of “knots”, or breakpoints
in the data that become start and end points for independent estimation of linear
relationships for that segment. Rather than manually setting knots for the analysis,
PROC TRANSREG fits a transformation to each variable based upon the shape of the
observed data. Figure 5.5 shows the transformation plots for the variables in the model.

142

As can be seen from the plots, the relationships between the variables are nonlinear,
and are both bimodal and monotonic.
Due to the fact that the dependent variable is non-normally distributed, and the
relationships between the ratio variables are nonlinear, transformation of the data
before analysis was necessary.
Transformation of Data
When dealing with non-normal data distributions, a common practice is to
attempt to transform the data to create a normal distribution. Several standard
transformations were attempted on the Quality, BenFac, and Agile factors, none of
which allowed the data to approximate normality.

143

Quality

Spline(Agile)

Spline(Dynamism)

Spline(Technical Complexity)
Spline(Structural Complexity)
Figure 5.5. Spline Transformation Plots
144

As the survey items presented the respondents with several likert-scale response
options, the original survey data was coded as ordinal. After the factoring and scoring,
the variables had been converted to a continuous measure. Because the distance
between the intervals in the ordinal categories is arbitrary between subjects, and
because the data was originally interpreted by the subjects as ordinal, the factor score
data was recoded as ordinal. Ordinal data categorized into at least three categories
allows for only slightly less fidelity to generate substantive inferences than does ratio
data (Weiss 1986). Because the scale on which the items were answered was a five
category scale the data was discretized utilizing five equal width categories. After
recoding, all of the discretized variables were again tested with the Shapiro-Wilk test,
and the null hypothesis that the variables were normal could not be rejected for any of
them.
Analysis Results
This section presents the statistical results for the data analysis that was
performed to test the hypotheses of the theoretical model.
Because the theory and data are multilevel in nature (i.e. – people in project
teams), the sample violates several statistical assumptions. The first is that the
observations are independent, and that the random errors are independent,
homoscedastic and normally distributed. If the sample was tested without regard to the
expected similarity of responses it would imply that the members of a group share no
common attitudes or other characteristics that might influence their responses regarding
the project. This is specifically contrary to the assumption about the responses from the
same group. Alternatively, while the data could be aggregated a single group-level

145

mean be utilized for the analysis, this would reduce statistical power, and potentially
provide biased estimates of group-level relationships, and lead to incorrect causal
inferences (Bovaird 2007).
There are several robust techniques to test multilevel models, and to separate
the between group variance (random effects) from the within group variance (fixed
effects). However, these methodologies require sufficient sample size at each level. We
attempted to perform multilevel modeling with four different statistical packages and
seven different analysis techniques. While first-level effects were successfully
generated in all of the tools, none could fit the second-level effects.
Consequently, a population-averaged model was utilized. This type of analysis
fits a single model to all of the data, but controls for within-cluster correlation. While this
type of model does not estimate the within and between team variance like a multilevel
model would, it uses the concept of nesting to correct for the presence of similarity
within clusters of responses. The data in the sample is made up of people within project
teams. Because of this, it would be expected that their responses would be similar. In
order to correct for this nested effect, we used a nested generalized linear model to test
the second-order discretized factor data. We discretized the data due to the severe nonnormality of several of the predictors and the nonlinear relationships between the
variables. Additionally, the sample size limitations, limited testing each hypothesis with
an independent model, rather than testing the three moderating hypotheses
simultaneously.
After dropping the influential cases described above, the sample reduced to 326
cases. After this, any project team for which only one case remained was also dropped.

146

The final analysis was performed using 305 responses, and controlled for nesting within
56 teams.
Testing the Research Model
We proposed that the agile methods would positively impact project success, and
that the dimensions of uncertainty would moderate the impact of agile method use on
project success. These hypotheses were as follows:
H1:

The extent of agile methodology use will positively impact project success

H2a: Structural complexity will negatively moderate the impact of agile
methodology use on project success
H2b: Technical complexity will negatively moderate the impact of agile
methodology use on project success
H2c: Environmental dynamism will positively moderate the impact of agile
methodology use on project success
In order to explore the direct impacts of the extent of Agile Method Use on
Project Success, and to explore the moderating effects of the dimensions of uncertainty,
three models were tested. As described above, the limited sample size at the team level
did not allow us to test all the moderation effects simultaneously. Each of the direct and
moderation effects was tested in an independent model, as indicated in Table 5.12.
Structural Complexity
The results of Model 1 are presented below. First, Table 5.13 provides goodness
of fit statistics for each of the dependent variables, with the extent of agile method use
as the independent variable, and structural complexity as the moderator. All models
have a non-significant chi-square test, indicating reasonable goodness of fit, and the

147

omnibus test indicates that the model parameters are not all zero. Table 5.13 also
presents the tests of model effects (Wald Chi-Square tests) by parameter, both at the
individual and nested levels. This test describes the effect of the variable as a whole. As
can be seen, the explanatory variables are significant in all five models.
Table 5.12. Generalized Linear Models Used to Test Moderating Effects of Uncertainty
Variable
Model 1a-e Model 2a-e Model 3a-e
Quality
Benefits
Budget Outcome
Time Outcome
Scope Outcome
Control Variables
Extent of Agile Method Use
Structural Complexity
Agile * Structural Complexity
Technical Complexity
Agile * Technical Complexity
Dynamism
Agile * Dynamism
Nested Effects

A
B
C
D
E
Yes
Yes
Yes
Yes

A
B
C
D
E
Yes
Yes

A
B
C
D
E
Yes
Yes

Yes
Yes

Yes

Yes

Yes
Yes
Yes

Table 5.14 presents the tests of model effects by parameter level, as compared
with the reference level of each variable (=5). The parameter estimates indicate that the
relationship between the extent of agile method use and structural complexity and the
outcome variables is significant at almost all levels. Further the parameters indicate that
the relationships between the variables are likely to be nonlinear, as the trends across
levels are not of a consistent direction and magnitude. Even with this nonlinear
structure, and even while some levels are insignificant, Table 5.19 indicates that each
variable is significant as a whole.

148

Table 5.13. Goodness of Fit of Model 3 the Moderation Effects of Structural Complexity.
Quality
Benefits
Budget
Time
Model Fit
Outcome
Outcome
Model Chi-Square
6.198
5.087
7.804
18.670
(df, sig.)
(10, .620)
(10, .509)
(10, .780)
(10, 1.867)
Model Omnibus Test
1096.959
1211.773
1310.769
958.286
(df, sig.)
(294, .000)
(294, .000)
(294, .000)
(294, .000)

Scope
Outcome
7.241
(10, .724)
1202.672
(294, .000)

Population Effects (Wald Chi-Square Test)
Intercept
24.613**
Agile
320.720**
Structural Complexity
46.303**
Agile * SC
73.993**
Management Support
54.791**
Project Length
35.993**
Organization Size
61.706**
Org. Agile Experience
35.362**
Respondent Experience
66.037**
Respondent Agile Exp.
23.668**

14.040**
168.532**
247.533**
240.409**
170.755**
9.887**
104.574**
40.211**
.903
44.618**

10.505**
217.513**
133.946**
112.217**
94.613**
31.349**
50.678**
40.322**
28.178**
71.738**

65.916**
58.828**
66.576**
35.457**
32.740**
1.593
.025
10.268**
44.494**
15.699**

159.387**
198.608**
295.777**
103.381**
197.756**
13.408**
41.892**
114.721**
24.271**
31.295**

Nested Effects (Wald Chi-Square Test)
Agile
475.706**
Structural Complexity
793.760**
Agile * SC
125.267**
Management Support
344.725**
Respondent Experience
53.717**
Respondent Agile Exp.
126.562**

586.369**
915.211**
81.211**
211.007**
175.248**
266.256**

854.574**
1470.651**
41.658**
623.352**
46.716**
252.030**

211.290**
310.789**
49.599**
204.445**
42.157**
61.091**

754.558**
617.564**
107.967**
633.424**
156.380**
292.425**

149

Table 5.14. Parameter Estimates for Model 3, the Moderation Effects of Structural Complexity.
Scope
Quality
Benefits Budget Outcome Time Outcome
Outcome
Intercept
5.976**
1.0896**
-1.886
-4.875*
-13.491**
Agile=1
-21.820**
4.1921**
19.447**
13.340
44.661**
Agile=2
-8.916**
1.2445**
5.140**
10.734**
17.381**
Agile=3
-19.636**
1.9745**
1.216
21.870**
15.536**
Agile=4
-3.531**
.7041**
-.436
4.096**
7.058**
Agile=5
0
0
0
0
0
Structural Complexity=1
5.716**
1.7592**
3.237
3.759
14.783**
Structural Complexity=2
8.494
5.2091**
-21.760**
-4.912
-41.290**
Structural Complexity=3
15.649**
3.8085**
-18.194**
-9.048
-35.551**
Structural Complexity=4
18.266**
3.9590**
-18.400**
-10.347
-40.676**
Structural Complexity=5
0
0
0
0
0
Management Support=1
-.474
-2.111**
1.858**
-1.047
1.566**
Management Support=2
-1.077*
2.721**
-.880
-1.113
2.111**
Management Support=3
-.991**
-2.862**
-.761**
-1.463**
2.075**
Management Support=4
-.813*
-6.353**
-.893*
-3.126**
.059
Management Support=5
0
0
0
0
0
Project Length
-.172**
.082**
.180**
.063
.114**
Organization Size
.161**
.189**
.163**
-.006
-.143**
Org. Agile Experience
.508**
-.491**
-.609**
-.475**
-.990**
Respondent Experience
.012
.629**
.136**
.264**
.512**
Respondent Agile Exp.
-1.302**
-1.198**
.712**
.933**
1.639**

150

Next the moderation effect of structural complexity on the impact of the extent of
agile method use on project success was evaluated. Because the moderation is tested
using discretized data, the moderation effect is reported by combination of levels from
each variable. The breakdown of cases by category is presented in Table 5.15.
Table 5.15. Cases by Category – Structural Complexity
Agile
=1
=2
=3
=4
Structural Complexity=1
0
1
2
0
Structural Complexity=2
4
14
12
6
Structural Complexity=3
13
59
63
72
Structural Complexity=4
0
8
13
11
Structural Complexity=5
0
0
4
3
Total
17
82
94
92

=5
1
1
10
8
0
20

Total
4
37
217
40
7
305

Remember that the hypothesis for the moderating impacts of structural
complexity was in the negative direction, but the negative impacts of project
management outcomes indicate increases from baseline estimates. This means that, for
the project management outcome variables, a positive beta coefficient indicates a
negative (as hypothesized) moderation effect. For this reason, the summary of
interaction terms is presented using the terms of hypothesized vs. reversed direction of
effect, rather than as positive or negative. Table 5.16 presents a summary of the
directions of interaction effects.
Table 5.16. Summary of Interaction Effects - Structural Complexity
Model
Reversed
Not
Hypothesized
Direction
Significant
Direction
Quality
4
3
6
Organizational Benefits
2
1
10
Budget Outcome
1
2
10
Time Outcome
3
8
2
Scope Outcome
1
1
11
Total, (%)
11 (17%)
15 (23%)
39 (60%)

151

Table 5.17 provides support for the negative moderating impact of structural
complexity on project success. At each level of Structural Complexity, the anticipated
effect is usually in the direction predicted. Interestingly, those impacts that are reversed
from the predicted hypothesis all appear at the lowest level of agile method use or the
lowest level of structural complexity. As shown in Table 5.16, these groups had the
fewest number of cases, as few as 1. Because of this, caution should be used when
interpreting differences regarding these groups.
The interaction terms indicate the predicted difference in effect from the
reference categories. As can be seen from the table, interactions are significant at all
levels for the Quality and Organization Benefits models. Fewer of the interactions are
significant for the project management success indicators. Given the observation
discussed above of the nonlinear relationships between both the IV and the moderators
with the outcome variables, the fact that the interactions are significant but are, at times,
in a direction opposite to that hypothesized is not surprising.

152

Table 5.17. Nested Model Interaction Coefficients – Structural Complexity
Quality
Agile
=1
=2
=3
=4
=5
Structural Complexity=1
a
23.615** 12.222**
a
0
Structural Complexity=2
6.743**
-2.741
-12.883*
-27.019**
0
Structural Complexity=3
1.185** -12.189** -18.063** -14.751**
0
Structural Complexity=4
a
3.200
-3.173
-14.749**
0
Structural Complexity=5
0
a
0
0
0
Org Benefits
Structural Complexity=1
Structural Complexity=2
Structural Complexity=3
Structural Complexity=4
Structural Complexity=5

=1
a
22.967**
1.367**
a
0

=2
-6.605**
-24.937**
-44.658**
-28.441**
a

Budget Outcome
Structural Complexity=1
Structural Complexity=2
Structural Complexity=3
Structural Complexity=4
Structural Complexity=5

Agile
=3
9.489**
-12.873*
-25.011**
-3.946
0

=1
a
11.391**
4.982**
a
0

=2
-15.507**
16.330**
11.277*
11.593
a

Agile
=3
-.735
30.182**
22.891**
21.276**
0

=1
a
10.247**
4.601**
a
0

=2
-28.845**
-4.066
-1.148
-11.030
a

Agile
=3
-9.535**
11.880
9.892
-3.302
0

Time Outcome

=4
a
-35.317**
-25.796**
-24.529**
0

=4
a
31.328**
20.745**
19.316**
0

=5
0
0
0
0
0

=5
0
0
0
0
0

=4
a
21.852*
9.178
8.304
0

=5
0
0
0
0
0

Agile
=1
=2
=3
=4
Structural Complexity=1
a
-28.663**
-.084
a
Structural Complexity=2
28.004** 27.690**
50.471**
45.697**
Structural Complexity=3
6.886** 18.625**
36.084**
34.392**
Structural Complexity=4
a
20.828**
39.883**
36.140**
Structural Complexity=5
0
a
0
0
a: no observations, 0: reference group,* p<.05, ** p<.01

=5
0
0
0
0
0

Structural Complexity=1
Structural Complexity=2
Structural Complexity=3
Structural Complexity=4
Structural Complexity=5
Scope Outcome

153

Summary of results: Hypothesis 2a
Structural Complexity will negatively moderate the impacts of Agile Method Use
on Project Success.
The moderating effects of structural complexity on agile method use were found
to be significant on all components of project success. The relationship between
structural complexity and project success is bi-modal. Because of this nonlinearity, it
was likely that mixed results would be observed when testing for the moderating effects
of structural complexity. Even so, it was found that 60% of the tested combinations of
level interactions were significantly positive and, of those combinations that were
significant, 78% of them were positive in direction. The presence of such a high
proportion of positive effects in the presence of such strong nonlinearity lends strong
support for Hypothesis 2a.
Technical Complexity
Model 2 was utilized to investigate the negative moderation effects of technical
complexity on the extent of agile method use. The method and interpretation used for
Model 2 and Model 3 was identical to that for Model 1.
Table 5.18 provides goodness of fit statistics and tests of model effects for each
of the Model 4 instances. As in the case of Model 1, all instances of Model 2 have a
non-significant chi-square test, indicating reasonable goodness of fit, and the omnibus
test indicates that the model parameters are not all zero. The tests of model fit indicate
that the explanatory variables still maintain significance for all of the model instances,
although control variables are not significant on the project management success
models.

154

Table 5.19 presents the parameter estimates and significance tests for the direct
effects of agile methodologies and technical complexity.

155

Table 5.18. Goodness of Fit of Model 4. Moderation Effects of Technical Complexity.
Quality
Benefits
Budget
Time
Model Fit
Outcome
Outcome
Model Chi-Square
14.287
14.446
35.698
36.551
(df, sig.)
(19, .757)
(19.760)
(19, 1.879)
(19, 1.924)
Model Omnibus Test
842.249
893.450
847.032
753.388
(df, sig.)
(285, .000)
(285,
(285, .000)
(285, .000)
.000)
Population Effects (Wald Chi-Square Test)
Intercept
12.078**
.017
19.560**
74.054**
Agile
76.579**
10.828*
21.152**
19.595**
Technical Complexity
60.519**
87.206**
35.210**
13.989**
Agile * TC
19.328**
46.226**
7.501
8.698*
Management Support
57.910**
69.760**
30.019**
15.877**
Project Length
24.128**
4.010*
13.536**
5.358
Organization Size
25.241**
39.154**
.749
10.170*
Org. Agile Experience
48.455**
15.321**
.053
28.891**
Respondent Experience
20.898**
12.615**
.406
23.053**
Respondent Agile Exp.
4.142*
37.548**
1.046
1.798
Nested Effects (Wald Chi-Square Test)
Agile
544.492**
Technical Complexity
215.142**
Agile * TC
66.118**
Management Support
302.494**
Respondent Experience
252.087**
Respondent Agile Exp.
168.169**

Scope
Outcome
15.333
(19, .807)
973.834
(285, .000)

129.852**
20.304**
20.595**
14.422**
53.707**
23.809**
.781
6.831**
.135
.365

374.396**
182.395**

221.027**
103.238**

189.664**
105.983**

507.173**
235.176**

371.655**
112.135**
145.091**

247.582**
45.023**
67.655**

141.362**
31.082**
35.391**

298.996**
140.598**
65.502**

156

Table 5.19. Parameter Estimates for Model 4, the Moderation Effects of Technical Complexity.
Scope
Quality
Benefits Budget Outcome Time Outcome
Outcome
Intercept
6.093**
-.862
3.382
7.906**
1.071
Agile=1
-11.202**
-5.857**
1.582
-2.487
-.110
Agile=2
-10.388** -12.245**
2.379
-10.201**
-3.007
Agile=3
-4.462*
4.638*
5.014
1.194
5.558**
Agile=4
-10.149** -10.392**
5.237
-4.540
-3.306
Agile=5
0
0
0
0
0
Technical Complexity=1
-4.767*
5.845**
1.026
-.106
5.605**
Technical Complexity=2
-.024
.847
-.931
.997
.867
Technical Complexity=3
4.001**
6.967**
-3.458
.734
.209
Technical Complexity=4
1.144**
.371
-1.575
-1.295**
-.883**
Technical Complexity=5
0
0
0
0
0
Management Support=1
-.020
2.232**
-2.343
-.010
1.357**
Management Support=2
-.366
.293
-2.560
-2.004*
.070
Management Support=3
.044
.205
-1.781
-.657
.505**
Management Support=4
-.464
-3.308**
-1.255
-3.125**
-1.701
Management Support=5
0
0
0a
0
0
Project Length
.171**
-.070*
.202
.129*
-.176**
Organization Size
.148**
.185**
.040
.150**
.027
Org. Agile Experience
-.836**
-.473**
-.044
-1.032**
-.325**
Respondent Experience
.249**
.614**
.016
.406**
.436**
Respondent Agile Exp.
-.298**
-.055
-.074
-.235**
.005

157

Table 5.20 lists the breakdown of cases by category for technical complexity. In
comparing this breakdown to Table 5.15, the cases are more evenly dispersed,
although there are still three cells with no cases.
Table 5.20. Cases by Category – Technical Complexity
Agile
=1
=2
=3
=4
Technical Complexity=1
0
3
9
4
Technical Complexity=2
7
27
27
14
Technical Complexity=3
6
38
36
23
Technical Complexity=4
4
13
18
38
Technical Complexity=5
0
1
4
13
Total
17
82
94
92

=5
0
5
12
2
1
20

Total
16
80
115
75
19
305

Table 5.21 lists the summary of directionality of interaction effects for Technical
Complexity. Only 16% of the interactions at the discrete levels are in the hypothesized
direction, while 70% of the interactions are not significant.
Table 5.21. Summary of Interaction Effects - Technical Complexity
Model
Reversed
Not
Hypothesized
Direction
Significant
Direction
Quality
2
6
2
Organizational Benefits
1
3
6
Budget Outcome
1
9
0
Time Outcome
3
7
0
Scope Outcome
0
10
0
Total, (%)
7 (14%)
35 (70%)
8 (16%)

Table 5.22 presents the specific interaction results by model and by discrete
category. The results presented in Table 5.22 provide little support for the hypothesis of
a negative moderation effect of technical complexity.

158

Table 5.22. Nested Model Interaction Coefficients – Technical Complexity
Quality
Agile
=1
=2
=3
=4
=5
Technical Complexity =1
a
9.655**
2.262
0
a
Technical Complexity =2
5.211**
1.829
-2.931
-1.298
0
Technical Complexity =3
0
-1.452
-4.624**
-.811
0
Technical Complexity =4
0
0
-2.920**
0
0
Technical Complexity =5
a
0
0
0
0
Org Benefits
Technical Complexity =1
Technical Complexity =2
Technical Complexity =3
Technical Complexity =4
Technical Complexity =5

=2
.738
.522
-1.564
0
0

Agile
=3
-17.143**
-8.073**
-12.167**
-8.007**
0

=4
0
-2.738*
-3.277**
0
0

=5
a
0
0
0
0

=1
a
3.061
0
0
a

=2
-.827
-.760
.147
0
0

Agile
=3
-9.532*
-2.483
-.797
-.263
0

=4
0
-.490
-1.046
0
0

=5
a
0
0
0
0

=1
a
-1.984
0
0
a

=2
5.819
-1.990
1.390
0
0

Agile
=3
-5.262
-7.656**
-4.859**
-2.448
0

=4
0
-6.787**
-3.211
0
0

=5
a
0
0
0
0

=1
a
5.444**
0
0
a

Budget Outcome
Technical Complexity =1
Technical Complexity =2
Technical Complexity =3
Technical Complexity =4
Technical Complexity =5
Time Outcome
Technical Complexity =1
Technical Complexity =2
Technical Complexity =3
Technical Complexity =4
Technical Complexity =5
Scope Outcome

Agile
=1
=2
=3
=4
Technical Complexity =1
a
-4.102
-10.602
0
Technical Complexity =2
.516
-5.010
-4.687
-3.654
Technical Complexity =3
0
-1.503
-2.910
.038
Technical Complexity =4
0
0
-2.723
0
Technical Complexity =5
a
0
0
0
a: no observations, 0: reference group,* p<.05, ** p<.01

159

=5
a
0
0
0
0

Summary of results: Hypothesis 2b
Technical Complexity will negatively moderate the impact of Agile Method Use
on Project Success
We find partial support for the relationship between technical complexity and
project success. This relationship is an inverse u shaped curve. This means that when
technical complexity is very low or is very high, the relationship with project success is
negative, but as the level of technical complexity approaches the mean, the relationship
becomes positive. Because of this nonlinearity, mixed results of hypothesis testing were
observed.
Of the tested combinations, only 16% of the tests were significant and in the
hypothesized direction. Of the remaining combinations, 70% were found to be
insignificant, and 14% were significant in the opposite direction as was predicted.
These results indicate that support for hypothesis is weak for most dimensions of
project success, with the exception of organizational benefits, for which 86% of the
significant interaction combinations were positive.
Dynamism
Model 3 was used to complete the tests of the moderation effects of uncertainty
on project success. In contrast to hypotheses 2a and 2b, the moderation effects of
dynamism were predicted to be positive. Table 5.29 provides goodness of fit statistics
for each of the dependent variables, with the extent of as the independent variable, and
dynamism as the moderator. Again, all models have a non-significant chi-square test,
indicating reasonable goodness of fit, and the omnibus test indicates that the model
parameters are not all zero.

160

Table 5.23 presents the tests of model effects by parameter, both at the
individual and nested levels. As before, this test evaluates the effect of the variable as a
whole. Table 5.24 lists the breakdown of cases by category for technical complexity.
The hypothesis for the moderating impacts of dynamism was in the positive
direction, but the positive impacts of project management outcomes indicate decreases
from baseline estimates. This means that, for the project management outcome
variables, a negative beta coefficient indicates a positive (as hypothesized) moderation
effect. Table 5.32 lists the summary of directionality of interaction effects for Technical
Complexity. About 46% of the interactions at the discrete levels are in the hypothesized
direction, while 14% of the interactions in the opposite direction from the hypothesis. Of
those combinations that had significant effects, 70% were in the hypothesized direction.
As the spline transformation regression showed a nearly linear positive relationship
between dynamism and quality, these results are not unexpected.

161

Table 5.23. Goodness of Fit of Model 5, the Moderation Effects of Dynamism.
Quality
Benefits
Budget
Time
Model Fit
Outcome
Outcome
Model Chi-Square
5.635
1.809
11.550
9.521
(df, sig.)
(10, .563)
(10, .181)
(10, 1.155)
(10, .952)
Model Omnibus Test
1126.022
1527.079
1191.213
1163.668
(df, sig.)
(294, .000)
(294, .000)
(294, .000)
(294, .000)

Scope
Outcome
4.474
(10, .447)
1349.540
(294, .000)

Population Effects (Wald Chi-Square Test)
Intercept
61.566**
Agile
263.890**
Dynamism
93.424**
Agile * Dynamism
225.243**
Management Support
231.467**
Project Length
69.305**
Organization Size
11.180**
Org. Agile Experience
27.006**
Respondent Experience
74.014**
Respondent Agile Exp.
46.127**
Nested Effects (Wald Chi-Square Test)
Agile
310.885**
Technical Complexity
1078.846**
Agile * Dynamism
Management Support
435.109**
Respondent Experience
313.788**
Respondent Agile Exp.
146.601**

96.116**
909.798**
906.354**
877.381**
533.074**
92.106**
141.432**
28.762**
117.899**
89.873**

2.013
37.195**
102.049**
198.021**
65.890**
48.395**
1.813
1.116
5.520*
1.248

2.641
133.621**
45.870**
163.838**
54.667**
29.170**
52.786**
24.686**
.054
2.017

16.498**
224.609**
87.185**
390.676**
61.484**
.469
.562
15.859**
31.824**
33.239**

1007.804**
1999.906**

393.487**
847.630**

521.989**
711.497**

378.755**
881.098**

1525.429**
765.330**
803.866**

269.694**
81.166**
206.406**

134.000**
64.763**
200.650**

169.585**
56.396**
54.435**

162

Table 5.24. Parameter Estimates for Model 5, the Moderation Effects of Dynamism.
Scope
Quality
Benefits Budget Outcome Time Outcome
Outcome
Intercept
-6.164**
-6.966**
-1.543
1.692*
4.065**
Agile=1
-54.951**
-.005
-42.326**
50.735**
-12.813
Agile=2
-27.354**
16.122**
-3.138
44.376**
19.658**
Agile=3
-32.324**
6.084
-6.905
33.635**
13.831*
Agile=4
14.718**
20.000**
5.631**
-1.135**
7.201**
Agile=5
0
0
0
0
0
Dynamism=1
-36.051**
10.187*
-19.233
50.690**
6.399
Dynamism=2
-37.727**
9.730*
-18.817
52.111**
6.886
Dynamism=3
-103.620**
13.230
-68.089*
141.424**
-2.820
Dynamism=4
-36.440**
10.826*
-15.383
53.968**
11.225
Dynamism=5
0
0
0
0
0
Management Support=1
39.756**
-7.053
20.433*
-51.588**
-5.439
Management Support=2
39.197**
-5.529
17.141
-49.102**
-11.897
Management Support=3
41.543**
-2.708
20.487*
-49.147**
-6.780
Management Support=4
45.150**
1.354
22.755*
-49.311**
-3.865
Management Support=5
0
0
0
0
0
Project Length
.198**
.129**
.237**
.167**
.014
Organization Size
.081**
-.164**
.047
.229**
-.016
Org. Agile Experience
-.393**
-.230**
.114
-.488**
.268**
Respondent Experience
-.204**
-.148**
-.170**
.125**
-.117**
Respondent Agile Exp.
.545**
-.055
.309
-1.066**
-.076

163

Table 5.25 presents a summary of the directions of interaction effects, in
comparison to the hypothesized direction.

Dynamism=1
Dynamism=2
Dynamism=3
Dynamism=4
Dynamism=5
Total

Table 5.25. Cases by Category – Dynamism
Agile
=1
=2
=3
=4
0
4
3
3
6
15
20
36
7
34
40
29
4
25
25
19
0
4
6
5
17
82
94
92

=5
1
5
9
4
1
20

Total
11
82
119
77
16
305

Table 5.26. Summary of Interaction Effects - Dynamism
Model
Reversed
Not
Hypothesized
Direction
Significant
Direction
Quality
2
0
12
Organizational Benefits
6
7
1
Budget Outcome
3
9
2
Time Outcome
2
0
12
Scope Outcome
1
8
5
Total, (%)
14 (20%)
24 (34%)
32 (46%)

Table 5.27 presents the specific interaction results by model and by discrete
category.

164

Table 5.27. Nested Model Interaction Coefficients – Dynamism
Quality
Agile
=1
=2
=3
=4
Dynamism=1
a
36.649**
32.325**
77.307**
Dynamism=2
1.517**
76.339**
68.349**
-10.770**
Dynamism=3
68.314**
98.264**
49.827**
56.322**
Dynamism=4
0
75.967**
36.845**
-9.852**
Dynamism=5
a
0
0
0
Org Benefits

=5
0
0
0
0
0

=2
-4.533
-13.756
-13.690
-14.736
0

Agile
=3
-5.175
-18.097*
-14.856*
-11.179*
0

=4
-14.531
-16.786*
-14.755*
-13.789*
0

=5
0
0
0
0
0

=1
a
-1.628**
52.554**
0
a

=2
11.729
35.825
56.160*
31.363
0

Agile
=3
8.536
23.270
26.294
15.366
0

=4
36.797
-2.706
45.795**
-3.592*
0

=5
0
0
0
0
0

=1
a
3.145**
-82.196**
0a
a

=2
-48.990**
-94.285**
-136.561**
-100.860**
0

Agile
=3
-32.628**
-82.919**
-80.712**
-50.372**
0

=4
-102.459**
.871**
-86.013**
-1.011**
0

=5
0
0
0
0
0

Agile
=1
=2
=3
=4
a
-13.475*
-13.864**
-14.206
.569*
-17.094
-27.014**
-7.341**
11.896
-6.990
-7.521
.711
0
-19.957
-8.287
-6.246**
a
0
0
0
a: no observations, 0: reference group,* p<.05, ** p<.01

=5
0
0
0
0
0

=1
a
.421*
-2.124
0
a

Dynamism=1
Dynamism=2
Dynamism=3
Dynamism=4
Dynamism=5
Budget Outcome
Dynamism=1
Dynamism=2
Dynamism=3
Dynamism=4
Dynamism=5
Time Outcome
Dynamism=1
Dynamism=2
Dynamism=3
Dynamism=4
Dynamism=5
Scope Outcome
Dynamism=1
Dynamism=2
Dynamism=3
Dynamism=4
Dynamism=5

The results in Table 5.27 show that the nature of the moderating relationship
between the extent of agile method use, dynamism, and quality is as hypothesized.
165

Summary of results: Hypothesis 2c
Environmental Dynamism will positively moderate the impact of Agile Method
Use on Project Success
Significant results in the hypothesized direction for four of the five dimensions of
project success. For all of components of project success, the presence of higher levels
of dynamism is related to higher impacts of agile method use. While this relationship is
not linear, at all levels of agile method use, the as dynamism rises, the model indicates
the expectation of more positive expected results for all components of project success.
Summary of results: Hypothesis 1
The Extent of Agile Method Use will positively impact Project Success
We tested the direct effects of the extent of agile method use and the impacts of
agile method use in the presence of three moderating variables. In the majority of the
models, the results indicated that the impact of the extent of agile method use on project
success is positive, although the effects exhibited heterogeneity. On the Quality factor,
agile impact use consistently demonstrated positive direct effects. In most models,
higher levels of agile method use were associated with higher quality, but these effects
were not linear. On the organizational benefits factor, the results were mixed. While
agile method use was consistently significant, its effects were extremely nonlinear, with
some effects being positive, and some negative, even in the same model.
The project management outcome variables also exhibited direct effects from
agile method use, but these results were also mixed. While in the direct effects models,
the impacts of agile method use were not significant, but in the moderated models,
significant direct effects were found. In the structural complexity tests, all effects were

166

significant and in the hypothesized direction. In the technical complexity model, the
effect on budget outcome was non significant, the effect on time outcome was mixed in
direction, and the effect on scope outcome was as hypothesized. Finally, in the
dynamism model, the impact on budget outcome was significant, but reversed, while the
effect on time outcome and scope outcome was as hypothesized.
In summary, the observed impacts of agile methodologies on project success do,
on the whole, support Hypothesis 1. However, the nonlinear relationship between agile
method use and project success suggests that the hypothesis is most likely overly
simplistic, and requires further development.
Table 5.28 presents a summary of the results of all of the hypothesis tests.

167

Table 5.28. Summary of Hypothesis Findings
Budget
Hypothesis
Product Quality Org. Benefits Outcome
H1: Agile Method Use will
Full
Mixed
Mixed
positively impact project success

Time
Outcome
Full

Scope
Outcome
Full

H2a: Structural Complexity will
negatively moderate the impact of
Agile Method Use

Full

Full

Full

No

Full

H2b: Technical Complexity will
negatively moderate the impact of
Agile Method Use

No

Full

No

No

No

H2c: Dynamism will negatively
moderate the impact of Agile
Method Use

Full

Full

Full

Full

Full

168

Chapter Summary
Significant, nonlinear relationships were found between the theorized explanatory
variables and outcome variables. This nonlinearity contributed to mixed results of the
hypothesis testing.
Support was found for Hypothesis 1, which stated: “The extent of agile method
use will positively impact project success”. Support was found for Hypothesis 2a, which
predicted the negative moderation effects of Structural Complexity on the impact of
Agile Method Use on Project Success. However, only partial support was found for the
impact of Hypothesis 2b, with the hypothesized negative moderating effects of technical
complexity only being significant for the organizational benefits component of project
success.
Finally, strong support was found for Hypothesis 2c, which stated that dynamism
would positively moderate the impact of project success. A significant positive
moderating effect was found across all measures of project success.
The results of these hypothesized tests are summarized by project success
dimension in Table 5.28

169

Chapter 6: Discussion and Implications of the Results
Introduction
This chapter discusses the findings, implications, and limitations of this research
study. First, the empirical results will be compared with the conceptual models that were
proposed and motivated the study. Next, the limitations of the study will be discussed.
The chapter will conclude with implications and directions for future research.
The Effectiveness of Agile Methodologies in Software Development
Agile development methodologies have emerged in the past decade as a
significant new way of organizing and executing software development projects. Agile
practitioners have made numerous claims about the impacts of agile methodologies
including associated improvements in team efficiency and performance, higher software
quality, and greater organizational benefits. While recent research on agile method use
have shown significant results, a number of the normative claims of agile method
practitioners have not yet been tested.
Software development practices have long been hypothesized to be contingent
upon environmental and other factors (Barki et al. 2001). This means that the
performance of a software development method is the product of the interactions
between the characteristics of the environment with the characteristics of the
development method in use. Agile practitioners have argued that because the software
development process is inherently uncertain and empirical, traditional software
methodologies have been unable to successfully develop successful plans of execution,
even with very high initial investment (Highsmith 2002; Schwaber and Beedle 2002).
Instead, agile practitioners have argued that the software development process is much

170

more like the new product development process than an engineering process
(Highsmith 2000). Because of this, agile methodologies have adopted practices,
techniques and supporting technologies that are said to enhance delivery success.
However, as explained in Chapter 1, many of agile methodologies’ normative claims
have not been tested, and the methodologies themselves display only partial
homogeneity.
In this dissertation we explained that, while the methodologies have diverse
enabling practices, and different points of focus, they share a philosophy of the
importance of feedback. Whereas previous studies have studied teams performing a
particular agile method, we conceptualized agile method use at a level that is
measurable across the multiple agile methodologies in practice today, specifically the
various feedback processes that are based upon the shared philosophy of the agile
manifesto.
The results show that the extent of agile method use positively impacts project
success. Significant effects were found on the proposed dimensions of project success,
including project management metrics, product quality, and perceived organizational
impacts. However, the impact of the extent of agile method use on the project success
dimensions was found to be nonlinear. While the effects of agile use were generally
positive, the slope of effect is greatly reduced near the mean, and then increases
sharply again. This result is intuitive when considering the recommendations of agile
practitioners. Agile methodologies advocates argue that the method should be adopted
piecemeal, and that a team’s greatest pain points should be addressed first. It is
therefore likely that there are fast gains from agile methodologies early in the use cycle,

171

as those practices that are most likely to increase performance are implemented first.
However, agile practitioners also stress that, due to the mutually reinforcing nature of
the practices of agile, the full performance impact of the method is only achieved when
the majority of the practices are put into place. However, the interaction of these
practices is complex, and because teams must understand the manner in which to
apply them through learning. This means is that after observing the immediate early
performance impacts, teams will continue to add agile practices with the anticipation of
continued performance gains. However, it should be anticipated that since the “low
hanging” fruit has already been addressed, the impacts of these new practices would be
lowered. However, once a more complete network of reinforcing practices is put into
place, the impact is likely to increase again as the synergistic nature of the practices
come into play. This phenomenon requires additional study.
The Moderating Effects of Uncertainty
Hypotheses 2a, 2b, and 2c predicted that three dimensions of uncertainty would
moderate the impact of agile methodologies on project success. Uncertainty has been
theorized to consist of multiple dimensions of complexity and dynamism. Technical
complexity and structural complexity have been explained in the past to be barrier to
successful software development. Structural complexity makes the accessing and
interpretation of environmental feedback more difficult. Teams within complex structural
environments have greater difficulty in establishing goals, interpreting the needs of
multiple stakeholders, and in building shared mental models (Xia and Lee 2005).
Consistent with prior theory, structural complexity was found to negatively moderate the
impact of the extent of agile method use on project success. Structural complexity is

172

likely to manifest itself through negative impacts on human processes, such as
developing shared mental models and processing feedback. Thus, it’s presence is
intuitively associated with a reduction in the ability of an adaptive team to react to salient
cues from the environment.
The moderating impacts of technical complexity on the effects of the extent of
agile method use on project success were, in general, insignificant. The impacts of
technical complexity were hypothesized to be negative. Significant impacts were found
only on the organizational benefits component of project success. This may indicate that
the processes of agile methodologies, such as technical feedback and supporting
technologies, potentially mitigate the impacts of technical complexity. Further
implications of this finding are discussed below.
The third dimension of uncertainty that was hypothesized as moderating the
impacts of the extent of agile method use is dynamism. Agile methodologies are
designed to manage dynamic environments. Therefore, the use of agile methodologies
in a highly dynamic environment is indicative of fit between the environment and the
development method in use. We hypothesized that the extent of dynamism in an
environment would positively moderate the impacts of agile method use on project
success. We proposed a positive interaction between these constructs due to the fact
that agile methodologies’ feedback processes are designed to recognize and react to
environmental change. In the presence of high dynamism, these feedback processes
are crucial to project success. However, in the presence of low dynamism, the feedback
processes are additional overhead that provide little benefit (Eisenhardt and Tabrizi
1995).

173

The empirical evidence provided by this study confirms some of the practitioner
claims of the impacts of the extent of agile method use on project success. However,
the observed relationships between the explanatory and outcome variables was found
to be complex. However, this research study has several important limitations.
Limitations of this Study
Because the conceptualization of a generalized extent of agile method use
construct was new, and because the impacts of agile on various dimensions of project
success had not yet been tested, this research study was exploratory in nature. The first
limitation is that the theorized impacts of agile method use were motivated from the
perspective of team adaptability and organizational learning. These theories are
inherently reinforcing and cyclical in nature, but the research study was performed using
a cross-sectional design that indicates the state of the theoretical model at a particular
point in time. Learning and adaptation are cyclical processes that occur over time, but
cross-sectional

research

methodologies

do

not

allow

researchers

to

make

determinations about rates of change over time, magnitude and changes in magnitude
of change over time, and trends. Further, the nature of a survey instrument does not
allow for the understanding of specific local variation, but rather simply identifying that
the local variation exists. Even so, we believe that cross-sectional research
methodologies were appropriate to answer the research question being studied, and to
explore the relationship between the use of agile methodologies and project success.
Second, the sample in this study was made up only of agile teams, and was as
such as purposive sample. This created intentional sampling bias by including in the
research only those teams that self-identified as using agile methodologies. However,

174

this bias was necessary as the concepts and constructs of the agile team’s process and
practices would not necessarily be interpretable by non-agile practitioners. This limits
the generalizability of the study, and does not indicate the applicability of the study’s
model or conclusions outside of agile teams.
Third, although the study sought to include the constructs, dimensions, and
variables that were indicated by the review of the literature, the final survey that we
implemented included less than half of the variables that were initially identified. Even
so, the survey took many respondents more than 30 minutes to complete. While
respondent participation was encouraged by a small reward, and while respondent
fatigue was reduced via a variety of survey design techniques, we could not include all
of the variables that the focus groups identified as being salient in agile development.
Specifically the nature of interaction with the customer and the frequency and quality of
interaction of the team and stakeholders were dimensions that were identified as
extremely important by two of the five focus groups. To the extent that the model
neglected to measure and consider these variables, we must be cautious in applying
the findings.
Implications for Practice
Even today, practitioners look for “silver bullets” with regards to software
development methodologies. However, as the prior literature and this study show, the
impacts of the use of methodologies such as those designated as agile methodologies
is contingent upon fit with the environment. As has been explained, software
development methodologies have the greatest impact when there is high fit between
environmental factors and the practices of the method. We found this to be the case.

175

This research reinforces prior research that indicates that structural complexity
negatively impacts teams in general. This research shows a significant negative impact
on the effects of the extent of agile method use when teams operate in the presence of
higher levels of structural complexity. IS practitioners should take care when adopting
agile teams to minimize the structural complexity that impacts the teams. When
possible, teams should be collocated, and the number of stakeholders and reporting
relationships should be minimized.
Also this research found that the impact of technical complexity on success is
extremely nonlinear. At both low and high levels of technical complexity, negative
interactions between agile use and success were found. This is somewhat contradictory
to previous assertions in the practitioner literature that indicate that agile methodologies
are best suited for less technically complex projects, and that as technical complexity
rises, the need for structure and up front planning rise (Boehm and Turner 2004;
Cockburn 2001). One reason for this result may be that if a project is extremely simple,
the perceptions of project success and impacts are likely to be low. Conversely,
extremely high technical complexity may overwhelm the technical feedback processes
of many agile teams. If high system complexity impedes the effect of the technical
feedback, the impacts of technical feedback would be mitigated. For instance, if high
levels of integration with external systems are necessary, an agile development team
may be unable to write tests that fully test the system.
The research suggests that the single most significant component of uncertainty
that would indicate a good fit for the use of agile methodologies is the presence of high
levels of dynamism. In these environments, the ability to sense and respond to change

176

has been shown to positively impact performance. However, in environments with lower
levels of dynamism, the impacts of agile method adoption and use are not expected to
be as high. Organizations should consider the extent to which the team must be able to
respond to change as a key indicator signals potentially high performance for agile
method use.
In summary, the results of this research indicate that the organizations should
generally expect the impacts of agile method use on project success to be generally
positive, and to be the most positive in the presence of low structural complexity and
high dynamism.
Implications for Future Research
Studying the impacts of agile methodologies remains a rich area for future
information systems research. While this research study has been primarily exploratory
in nature, it provides a number of new opportunities for future research. This study acts
as a useful new data point for the research stream, and provides evidence that more
development and empirical research is required.
First, further empirical and theoretical elaboration of the general constructs of
agile method use is needed. The most obvious need is to adopt a longitudinal research
design that will allow the researcher to investigate the impacts of agile methodologies
over time. This would allow the field to better understand whether the impacts of agile
and the moderating effects of environmental uncertainty and complexity are consistent
over time. Further, the nonlinear nature of the data obtained in this study may imply the
presence of a recursive process. A longitudinal or experimental research design would

177

be able to potentially detect a cyclical or reinforcing effect of agile method use over
time.
Second, while we theorized a greater nomological network for the constructs that
comprise agile methodologies. While we utilized this theorized nomological network to
motivate the hypotheses, the network itself was not tested in this study. Understanding
this network would be an interesting extension of this research. Further the nomological
network proposed may be generalizable to all software development methodologies.
This concept of a nomological network of the components of software development
methodologies has not previously been tested. Testing this concept both within the area
of agile methodologies, as well as other software development methodologies would be
a fruitful addition to the IS field.
Third, the relationships between the constructs are contrary to prior research that
propose linear negative relationships between complexity and project success, and
linear positive relationships between the extent of agile method use and project
success. These findings indicate the need for the development of new theory.
Significant theoretical development that recognizes nonlinearity has occurred in the
organizational behavior literature. However, IS theories have, with few exceptions,
proposed linear relationships, and studies have designed measurement and tests that
assume linear relationships. These results add to previous calls for the development of
theories of nonlinear effects (Venkatesh and Goyal 2010).

178

Conclusion
Agile methodologies proceed from the philosophy that the software development
effort is one that must consistently and efficiently deal with change. These
methodologies prescribe complex networks of processes, practices and procedures, as
well as supporting technologies that together are proposed to positively impact the
delivery of software in uncertain environments. While these methodologies are
extremely heterogeneous in their defined practices, practices that are common to all of
the methodologies are those that are designed to elicit feedback cues from the
environment. It is only through processing these environmental cues that software
development teams can develop adaptability and agility (Burke et al. 2006; Eisenhardt
and Tabrizi 1995; Kozlowski 1998). This research focused specifically on the impacts of
agile method use, as indicated by the level of use of feedback processes.
Numerous claims have been made about the impacts of agile method use on
project success, and the environmental conditions for which the methodologies are
most well-suited. Agile method use is expected to have the highest impacts in
environments that are highly dynamic. Further, agile method use is most effect when
structural complexity barriers are low. Because of the methodologies’ focus on human
processes of feedback, sense and response high levels of structural complexity
impedes the ability of these methodologies to impact the project fully.
The results of this research highlight the importance of feedback in managing the
uncertain nature of the software development process. However, the findings indicate a
significant need for further research on the impacts of agile methodologies, and
environmental conditions on the successful delivery of software development projects.

179

APPENDICES

180

APPENDIX A: QUESTIONNARIE ITEMS
Items for Structural Complexity
1. Approximately how many people worked on the project, IN TOTAL?
Include everyone who worked on the core project team, even if they only
participated in part of the project. (Scale: Don’t Know, 1- 6-10, 11-15,
-5,
16-20, 21-25, Over 25)
2. How many companies did members of the project team work for? (i.e., Did
everyone work for the same company, or were people paid by multiple
companies?) (Scale: Don’t Know, 1, 2, 3, 4, 5, 6+)
3. How many departments were represented on the project tea
team? (Scale:
Don’t Know, 1, 2, 3, 4, 5, 6+)
4. How many different groups of users provided requirements for the project?
(Scale: Don’t Know, 1, 2, 3, 4, 5, 6+)
5. Please indicate how close or how far away the members of the team were
located when working:
Scale:
The entire team worked in the same room/area
he
The entire team worked in the same building or campus
The entire team worked in multiple locations in the same city
The entire team worked in the same state
state/province
The entire team worked in the same country
Some of the team worked in a different country
The entire team was spread across different countries

181

Items for Project Criticality
Scale: 5-point Likert Scale + Don’t know (Strongly Disagree – Strongly Agree)
1. The project deliverable was part of the strategic plan of the organization.
2. The project deliverable was required in order to respond to competition.
3. The project deliverable was required in order to respond to government
requirements.
4. If the project deliverable was delivered late, it would have significant
financial impact to the organization.

Items for Technical Complexity
Scale: 5-point Likert Scale + Don’t know (Not Complex At All – Extremely Complex)
1. How technically complex was the organization's system environment?
2. How technically complex was the project system?

Scale: 5-point Likert Scale + Don’t know (Strongly Disagree – Strongly Agree)
3. The system involved multiple software environments.
4. The system involved multiple technology platforms.
5. The system involved a lot of integration with other, external systems.
6. The project's integration with external systems was complex.

182

Items for Dynamism
Scale: 5-point Likert Scale + Don’t know (Strongly Disagree – Strongly Agree)
Requirements Dynamism
1. Project requirements fluctuated quite a bit in early phases of the project.
2. Project requirements fluctuated quite a bit in later phases of the project.
3. Project requirements identified at the beginning of the project were quite
different from those toward the end.
Technical Dynamism
1. The IT infrastructure that the project depended on changed a lot during
the project.
2. Software development tools that the project depended on changed a lot
during the project.
External Dynamism
1. The project was associated with changes in the users’ business
processes.
2. The project was associated with changes in the users’ organizational
structure.
3. The system users’ information needs changed a lot during the project.

183

Items for Agile Method Use
Scale: 5-point Likert Scale + Don’t know (Strongly Disagree – Strongly Agree)
Reduced Up Front Planning
1. The team spent less than 10% of the total project timeline on up-front
planning (planning that occurred before ANY coding began).
2. At the beginning of the project, the team tried to make only the decisions
that were necessary for coding to begin.
Technical Feedback
1. Every programmer was responsible for writing automated tests for the
code he or she wrote.
2. Programmers ran a set of automated tests until they all ran successfully
before checking in changes.
Environmental Feedback
1. At the beginning of each development cycle, the team and business
owners agreed on what would be delivered during the development cycle.
2. The team had a short meeting every day to discuss what was going on
that day.
3. The team had a review/verification meeting with stakeholders to
demonstrate when software features were complete.
4. On a regular basis, the team reflected on previous work, and looked for
ways to improve team performance.
5. At the beginning of each development cycle, the team and business
owners agreed on what would be delivered during the development cycle.

184

Items for Product Quality
Scale: 5-point Likert Scale + Don’t know (Definitely Not True – Definitely True)
Prompt: Please give your opinion about the following statements about the project
system:
Quality
1. In terms of system quality, I would rate [the project system] highly.
2. Overall, [the project system] is of high quality.
3. I would give the quality of [the project system] a high rating.
Usefulness
1. [The project system] improves users abilities to perform their tasks.
2. The project system] allows users to get work done more effectively.
3. [The project system] allows users to get their tasks done more quickly.
Completeness
1. [The project system] provides users with a complete set features and information.
2. [The project system] is a comprehensive solution.
3. [The project system] provides users with all needed information to do their tasks
in the system.
Reliability
1. [The project system] operates reliably.
2. The company can rely on [the project system].
3. The operation of [the project system] is dependable.
Accuracy
1. [The project system] properly performs the tasks it was intended to perform.
2. There are few errors or bugs in [the project system].
3. The information provided by [the project system] is accurate.
Suitability
1. The [the project system] delivered the desired project outcome.
2. The [the project system] accomplishes what was needed.
3. The [the project system] does what was it is supposed to.

185

Items for Organizational Benefits
Scale: 5-point Likert Scale + Don’t know (Definitely Not True – Definitely True)
Prompt: Please give your opinion about these statements related to YOUR
PERCEPTION of satisfaction with the project system:
1. Because of this project, our organization can better realize its goals.
2. This project helped our organization to perform better.
3. Our organization is more competitive because of this project.

Items for Project Management Outcomes
Project Budget Outcome
Scale: 5-point Likert Scale + Don’t know (Very Much Lower – Very Much Higher)
1. In comparison to the initial budget estimate the final budget was:
Project Time Outcome
Scale: 5-point Likert Scale + Don’t know (Very Much Shorter – Very Much Longer)
1. In comparison to the initial time estimate the final project duration was:
Project Scope Outcome
Scale: 5-point Likert Scale + Don’t know (Very Much Smaller – Very Much Larger)

186

APPENDIX B: SUPPLEMENTAL DATA TABLES
Team ID
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38

Table B.1. Response Breakdown by Team and Response Type
Stakeholder
IT Mgmt Team
Team Total
0
0
2
2
1
0
0
1
0
0
1
1
1
0
4
5
0
0
1
1
4
0
12
16
0
0
2
2
1
0
5
6
1
1
2
4
0
1
5
6
0
0
4
4
0
0
6
6
1
0
5
6
1
0
2
3
0
1
1
2
1
0
3
4
0
0
5
5
1
0
7
8
0
0
3
3
0
0
4
4
1
0
4
5
0
0
6
6
0
0
4
4
1
0
5
6
1
1
9
11
1
0
10
11
0
0
6
6
2
0
5
7
0
0
1
1
0
1
2
3
0
1
2
3
0
0
1
1
0
0
1
1
0
0
5
5
2
2
5
9
1
3
3
7
2
2
5
9
0
1
3
4
187

Team ID
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78

Table B.1. Response Breakdown by Team/Role (cont’d)
Stakeholder
IT Mgmt Team
1
0
2
0
1
0
0
1
0
0
0
15
0
1
4
2
1
11
3
1
16
0
0
1
0
0
1
0
1
3
0
0
1
1
0
3
1
1
2
1
0
2
0
0
1
0
0
1
1
0
0
0
0
1
0
0
1
0
0
1
0
0
1
0
0
1
0
0
1
1
0
0
1
0
0
0
0
1
0
0
1
0
1
2
0
1
2
0
1
5
1
1
4
0
1
3
0
1
2
1
1
8
0
1
8
1
1
7
2
0
10
0
1
5
1
0
5
0
0
4
188

Team Total
3
1
1
15
5
14
20
1
1
4
1
4
4
3
1
1
1
1
1
1
1
1
1
1
1
1
1
3
3
6
6
4
3
10
9
9
12
6
6
4

Table B.1. Response Breakdown by Team/Role (cont’d)
79
0
0
3
80
0
0
1
81
1
0
0
82
0
0
5
83
0
0
4
Grand Total
42
31
300
Average Responses per Team

189

3
1
1
5
4
373
4.49

REFERENCES

190

REFERENCES
2009. "Chaos Report," The Standish Group International, Inc., West Yarmouth, MA.
Abrahamsson, P., Salo, O., Ronkainen, J., and Warsta, J. 2002. "Agile Software
Development Methods," VTT Technical Research Centre of Finland.
Agerfalk, P., and Fitzgerald, B. 2006. "Old Petunias in New Bowls?," Communications
of the ACM (49), pp. 10-27.
Ambler, S.W. 2009. "The Agile Scaling Model (Asm): Adapting Agile Methods for
Complex Environments." IBM Corporation.
Ancona, D. 1990. "Outward Bound: Strategies for Team Survival in an Organization,"
Academy of management journal (33:2), pp. 334-365.
Ancona, D., and Caldwell, D. 1992. "Bridging the Boundary: External Activity and
Performance in Organizational Teams," Administrative Science Quarterly (37:4),
pp. 634-665.
Anderson, D.J. 2004. Agile Management for Software Engineering: Applying the Theory
of Constraints for Business Results. Prentice Hall.
Argyris, C., and Schoen, D. 1978. Organizational Learning: A Theory of Action
Perspective. Reading, MA: Addison-Wesley.
Atkinson, R. 1999. "Project Management: Cost, Time and Quality, Two Best Guesses
and a Phenomenon, Its Time to Accept Other Success Criteria," International
Journal of Project Management (17:6), pp. 337-342.
Austin, R., and Devin, L. 2009. "Weighing the Benefits and Costs of Flexibility in Making
Software: Toward a Contingency Theory of the Determinants of Development
Process Design," Information Systems Research (20:3), pp. 462-477.
Avison, D., and Fitzgerald, G. 1998. Information Systems Development: Methodologies,
Techniques and Tools. Oxford: Blackwell Scientific Publications.

191

Avison, D., and Taylor, V. 1997. "Information Systems Development Methodologies: A
Classification According to Problem Situation," Journal of Information Technology
(12:1), pp. 73-81.
Baccarini, D. 1996. "The Concept of Project Complexity--a Review," International
Journal of Project Management (14:4), pp. 201-204.
Barki, H., Rivard, S., and Talbot, J. 2001. "An Integrative Contingency Model of
Software Project Risk Management," Journal of Management Information
Systems (17:4), pp. 37-69.
Beck, K. 1999. Extreme Programming Explained: Embrace Change, (First ed.).
Addison-Wesley Professional.
Beck, K., and Andres, C. 2004. Extreme Programming Explained: Embrace Change,
(Second ed.). Addison-Wesley Professional.
Beck, K., Beedle, M., van Bennekum, A., Cockburn, A., Cunningham, W., Fowler, M.,
Grenning, J., Highsmith, J., Hunt, A., Jeffries, R., Kern, J., Marick, B., Martin,
R.C., Mellor, S., Schwaber, K., Sutherland, J., and Thomas, D. 2001. "Manifesto
for Agile Software Development, Accessed July 1, 2011." Retrieved July 1,
2011, 2001, from http://www.agilemanifesto.org
Bhattacharya, S., Krishnan, V., and Mahajan, V. 1998. "Managing New Product
Definition in Highly Dynamic Environments," Management Science), pp. 50-64.
Boehm, B. 1988. "A Spiral Model of Software Development and Enhancement,"
Computer (21:5), pp. 61-72.
Boehm, B. 2002. "Get Ready for Agile Methods, with Care," Computer), pp. 64-69.
Boehm, B., and Turner, R. 2004. Balancing Agility and Discipline, a Guide for the
Perplexed. Boston, MA: Addison Wesley.
Bollen, K.A., and Jackman, R.W. 1990. "Regression Diagnostics: An Expository
Treatment of Outliers and Influential Cases," in Modern Methods of Data
Analysis, J. Fox and J. Long (eds.). Newbury Park, CA: Sage, pp. 257-291).

192

Bovaird, J. 2007. "Multilevel Structural Equation Models for Contextual Factors," in
Modeling Contextual Effects in Longitudinal Studies, T. Little, J. Bovaird and N.
Card (eds.). Mahwah, NJ: Lawrence Erlbaum Associates, pp. 149-182.
Brooks, F. 1987. "No Silver Bullet: Essence and Accidents of Software Engineering,"
IEEE Computer (20:4), pp. 10-19.
Brooks, F. 1995. The Mythical Man-Month (Anniversary Ed.). Addison-Wesley Longman
Publishing Co., Inc. Boston, MA, USA.
Burke, C., Stagl, K., Salas, E., Pierce, L., and Kendall, D. 2006. "Understanding Team
Adaptation: A Conceptual Analysis and Model," Journal of Applied Psychology
(91:6), pp. 1189-1207.
Burns, T., and Stalker, G.M. 1994. The Management of Innovation. Oxford University
Press, USA.
Cao, L., Mohan, K., Xu, P., and Ramesh, B. 2009. "A Framework for Adapting Agile
Development Methodologies," European Journal of Information Systems (18:4),
pp. 332-343.
Coad, P., Lefebvre, E., and DeLuca, J. 1999. Java Modeling in Color with Uml. Prentice
Hall.
Cockburn, A. 2001. Agile Software Development. Boston, MA: Addison-Wesley.
Cockburn, A., and Williams, L. 2001. "The Costs and Benefits of Pair Programming."
Citeseer, pp. 223-248.
Conboy, K. 2009. "Agility from First Principles: Reconstructing the Concept of Agility in
Information Systems Development," Information Systems Research (20:3),
September 1, 2009, pp. 329-354.
Conboy, K., and Fitzgerald, B. 2004. "Toward a Conceptual Framework of Agile
Methods," Extreme Programming and Agile Methods-XP/Agile Universe 2004),
pp. 105-116.

193

Consortium, D. 2002-2011. "Dsdm Public Version 4.2."
http://www.dsdm.org/version4/2/public/

Retrieved 1/20, 2011, from

Converse, J.M., and Presser, S. 1986. Survey Questions: Handcrafting the
Standardized Questionnaire. Sage Publications, Inc.
Creswell, J.W. 2003. Research Design: Qualitative, Quantitative, and Mixed Methods
Approaches. Sage Publications.
Curtis, B., Krasner, H., and Iscoe, N. 1988. "A Field Study of the Software Design
Process for Large Systems," Communications of the ACM (31:11), pp. 12681287.
DeLone, W.H., and McLean, E.R. 1992. "Information Systems Success: The Quest for
the Dependent Variable," Information Systems Research (3:1), pp. 60-95.
Delone, W.H., and McLean, E.R. 2003. "The Delone and Mclean Model of Information
Systems Success: A Ten-Year Update," Journal of Management Information
Systems (19:4), pp. 9-30.
DeSanctis, G., and Poole, M.S. 1994. "Capturing the Complexity in Advanced
Technology Use: Adaptive Structuration Theory," Organization Science (5:2), pp.
121-147.
Dickinson, T.L., and McIntyre, R.M. 1997. "A Conceptual Framework for Teamwork
Measurement," in Team Performance Assessment and Measurement: Theory,
Methods, and Applications. pp. 19-43.
Donaldson, L. 2001. The Contingency Theory of Organizations. Thousand Oaks, CA:
Sage Publications, Inc.
Duncan, R.B. 1972. "Characteristics of Organizational Environments and Perceived
Environmental Uncertainty," Administrative Science Quarterly (17:3), pp. 313327.
Dyba, T., and Dingsoyr, T. 2008. "Empirical Studies of Agile Software Development: A
Systematic Review," Information and Software Technology (50:9-10), pp. 833859.

194

Eisenhardt, K., and Tabrizi, B.N. 1995. "Accelerating Adaptive Processes: Product
Innovation in the Global Computer Industry," Administrative Science Quarterly
(40:1).
Endsley, M.R. 1995. "Toward a Theory of Situation Awareness in Dynamic Systems,"
Human Factors: The Journal of the Human Factors and Ergonomics Society
(37:1), pp. 32-64.
Estublier, J. 2000. "Software Configuration Management: A Roadmap," ACM, pp. 279289.
Eveleens, J.L., and Verhoef, C. 2009. "The Rise and Fall of the Chaos Report Figures,"
IEEE software), pp. 30-36.
Fowler, M. 1999. Refactoring. Improving the Design of Exiting Code. Reading, MA:
Addison Wesley Longman.
Fowler, M. 2006. "Continuous Integration." Retrieved March 19th, 2011, 2011, from
http://www.martinfowler.com/articles/continuousIntegration.html
Fowler, M., and Highsmith, J. 2001. "The Agile Manifesto," in: Software Development.
pp. 28-35.
Fruhling, A., and Vreede, G. 2006. "Field Experiences with Extreme Programming:
Developing an Emergency Response System," Journal of Management
Information Systems (22:4), pp. 39-68.
Gersick, C. 1988. "Time and Transition in Work Teams: Toward a New Model of Group
Development," Academy of management journal (31:1), pp. 9-41.
Gersick, C.J.G., and Hackman, J.R. 1990. "Habitual Routines in Task-Performing
Groups," Organizational behavior and human decision processes (47:1), pp. 6597.
Group, S. 2009. "Chaos Report, 2009," The Standish Group International, Inc., West
Yarmouth, MA.

195

Harris, M., Collins, R., and Hevner, A. 2009. "Control of Flexible Software Development
under Uncertainty," Information Systems Research (20:3), pp. 400-419.
Highsmith, J. 2000. Adaptive Software Development. Dorset House New York NY.
Highsmith, J. 2002. Agile Software Development Ecosystems. Boston: Addison Wesley.
Hinds, P.J., and Mortensen, M. 2005. "Understanding Conflict in Geographically
Distributed Teams: The Moderating Effects of Shared Identity, Shared Context,
and Spontaneous Communication," Organization Science), pp. 290-307.
Hirschheim, R., Klein, H., and Lyytinen, K. 1995. Information Systems Development and
Data Modeling: Conceptual and Philosophical Foundations. Cambridge Univ Pr.
Ilgen, D., Hollenbeck, J., Johnson, M., and Jundt, D. 2005. "Teams in Organizations:
From Input-Process-Output Models to Imoi Models," Annual Review of
Psychology (56), pp. 517-543.
Indeed.com. 2011. "Agile, Scrum, Extreme Programming, Test Driven Job Trends."
Retrieved
May
6,
2011,
2011,
from
http://www.indeed.com/jobtrends?q=agile%2C+scrum%2C+%22extreme+progra
mming%22%2C+%22test+driven%22&l=
Keil, M. 1995. "Pulling the Plug: Software Project Management and the Problem of
Project Escalation," Mis Quarterly (19:4), pp. 421-447.
Keil, M., Cule, P.E., Lyytinen, K., and Schmidt, R.C. 1998. "A Framework for Identifying
Software Project Risks," Communications of the ACM (41:11), pp. 76-83.
Keil, M., Mann, J., and Rai, A. 2000. "Why Software Projects Escalate: An Empirical
Analysis and Test of Four Theoretical Models," Mis Quarterly), pp. 631-664.
Kiesler, S., and Cummings, J.N. 2002. "What Do We Know About Proximity and
Distance in Work Groups? A Legacy of Research," Distributed work (1), pp. 5780.
Kirsch, L.J. 1996. "The Management of Complex Tasks in Organizations: Controlling the
Systems Development Process," Organization Science), pp. 1-21.
196

Kirsch, L.J., Sambamurthy, V., Ko, D.G., and Purvis, R.L. 2002. "Controlling Information
Systems Development Projects: The View from the Client," Management
Science), pp. 484-498.
Kozlowski, S. 1998. "Training and Developing Adaptive Teams: Theory, Principles, and
Research,").
Kozlowski, S., Gully, S., Nason, E., and Smith, E. 1999. "Developing Adaptive Teams: A
Theory of Compilation and Performance across Levels and Time," Pulakos
(Eds.), The changing nature of performance: Implications for staffing, motivation,
and development), p. 240ñ292.
Kuchta, D. 2001. "Use of Fuzzy Numbers in Project Risk (Criticality) Assessment,"
International Journal of Project Management (19:5), pp. 305-310.
Lave, J. 1991. Cognition in Practice. Cambridge Univ. Pr.
Lawrence, P.R., and Lorsch, J.W. 1967. Organization and Enviroment.
Lee, G., and Xia, W. 2010. "Toward Agile: An Integrated Analysis of Quantitative and
Qualitative Field Data on Software Development Agility," MIS Quarterly (34:1),
pp. 87-114.
Lindvall, M., Muthig, D., Dagnino, A., Wallin, C., Stupperich, M., Kiefer, D., and May, J.
2004. "Agile Software Development in Large Organizations," Computer), pp. 2634.
Louis, M.R., and Sutton, R.I. 1991. "Switching Cognitive Gears: From Habits of Mind to
Active Thinking," Human Relations (44:1), p. 55.
Lyytinen, K., and Hirschheim, R. 1988. "Information Systems Failuresóa Survey and
Classification of the Empirical Literature," Oxford University Press, Inc., pp. 257309.
Lyytinen, K., and Rose, G. 2006. "Information System Development Agility as
Organizational Learning," European Journal of Information Systems (15:2), pp.
183-199.

197

Mabert, V.A., Soni, A., and Venkataramanan, M. 2003. "The Impact of Organization
Size on Enterprise Resource Planning (Erp) Implementations in the Us
Manufacturing Sector," Omega (31:3), pp. 235-246.
MacCormack, A. 2001. "How Internet Companies Build Software," MIT Sloan
Management Review (42:2), p. 75ñ84.
MacCormack, A., and Verganti, R. 2003. "Managing the Sources of Uncertainty:
Matching Process and Context in Software Development," Journal of Product
Innovation Management (20:3), pp. 217-232.
Mangalaraj, G., Mahapatra, R.K., and Nerur, S. 2009. "Acceptance of Software Process
Innovations - the Case of Extreme Programming," European Journal of
Information Systems (18:4), pp. 344-354.
Mann, C., and Maurer, F. 2005. "A Case Study on the Impact of Scrum on Overtime and
Customer Satisfaction," Agile Conference, 2005. Proceedings, pp. 70-79.
Marakas, G.M., and Elam, J.J. 1998. "Semantic Structuring in Analyst Acquisition and
Representation of Facts in Requirements Analysis," Information Systems
Research (9:1), p. 37.
Marks, M.A., and Panzer, F.J. 2004. "The Influence of Team Monitoring on Team
Processes and Performance," Human Performance (17:1), pp. 25-41.
Maruping, L., Venkatesh, V., and Agarwal, R. 2009a. "A Control Theory Perspective on
Agile Methodology Use and Changing User Requirements," Information Systems
Research (20:3), pp. 377-399.
Maruping, L.M., Zhang, X., and Venkatesh, V. 2009b. "Role of Collective Ownership
and Coding Standards in Coordinating Expertise in Software Project Teams,"
European Journal of Information Systems (18:4), pp. 355-371.
McAvoy, J., and Butler, T. 2009. "The Role of Project Management in Ineffective
Decision Making within Agile Software Development Projects," European Journal
of Information Systems (18:4), pp. 372-383.
McConnell, S. 1996. Rapid Development: Taming Wild Software Schedules. Microsoft
Press Redmond, WA, USA.
198

McGrath, J. 1964. Social Psychology: A Brief Introduction. Holt, Rinehart and Winston.
McGrath, J., Arrow, H., and Berdahl, J. 2000. "The Study of Groups: Past, Present, and
Future," Personality and Social Psychology Review (4:1), pp. 95-105.
McIntyre, R.M., and Salas, E. 1995. "Measuring and Managing for Team Performance:
Emerging Principles from Complex Environments," in Team Effectiveness and
Decision Making in Organizations. pp. 9-45.
McKeen, J.D., Guimaraes, T., and Wetherbe, J.C. 1994. "The Relationship between
User Participation and User Satisfaction: An Investigation of Four Contingency
Factors," MIS Quarterly), pp. 427-451.
Meyer, M.H., and Curley, K.F. 1991. "An Applied Framework for Classifying the
Complexity of Knowledge-Based Systems," Mis Quarterly), pp. 455-472.
Miller, D. 1992. "Environmental Fit Versus Internal Fit," Organization Science (3:2), pp.
159-178.
Moran, D. 2010. "Top 10 Reasons to Use Agile Development."
2011, from http://www.devx.com/enterprise/Article/44619

Retrieved 3/9/2011,

Nerur, S., Mahapatra, R.K., and Mangalaraj, G. 2005. "Challenges of Migrating to Agile
Methodologies," Communications of the ACM (48:5), pp. 72-78.
Nidumolu, S. 1995. "The Effect of Coordination and Uncertainty on Software Project
Performance: Residual Performance Risk as an Intervening Variable,"
Information Systems Research (6:3), p. 191.
Ogunnaike, B.A., and Ray, W.H. 1994. Process Dynamics, Modeling, and Control.
Oxford University Press New York:.
Okhuysen, G.A. 2001. "Structuring Change: Familiarity and Formal Interventions in
Problem-Solving Groups," The Academy of Management Journal (44:4), pp. 794808.

199

Okhuysen, G.A., and Waller, M.J. 2002. "Focusing on Midpoint Transitions: An Analysis
of Boundary Conditions," The Academy of Management Journal (45:5), pp. 10561065.
One, V. 2010. "5th Annual State of Agile Development Survey Final Summary Report."
Orlikowski, W.J. 1992. "The Duality of Technology: Rethinking the Concept of
Technology in Organizations," Organization Science (3:3), pp. 398-427.
Parrish, A., Smith, R., Hale, D., and Hale, J. 2004. "A Field Study of Developer Pairs:
Productivity Impacts and Implications," Software, IEEE (21:5), pp. 76-79.
Paulk, M., Weber, C., Curtis, B., and Chrissis, M. 1995. The Capability Maturity Model:
Guidelines for Improving the Software Process. Indianapolis, IN: Addison
Wesley.
Petter, S., DeLone, W., and McLean, E. 2008. "Measuring Information Systems
Success: Models, Dimensions, Measures, and Interrelationships," European
Journal of Information Systems (17:3), pp. 236-263.
Pfeffer, J., and Salancik, G.R. 2003. The External Control of Organizations: A Resource
Dependence Perspective. Stanford, CA: Stanford University Press.
Phillips, J.J., Bothell, T.W., and Snead, G.L. 2002. The Project Management Scorecard:
Measuring the Success of Project Management Solutions. ButterworthHeinemann.
Pich, M.T., Loch, C.H., and De Meyer, A. 2002. "On Uncertainty, Ambiguity, and
Complexity in Project Management," Management Science), pp. 1008-1023.
PMBOK. 2000. A Guide to the Project Management Body of Knowledge (Pmbok
Guide). Project Management Institute.
Port, D., and Bui, T. 2009. "Simulating Mixed Agile and Plan-Based Requirements
Prioritization Strategies: Proof-of-Concept and Practical Implications," European
Journal of Information Systems (18:4), pp. 317-331.

200

Purvis, R.L., Sambamurthy, V., and Zmud, R.W. 2001. "The Assimilation of Knowledge
Platforms in Organizations: An Empirical Investigation," Organization Science),
pp. 117-135.
Rai, A., Lang, S.S., and Welker, R.B. 2002. "Assessing the Validity of Is Success
Models: An Empirical Test and Theoretical Analysis," Information Systems
Research (13:1), p. 50.
Ribbers, P.M.A., and Schoo, K.C. 2002. "Program Management and Complexity of Erp
Implementations," Engineering Management Journal (14:2), pp. 45-52.
Salas, E., Cannon-Bowers, J., Fiore, S., and Stout, R. 2001. "Cue-Recognition Training
to Enhance Team Situation Awareness," New trends in cooperative activities:
understanding system dynamics in complex environments. Santa Monica, CA:
Human Factors and Ergonomics Society), p. 169ñ190.
Salas, E., Prince, C., Baker, D.P., and Shrestha, L. 1995. "Situation Awareness in Team
Performance: Implications for Measurement and Training," Human Factors: The
Journal of the Human Factors and Ergonomics Society (37:1), pp. 123-136.
Salo, O., and Abrahamsson, P. 2008. "Agile Methods in European Embedded Software
Development Organisations: A Survey on the Actual Use and Usefulness of
Extreme Programming and Scrum," Software, IET (2:1), pp. 58-64.
Schmidt, R., Lyytinen, K., Keil, M., and Cule, P. 2001. "Identifying Software Project
Risks: An International Delphi Study," Journal of Management Information
Systems (17:4), pp. 5-36.
Schwaber, C., and Fichera, R. 2005. "Corporate It Leads the Second Wave of Agile
Adoption," Cambridge, MA.
Schwaber, K. 1996. "Controlled Chaos: Living on the Edge," American Programmer (9),
pp. 10-16.
Schwaber, K., and Beedle, M. 2002. Agile Software Development with Scrum. Prentice
Hall Upper Saddle River, NJ.
Seddon, P.B. 1997. "A Respecification and Extension of the Delone and Mclean Model
of Is Success," Information Systems Research (8:3), pp. 240-253.
201

Shenhar, A.J., and Dvir, D. 1996. "Toward a Typological Theory of Project
Management," Research Policy (25:4), pp. 607-632.
Stephens, M., and Rosenberg, D. 2003. Extreme Programming Refactored: The Case
against Xp. New York, NY: Apress.
Tait, P., and Vessey, I. 1988. "The Effect of User Involvement on System Success: A
Contingency Approach," Mis Quarterly), pp. 91-108.
Tesluk, P.E., and Mathieu, J.E. 1999. "Overcoming Roadblocks to Effectiveness:
Incorporating Management of Performance Barriers into Models of Work Group
Effectiveness," Journal of Applied Psychology (84:2), p. 200.
Thompson, J. 2003. Organizations in Action: Social Science Bases of Administrative
Theory. Transaction Pub.
Turner, J.R., and Cochrane, R.A. 1993. "Goals-and-Methods Matrix: Coping with
Projects with Ill Defined Goals and/or Methods of Achieving Them," International
Journal of Project Management (11:2), pp. 93-102.
Venkatesh, V., and Goyal, S. 2010. "Expectation Disconfirmation and Technology
Adoption: Polynomial Modeling and Response Surface Analysis," Mis Quarterly
(34:2), pp. 281-303.
Venkatraman, N. 1989. "The Concept of Fit in Strategy Research: Toward Verbal and
Statistical Correspondence," Academy of Management Review), pp. 423-444.
Wallace, L., and Keil, M. 2004. "Software Project Risks and Their Effect on Outcomes,"
Communications of the ACM (47:4), pp. 68-73.
Waller, M.J. 1999. "The Timing of Adaptive Group Responses to Nonroutine Events,"
The Academy of Management Journal (42:2), pp. 127-137.
Weiss, D.J. 1986. "The Discriminating Power of Ordinal Data," Journal of Social
Behavior and Personality (1:38), pp. 1-389.
West, D., and Grant, T. 2010. "Agile Development: Mainstream Adoption Has Changed
Agility," Forrester Research.
202

Williams, L., Kessler, R.R., Cunningham, W., and Jeffries, R. 2000. "Strengthening the
Case for Pair Programming," Software, IEEE (17:4), pp. 19-25.
Williams, T.M. 1999. "The Need for New Paradigms for Complex Projects," International
Journal of Project Management (17:5), pp. 269-273.
Wixom, B.H., and Todd, P.A. 2005. "A Theoretical Integration of User Satisfaction and
Technology Acceptance," Information Systems Research (16:1), pp. 85-102.
Wynekoop, J., and Russo, N. 1995. "Systems Development Methodologies:
Unanswered Questions," Journal of Information Technology (10:2), pp. 65-73.
Xia, W., and Lee, G. 2005. "Complexity of Information Systems Development Projects:
Conceptualization and Measurement Development," Journal of Management
Information Systems (22:1), pp. 45-83.
Zmud, R.W., Anthony, W.P., and Stair Jr, R.M. 1993. "The Use of Mental Imagery to
Facilitate Information Identification in Requirements Analysis," Journal of
Management Information Systems (9:4), pp. 175-191.

203