Answer questions below.....
latnia2
Week8Article2_Toward_a_Diagnostic-Dialogic_M2.pdf
All rights reserved
INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.
In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.
Microform Edition © ProQuest LLC. All rights reserved. This work is protected against
unauthorized copying under Title 17, United States Code
ProQuest LLC. 789 East Eisenhower Parkway
P.O. Box 1346 Ann Arbor, MI 48106 - 1346
UMI 3536439
Published by ProQuest LLC (2013). Copyright in the Dissertation held by the Author.
UMI Number: 3536439
© Copyright William Bartlett Brock 2013
iii
Abstract
It has become axiomatic within the field of organization development that
organizations operate within increasingly dynamic, even volatile, environments and
must improve their ability to consciously and continuously develop, change, and
adapt (i.e., implement planned organizational change) in order to survive and thrive.
While the need for continual organizational change is axiomatic, the approach for
effecting planned organizational change is not. Based on organizational change
theory, accounting metrics, and systems theory, this study develops a diagnostic-
dialogic organizational change model as a resolution tool for three dilemmas
identified within the literature on organizational change: organizational change
approaches, a systemic perspective, and measurable change.
The diagnostic-dialogic model of organizational change is empirically tested through
a retrospective, longitudinal field study with the application of rigorous statistical
methods: survival analysis, generalized estimating equations, and logistic regression.
Empirically tested within the context of an organizational customer proposal process,
results indicate that a diagnostic-dialogic intervention provides superior results
compared with either a laissez-faire approach or a traditional, diagnostic intervention.
The diagnostic-dialogic intervention realized greater improvements in process
iv
completion time, in level of commentary, in degree of systemic commentary, in
degree of profitability alignment, and in the likelihood of customer acceptance.
v
Dedication
To Mary –
My wife, my partner for the past 30 years—by any metric, the love of my life!
To Mom –
What an incredible path of learning you forged for me!
vi
Acknowledgments
I am honored to acknowledge the help of so many in this endeavor called the Ph.D.
journey. To Mary, there are just not words enough to express the value of your
unwavering support through this process. In spite of a road so deeply marked by the
loss of sister and mother along the way, you never hesitated in cheering me to the
finish line!
To my exemplary committee—Dr. Ram Tenkasi, Dr. Bill Pasmore, Dr. Dick Woodman,
and Dr. Kala Visvanathan—a general acknowledgment is insufficient, so allow me to
thank each of you specifically.
Ram – your vision, your persistence, and your quiet demand for the highest
quality of scholarly research, combined with the highest level of practitioner
applicability, made for a truly epic journey. I am proud for this work to
represent that journey. Thank you!
Bill – your continuing quest for greater context pushed me well beyond my
comfort zone. The result is a work immeasurably richer than I would have
ever thought possible. Thank you!
Dick – you so deftly overcame my objections to adding that final hypothesis.
It seemed a great price at the time, but the price pales in significance to the
value added to WholeCo and the diagnostic-dialogic paradigm. Thank you!
vii
Kala – how patient you were during those many Saturday and Sunday
whiteboard sessions. Because of you, I had confidence to delve into the
statistics of epidemiology, and OD now boasts a higher level of rigor. Thank
you!
To Dr. Peter Sorensen, thank you for creating a world-class program and allowing me
to be a part of it. I can only hope to emulate your passion for and contribution to the
field of OD.
To Dr. Ron Fisher, you absolutely amaze me. Your generosity and willingness to help
so many along the way continually inspired me.
To our Quigley’s group – you know who you are – I cannot imagine how lonely this
journey would have been without you all. So many shared hardships and successes –
may our paths cross often.
This work would not be possible without the WholeCo organization and in particular
its CEO. Though he must remain nameless, his willingness to endure countless hours
of coding, debating, and proofreading will never be forgotten.
viii
Table of Contents
List of Figures ............................................................................................................... x
List of Tables ............................................................................................................... xi
Chapter 1: Introduction ................................................................................................. 1
Background of Study ................................................................................................ 3
Overview of Chapters ............................................................................................... 4
Chapter 2: Literature Review ........................................................................................ 6
Dilemma of Organizational Change Approaches ..................................................... 6
Synthesizing Diagnostic and Dialogic Approaches .............................................. 9
Diagnostic-Dialogic Construct............................................................................ 11
A Systemic Perspective........................................................................................... 12
The Dilemma of a Systemic Perspective ............................................................ 13
Accounting Metrics: Promoting a Systemic Perspective within the Diagnostic-
Dialogic Framework ........................................................................................... 15
Accounting Metrics as Cybernetic, Double-Loop Systems ................................ 20
Dilemma of Measurable Change ............................................................................ 22
Accounting Metrics as Change Measurement Instruments ................................ 23
Diagnostic-Dialogic Organizational Change Model ............................................... 24
Chapter 3: Methodology ............................................................................................. 26
Methodological Structure........................................................................................ 26
Field Study .......................................................................................................... 26
Longitudinal Study.............................................................................................. 28
Retrospective Study ............................................................................................ 29
Research Setting...................................................................................................... 29
The Company ...................................................................................................... 30
The Situation ....................................................................................................... 31
Customer Proposal Process ................................................................................. 32
Interventions ....................................................................................................... 37
Research Question .................................................................................................. 38
Quasi-Experimental Groups.................................................................................... 39
Laissez-faire ........................................................................................................ 39
Traditional ........................................................................................................... 39
Diagnostic-Dialogic ............................................................................................ 40
Hypotheses .............................................................................................................. 49
Hypothesis 1: Completion Time ......................................................................... 49
Hypothesis 2: Quantity of Commentary ............................................................. 50
Hypothesis 3: Quality of Commentary ............................................................... 51
Hypothesis 4: Profitability Alignment ................................................................ 58
Hypothesis 5: Customer Acceptance .................................................................. 60
Control Variables .................................................................................................... 61
General Control Variables .................................................................................. 61
Specific Control Variables .................................................................................. 62
ix
Data Collection ....................................................................................................... 65
Longitudinal Dataset ........................................................................................... 66
Testing Methods...................................................................................................... 67
Survival Analysis ................................................................................................ 67
Generalized Estimating Equations ...................................................................... 69
Logistic Regression ............................................................................................. 72
Summary ................................................................................................................. 73
Chapter 4: Data Analysis and Results ......................................................................... 75
Data Verification ..................................................................................................... 75
Descriptive Statistics and Correlations ................................................................... 76
Results of Hypotheses Testing ................................................................................ 79
Hypotheses 1a and 1b: Effect on Completion Time ........................................... 79
Hypotheses 2a and 2b: Effect on Quantity of Commentary ............................... 82
Hypotheses 3a and 3b: Effect on Quality of Commentary ................................. 85
Hypothesis 4: Effect on Profitability Alignment ................................................ 87
Hypotheses 5a and 5b: Customer Acceptance .................................................... 89
Summary of Findings .............................................................................................. 91
Chapter 5: Conclusion and Implications ..................................................................... 94
General Findings ..................................................................................................... 94
Completion Time .................................................................................................... 95
Quantity of Commentary ........................................................................................ 97
Quality of Commentary ........................................................................................ 100
Profitability Alignment ......................................................................................... 102
Customer Acceptance ........................................................................................... 103
Limitations ............................................................................................................ 105
Implications for Future Research .......................................................................... 106
Appendix A: GEE Distributions and Link Functions ............................................... 108
Appendix B: Disposition of Outlier Records ............................................................ 109
References ................................................................................................................. 110
x
List of Figures
Figure 1. Customer Proposal Form ............................................................................. 34
Figure 2. Customer Proposal Approval Form ............................................................. 35
Figure 3. Flow of Customer Proposal Process ............................................................ 36
Figure 4. Longitudinal Timeline ................................................................................. 38
Figure 5. Diagnostic-Dialogic, Accounting Metrics Tool .......................................... 43
Figure 6. Histogram of Completion Time................................................................... 80
xi
List of Tables
Table 1. Comparison of Diagnostic and Dialogic Change ........................................... 8
Table 2. Diagnostic-Dialogic Organizational Change Schema .................................. 25
Table 3. Traditional Intervention - Diagnostic Tool ................................................... 41
Table 4: Preliminary Coding Definitions for Quality of Commentary ....................... 55
Table 5: Final Coding Definitions for Quality of Commentary ................................. 56
Table 6. Comment Coding Matrix .............................................................................. 58
Table 7. WholeCo Customer Base .............................................................................. 63
Table 8. WholeCo Product Groups ............................................................................. 63
Table 9: Summary of Operationalized Variables........................................................ 64
Table 10: Stratified Record Set ................................................................................... 66
Table 11. Summary of Hypotheses and Testing Methods .......................................... 74
Table 12. Descriptive Statistics................................................................................... 77
Table 13. Correlations for Selected Variables ............................................................ 78
Table 14. AFT Models for Completion Time ............................................................. 81
Table 15. PA-GEE Models for Quantity of Commentary .......................................... 83
Table 16. PA-GEE Models for Quality of Commentary ............................................ 86
Table 17. PA-GEE Models for Profitability Alignment ............................................. 88
Table 18. LR Models for Customer Acceptance ........................................................ 90
Table 19. Results of Hypotheses Testing .................................................................... 92
Table 20. Results of AFT, PA-GEE and LR Testing .................................................. 93
Table 21. Mean Comments per Record by Company ................................................. 99
1
Chapter 1: Introduction It has become axiomatic within the field of organization development (OD) that
organizations operate within increasingly dynamic, even volatile, environments and
must improve their ability to consciously and continuously develop, change, and
adapt (i.e., implement planned organizational change) in order to survive and thrive.
But while the necessity of continual organizational change is axiomatic, the approach
for effecting planned organizational change is not.
The literature wrestles with the merits of traditional, diagnostic-based change
methods versus those of contemporary, dialogic-based change methods. Traditional,
diagnostic-based change methods are recognized as those that gather organizational
data for “diagnosing” organizational issues using normative models (Bushe &
Marshak, 2009; Cummings & Worley, 2009; Marshak & Grant, 2008); whereas
contemporary dialogic-based change methods view data as “generative” rather than
diagnostic (Bushe & Marshak, 2008, 2009; Marshak & Grant, 2008) and the purpose
of its collection as a “collective inquiry into the multiple [organizational] ‘realities’ in
order to generate new collective understandings and cognitive maps” (Bushe &
Marshak, 2008, pp. 10-11). This dichotomy of diagnostic and dialogic approaches
results in a dilemma of organizational change approaches.
Whether using a diagnostic or dialogic organizational change approach, OD scholars
and practitioners assert the necessity of promoting a systemic perspective within the
2
organization in order to avoid a cycle of unintended consequences (Burke, 2008b;
Carucci & Pasmore, 2002; Katz & Kahn, 1978; Mohrman Jr., Mohrman, Ledford Jr,
Cummings, & Lawler III, 1989; Mohrman, Mohrman, & Tenkasi, 1996; Senge, 1990;
Weisbord, 2004); however, the literature recognizes that achieving a systemic
perspective is difficult and uncommon (Ackoff, 2006; Argyris, 1990; Senge, 1990;
Sterman, 1994; Tenkasi, Mohrman, & Mohrman, 2007); thus creating a dilemma of
systemic perspective.
Concurrent with these dilemmas is the challenge for OD to provide measurable
evidence of the impact of organizational change interventions on organizational
performance. The need for a measureable link between organizational change and
organizational performance is clearly evident, but the response within the literature is
inadequate (Burke, 2008a; Pettigrew, Woodman, & Cameron, 2001; Woodman, 1989;
Woodman, Bingham, & Yuan, 2008), thus highlighting the dilemma of measureable
change.
How can OD approach resolution of these dilemmas of organizational change
approaches, a systemic perspective, and measurable change? This study proposes
resolution through a diagnostic-dialogic model of organizational change drawn from a
synthesis of organizational change theory with accounting metrics. It also seeks to
provide empirical evidence of the efficacy of this model through a longitudinal field
study testing the differential effects of three organizational change approaches:
Laissez-faire (baseline), Traditional (diagnostic), and Diagnostic-Dialogic.
3
Recognizing that organizational accounting metrics have the ability to create a
synthesis of diagnostic- and dialogic-based change, to promote a systemic
perspective, and to provide a method for rigorous measurement, I argue that the
diagnostic-dialogic intervention in this study should accomplish the following:
Faster organizational processes and decision making;
A greater level of dialogue within the organization;
A greater degree of systemic perspective within the organization;
Better alignment of organizational proposals with organizational goals;
Improved likelihood of success with organizational customers.
Background of Study
A retrospective, longitudinal field study was conducted from archived records of a
customer proposal process for the period from June 2008 through October 2011. The
field setting is a group of existing consumer products organizations with more than 50
years’ history, referenced in this study as WholeCo. The elements of the study, as
well as several elements of the WholeCo environment, are developed more fully in
Chapter 3; but a critical understanding is that the study occurs during a volatile,
rapidly changing period for WholeCo. The timeframe from June 2008 through
October 2011 is well recognized as a period of domestic and global economic
distress, including high unemployment, a depressed U.S. housing market, and bank
and business closures. As described in Chapter 3, many of these items directly
4
affected WholeCo and demonstrate that WholeCo operated in a volatile environment,
requiring integrated, measurable organizational change.
Overview of Chapters
Chapter 1 has provided a brief overview of the context of this research study along
with its proposal for a diagnostic-dialogic model, founded on accounting metrics to
synthesize diagnostic and dialogic organizational change approaches, to provide a
systemic perspective, and to provide a means for rigorous change measurement.
Chapter 2 establishes a dichotomic schema of organizational change approaches
through a brief review of the literature on the diagnostic and dialogic paradigms of
organizational change theory. It continues by establishing a new construct of
diagnostic-dialogic organizational change and introducing organizational accounting
metrics as a whole-system metaphor with the ability to enable diagnostic-dialogic
change in an integrative, and measurable, manner throughout the organizational
system.
Chapter 3 discusses the research methodology with a focus on the applicability of a
retrospective, longitudinal field study, the WholeCo research setting, and the
operationalization of quasi-experimental and outcome variables. Rigorous testing
methods are chosen and described as accelerated failure time survival analysis,
population-averaged generalized estimating equations, and logistic regression in order
5
to test five a priori hypotheses drawn from the literature and the proposed
characteristics of the diagnostic-dialogic model.
Chapter 4 reviews the data validation and verification process and presents a
summary of descriptive statistics. Each hypothesis is then tested separately. Results
are provided in summary tables along with a discussion of statistical significance and
a determination of whether the hypothesis is supported, partially supported, or
rejected.
Chapter 5 concludes the study with general findings and a discussion of the testing
results for each hypothesis and the implication of those results. This chapter
demonstrates the implications of a combined diagnostic-dialogic organizational
change approach, founded on accounting metrics, to integrate both diagnostic- and
dialogic- based organizational change, promote a systemic perspective, and provide
rigorous change measurement. The chapter ends with a summary of the research
limitations along with implications for areas of future research.
6
Chapter 2: Literature Review In this chapter, I examine the three dilemmas revealed in Chapter 1: the dilemma of
organizational change approaches, the dilemma of a systemic perspective, and the
dilemma of measurable change. This examination proceeds through sections
reviewing organizational change literature and systems theory literature, along with a
brief discussion of the need for OD interventions to demonstrate measurable results.
Each section includes a discussion of how accounting metrics can serve as a
resolution to the relevant dilemma. The chapter concludes with the development of a
diagnostic-dialogic organizational change model founded on accounting metrics.
Dilemma of Organizational Change Approaches
The current OD tools for addressing organizational change may be broadly grouped
into a diagnostic and dialogic dichotomy (Bushe & Marshak, 2008, 2009; Marshak &
Grant, 2008). Diagnostic change is seen as the traditional approach to organizational
change, generally recognized as a process of data gathering to discover the current
organizational reality, to diagnose dysfunctional issues, and to develop an
intervention for moving organizational reality toward a presupposed norm (e.g.,
Lewin’s three-step model, classical OD action research, survey feedback, socio-
technical system analysis, and SWOT analysis) (Burke, 2008a; Bushe & Marshak,
2009; Cummings & Worley, 2009; Marshak & Grant, 2008). In contrast, dialogic
change is more recently recognized as those change efforts (e.g., appreciative inquiry,
world café, future search, and open space) that assert that organizational reality
7
cannot be objectively “discovered” but can be only subjectively understood through
personal experience, since organizational reality is not an objective “thing” to be
defined but is rather the social construction of the human systems that constitute an
organization (Bushe & Marshak, 2008, 2009; Marshak & Grant, 2008). Here data is
gathered from the subjective experiences, assumptions, and beliefs of organizational
members in order to create understanding and generate a vision and strategy for
reaching a desired organizational future (Bushe & Marshak, 2009).
In addition to contrasting purposes for data gathering, diagnostic and dialogic
approaches may be contrasted in their roles of dialogue and in their objects of change.
Dialogue has historically been included in diagnostic change efforts for the discovery
of “misperceptions” held by organizational members that need to be corrected or
somehow reintegrated into the organizational reality (Marshak & Grant, 2008). This
dialogue is therefore comparative and diagnostic. Conversely, the dialogue
incorporated into dialogic change approaches is discursive, with its intent being the
development of conversations and narratives to reveal and generate meaning and
create organizational reality (Bushe & Marshak, 2008, 2009; Marshak & Grant,
2008).
These differences in purpose of data gathering and dialogue lead directly into
differences in change objects between the two organizational change approaches.
8
Both approaches recognize the systemic aspect of organizations, but diagnostic
approaches tend to view organizations as open systems, defined by structure and
function, and seek to change people’s behavior in light of the diagnosis of
dysfunctional behavior relative to the defined system (Bushe & Marshak, 2008, 2009;
Marshak & Grant, 2008), whereas dialogic approaches view organizations as human
systems, defined only through the meaning making of their participants, and seek to
change the mental frameworks that guide people’s thinking as they create the
organizational system (Bushe & Marshak, 2008, 2009; Marshak & Grant, 2008).
These contrasts between diagnostic and dialogic organizational change approaches
are summarized in Table 1.
Table 1. Comparison of Diagnostic and Dialogic Change
Characteristic Diagnostic Dialogic
Ontology Objectivist Subjectivist
Epistemology Positivist Interpretivist
View of Organizations
Open Systems
(Structurally-Functionally
Constructed)
Meaning-Making
Systems (Socially
Constructed)
Role of Dialog Comparative & Diagnostic Interactive &
Constructivist
Change Object Behavior Mental Frameworks
Process
Gather data, diagnose
issues, and prescribe
change effort
Gather data, interpret
data, generate new
reality and
achievement strategy
9
Synthesizing Diagnostic and Dialogic Approaches
While Table 1 indicates that the diagnostic and dialogic approaches to organizational
change share little in their characteristics and assumptions, researchers have argued
that, rather than being disparate methods, the two approaches may be viewed as
elements of a whole approach toward organizational change (Oswick, 2009;
Woodman, 2008). The combination, or synthesis, of diagnostic and dialogic methods
would allow a more holistic approach to organizational change; and this study argues
that such a synthesis can occur through the use of accounting metrics as both
diagnostic and dialogic instruments of change to promote outcome efficacy and foster
a systemic perspective. Note that accounting metrics are here defined not only as the
summarized metrics presented in organizational financial statements but also as any
level of summarized or transactional accounting metrics that represent intra- or
interorganizational activity, such as individualized invoice and payment transactions
as well as aggregated accounts receivable and accounts payable balance sheet
amounts.
Accounting Metrics as Diagnostic Instruments
The use of accounting metrics as diagnostic instruments is well established in the
literature. Accounting metrics are often equated with objective organizational reality
(i.e., discovering organizational reality) (Argyris, 1953, 1990), are useful in
“accumulating data on actual events and comparing them with planned objectives”
(i.e., diagnosing dysfunctions) (Lawrence & Lorsch, 1969, p. 225), and are the
foundation of budgetary controls, the pivotal tool of organizational control (i.e.,
10
interventions to change behavior toward normative goals) (Argyris, 1952; Lawrence
& Lorsch, 1969; Marginson & Ogden, 2005). These aspects of accounting metrics
clearly fit within the diagnostic approach, described earlier, as discovering the current
organizational reality, diagnosing dysfunctions, and developing interventions to
change behavior toward a normative reality.
Accounting Metrics as Dialogic Instruments
While the aspect of accounting metrics as diagnostic instruments is generally
assumed, the literature also clearly recognizes their potential as dialogic instruments.
Boland (1993) notes that accounting metrics are symbols that organizational actors
use to create both context and meaning and are “a central element in the shaping of
organizational reality” (1989, p. 598). Argyris (1990) suggests that accounting
metrics can be used to engage, rather than merely describe, the mental frameworks
which create organizational defensive routines. These aspects clearly demonstrate that
accounting metrics can fit within the dialogic approach, described earlier, as creating
organizational reality through context and meaning, and as developing interventions
to change the mental frameworks that guide people’s thinking as they create the
organizational system.
Accounting Metrics as Diagnostic-Dialogic Instruments
The prior sections demonstrate that accounting metrics can be both diagnostic and
dialogic instruments, but the literature provides a further path for managing this
dichotomic tension through a diagnostic-dialogic synthesis. Simons (1991, 1994)
charts a path for this synthesis by recognizing that diagnostic controls can be
11
converted into interactive mechanisms to encourage dialogue and learning in an
organization. He posits that the conversion of a diagnostic control into an interactive
mechanism can occur through a combination of four steps:
1. ensuring that the control is an important and recurring agenda in discussions,
2. ensuring that the control is a focus of operating managers,
3. management participation in face-to-face meetings, and
4. the continual challenge and debate of data, assumptions and action plans. (Simons, 1994)
Creating a diagnostic-dialogic synthesis by capturing the dual capabilities of
accounting metrics through this conversion path allows for a resolution to the
dilemma of organizational change approaches by providing diagnostic, comparative
abilities relative to a normative performance factor while simultaneously allowing
dialogic, interactive conversion of that normative performance factor and of the
organizational change process itself.
Diagnostic-Dialogic Construct
Based on the key elements of accounting metrics as both diagnostic and dialogic
change instruments in combination with Simons’ (1994) interactive conversion steps,
I define the construct of a diagnostic-dialogic organizational change process, for the
purpose of this research study, as follows:
An organizational change process, framed by organizational accounting metrics,
which retains its diagnostic, comparative abilities relative to a normative
performance factor while allowing interactive conversion of that normative
12
performance factor, and of the organizational change process, through a dialogic
process by:
ensuring that the factor and process are an important part of continuing discussions,
ensuring that the factor and process are a focal point for operating managers,
ensuring active participation in face-to-face meetings related to the factor and process, and
encouraging the challenge and debate of data, assumptions and action plans related to the factor and to the process.
Having now defined the diagnostic-dialogic organizational change construct for this
research study, and recognizing that both the diagnostic and dialogic paradigms of
organizational change incorporate a systemic perspective, the following sections
provide theoretical support for the diagnostic-dialogic model as a tool to foster a
systemic perspective.
A Systemic Perspective One of the recognized similarities between diagnostic and dialogic change approaches
is their focus on organizational systems. Bushe and Marshak (2009) note that at least
three common themes can be identified in both approaches: self-awareness, behavior,
and actualization of organizational systems. One can paraphrase their argument for
these three themes:
1. Self-awareness indicates that the more developed a system is, the greater
awareness it has of itself, and the greater ability organizational members
have to talk with one another about their perceptions of the organization.
13
2. Behavior becomes increasingly more goal-directed and rational versus
emotional and reactive the more developed a system becomes.
3. More developed systems have greater capacity to actualize their potentials
and greater capacity to achieve their purposes within their environments.
These systemic elements of greater discussion, goal-directed behavior, and actualized
potentials and purposes are each evident in the diagnostic-dialogic construct.
The Dilemma of a Systemic Perspective
The literature has long recognized that organizations are complex systems comprising
differentiated, yet interdependent, subsystems that must systematically integrate in
order successfully operate in changing environments, where uncertainty and
ambiguity tend to increase (Burke, 2008b; Katz & Kahn, 1978; Lawrence & Lorsch,
1969; Mohrman, et al., 1996; Senge, 1990; Weisbord, 2004). Indeed, this
understanding is a foundational element of a systemic or “whole-systems” approach
toward organizational change (Ackoff, 1994; von Bertalanffy, 1972), which
recognizes that, lacking a systems view, management makes decisions provoking
unforeseen reactions, thus creating future problems as the result of today’s solutions
(Senge, 1990; Sterman, 2002; Zexian & Xuhui, 2010).
But while the need for systemic organizational change is well recognized, it is not as
frequently applied. Ackoff, a prominent systems theorist, argues: “Very few
14
managers have any knowledge or understanding of systems thinking, and for good
reason. Very little of our literature and lectures are addressed to potential users”
(2006, p. 705). The literature also recognizes that managers are frequently not able to
apply academic knowledge that is presented since it is presented without contextual
relevance to their organizational systems (Tenkasi, et al., 2007).
In addition to the lack of practical and contextual approaches, Sterman calls attention
to several problems in learning and comprehending systems concepts.
The robustness of the misperceptions of feedback and the poor
performance they lead us to create across many domains result from
two basic and related deficiencies in our mental models of complexity.
First, our cognitive maps of the causal structure of systems are vastly
simplified compared to the complexity of the systems themselves.
Second, we are unable to infer correctly the dynamics of all but the
simplest causal maps. Both are direct consequences of bounded
rationality (Simon 1979; 1982); that is, the many limitations of
attention, memory, recall, information processing, and time that
constrain human decision making. (Sterman, 1994, p. 307)
Similarly, Gharajedaghi (2007) has noted a general difficulty in visualizing behavior
among interdependent variables and the effects of nonlinear feedback loops arising
from cognitive ability formed around independent variables and closed loop linearity.
And this difficulty is not restricted to the untrained organizational member: even
graduate-level students in elite schools have been found to have poor understanding
of basic systems thinking concepts (Sweeney & Sterman, 2000; Zulauf, 2007), as a
15
result of the imperceptibility, complexity, and counterintuitive nature of intercausal
relationships (Hung, 2008).
The literature’s clear recognition of both the need for a systemic perspective in
organizational change and the difficulty of achieving it reveal the dilemma of a
systemic perspective. This study argues that when accounting metrics are embedded
in a diagnostic-dialogic framework, they can aid in resolving the dilemma of a
systemic perspective by metaphorically revealing organizational systems and
enabling organizational members’ cognition and sense-making activities within the
context of a cybernetic, double-loop system.
Accounting Metrics: Promoting a Systemic Perspective within the Diagnostic-Dialogic Framework
1
If you are a member of an organization and you’ve never considered
yourself to be a part of a subsystem which in turn is part of a larger
system which has a myriad of additional subsystems, then you’ve never
been involved in a budget meeting.
(Burke, 1980, p. 209)
Having served in financial leadership roles for a number of years, personal experience
leads me to agree wholeheartedly with Burke’s implication that organizational
budgeting and accounting metric processes are among the most revealing processes
1 Note: Portions from “Synthesizing a Systems Perspective and Organizational
Change: Principles of a Whole-Systems Metrics Model,” by W. Brock, OD Journal,
30, (3), 17-28, 2012. Copyright 2012 by The International Society of Organizational
Development. Reprinted with permission.
16
for seeing organizational systems. While not exhibiting the full complexity of systems
dynamics or elaborated causal-loop diagrams, organizational accounting metrics
(particularly those of integrated financial models) are the closest experience that
many managers have had with a systemic perspective. But the apparent simplicity of
these metrics must not dissuade one from considering their application in
understanding the whole system, since research has shown that abstract, simplistic
concepts have great power to create and manage complex systems (Wolfram, 2002).
Organizations are replete with accounting metrics; but while this research study limits
discussion to customary accounting based metrics, personal experience shows that
sufficiently abstracted or translated nonfinancial metrics can be converted into
financial metrics that are readily linked to the organizational level.
Clearly, a financial, or accounting, metrics viewpoint of organizational reality is
prevalent in contemporary society. Its presence is ubiquitous in the quarterly and
annual reports of any publicly traded firm; constitutes a major component of
communications with investors in earnings calls, shareholder meetings, and analyst
conferences; and is referenced in news reports of national, state, or local budgetary
crises. It exists in codification of both its principles, many of which are captured
officially as local country GAAP (Generally Accepted Accounting Principles) and
through regulating mechanisms (e.g., chartered and certified public accountant
licensure, audited financial statements, and Sarbanes-Oxley testing). Both causal to
17
and resulting from its prevalence in society, the financial or accounting metrics
viewpoint is generally equated with an objective view of the organization (Morgan,
1988). Its foundational process of budgeting is recognized as the pivotal
organizational control tool (Argyris, 1952; Lawrence & Lorsch, 1969; Marginson &
Ogden, 2005), and it is regarded as a critical element in transforming organizational
strategy into reality (Melnyk, Stewart, & Swink, 2004).
In spite of this general acceptance, however, (Morgan, 1988) has posited that the view
provided by accounting metrics is not the objective view of reality but is instead a
lens for understanding the whole systems of organizations through numerical
metaphor. As discussed in the section on accounting metrics as dialogic instruments,
the OD literature recognizes that accounting metrics are symbols that organizational
actors use to create both context and meaning. (Boland, 1993) and (1989) calls them
“a central element in the shaping of organizational reality” (p. 598) and (Argyris,
1990) says that they may be used to engage, rather than merely describe, mental
frameworks.
But support for this concept of accounting metrics as abstract and metaphorical is not
limited to the OD literature. Measurement theory provides that metrics, such as
accounting metrics, are an abstraction of the attributes of tangible and intangible
things, but they are not the attributes themselves (Woodman, 1989). Enumerology
18
theory lends support by theorizing that these metrics are not objective and specific
but, instead, are subjective and holistic representations of phenomena. With its focal
concern around the social processes by which numbers are constructed and used for
everyday purposes (Bogdan & Ksander, 1980; Gephart, 2006), enumerology theory,
as indicated by Bogdan & Ksander (1980), Carlon, Downs, & Wert-Gray (2006), and
Gephart (2006), posits that
metrics of social phenomena are not objective but are the result of attitudes
toward those phenomena;
the act of measuring changes the perceptions of things, and metrics can make
the amorphous appear concrete;
metrics are both a reflection and an influencer of events;
metrics are “signs” that represent value but lack the attributes they represent;
metrics are central to how organizations experience their environment; and
organizational reality is structured in the belief of the concreteness of metrics.
Far from the popular perception of metrics as objective and specific, the literature
recognizes that accounting metrics are a subjective and holistic representation of
phenomena. They are:
the primary means of seeing and communicating organizational nature
(Swieringa & Weick, 1981);
19
“framed by an overarching metaphor encouraging a numerical view of
reality” (Morgan, 1988, p. 480); and
constructivist, serving as representations of organizational members’
cognition and sensemaking activities (Boland, 1981; Napier, 2006; Tenkasi &
Boland, 1993).
Based on this theoretical support, we might stop here, satisfied that accounting
metrics indeed serve as metaphorical representation of organizational systems. But as
a lifelong musician, I would be remiss if I failed to consider a final support for this
argument: the jazz metaphor.
A jazz metaphor of organizational change has been used in the literature to
demonstrate that the flexibility required to execute whole-system, integrated change
within the interdependent, complex systems of organizations is like jazz
improvisation techniques (Barrett & Peplowski, 1998; Hatch & Cunliffe, 2006;
Pasmore, 1998). But while a jazz metaphor calls to mind flexibility, one of its
counterintuitive features is that what appears fully flexible is actually a structured
form supporting maximum flexibility while always allowing participants to
understand the current harmonic reality (i.e., the whole system) of the song in play
(Barrett & Peplowski, 1998; Pasmore, 1998).
20
An even more fundamental observation of the jazz metaphor reveals that its structural
form is based on metrics. Indeed many, if not most, of its practitioners refer to song
progressions as a series of metrics, such as the II, V, I chord progression of many
popular songs (Barrett & Peplowski, 1998). It is easy to understand why. Even jazz
music relies on a structure of defined measure, beats, and rhythms; and its complex
structure with multiple iterations of named chord patterns in various keys can be
simplified into just a few holistic metric representations. When we consider that
metrics are the underlying, holistic representation of even one of the most
improvisational music forms, we have further theoretical support for using an
accounting metrics-based metaphor to view complex, organizational whole systems.
Accounting Metrics as Cybernetic, Double-Loop Systems
As described above, organizational accounting metrics are possibly the most relevant
lens for organizational members, as well as being extremely revealing lenses based on
their metaphorical and symbolic representation of the whole organizational system
and of members’ cognition and sensemaking activities (Argyris, 1990; Boland Jr,
1989; Boland, 1981, 1993; Morgan, 1988; Napier, 2006; Tenkasi & Boland, 1993).
However, organizational accounting metrics can go further than metaphorical and
symbolic representation. When embedded in a diagnostic-dialogic approach as
described in earlier sections, accounting metrics can serve as the foundation of
cybernetic, or double-loop, systems.
21
Cybernetics is generally ascribed as originating from Norbert Wiener’s work in the
1940s (Churchman & Ackoff, 1950; Jackson, 2006; Wolfram, 2002) and can be
thought of as systems control through communication (Ackoff, 1974; Churchman &
Ackoff, 1950). It focuses primarily on the structure, adaptation, and learning of
organizations in order to maintain stability when confronted by exogenous
disturbances (Schwaninger, 2001, 2006) and is based on “active learning and control
rather than the passive adaptability” of open systems (Flood & Jackson, 1991, p. 10).
This distinction relates to the ability of a cybernetic system to exhibit double-loop
feedback (Montuori, 2000), thus allowing the cybernetic system not simply to adjust
to a variation of inputs in order to continue toward a preset goal but also to determine
whether the preset goal is still appropriate in light of internal and external feedback
(Montuori, 2000; Senge, 1990). This cybernetic process underlies the concept of
learning organizations.
Within OD literature, cybernetic systems are probably best represented through the
double-loop learning model. This model extends the concept of single-loop feedback
and correction (i.e., diagnostic processes) to double-loop feedback and learning,
which do not merely make corrections to reach an existing normative goal but go
further to question whether a normative goal is still appropriate or relevant and to
provide organizational members the ability to contextually determine appropriate
goals and strategies (i.e., dialogic processes) (Argyris, 1976, 1977; Morgan, 1982),
22
with an understanding that individual elements may need to operate at suboptimal
levels (Pasmore & Khalsa, 1993).
Using accounting metrics in a synthesized diagnostic-dialogic approach, as described
in the section on organizational change approaches, creates a cybernetic, double-loop
system. The diagnostic element of the synthesized approach provides the traditional,
and necessary, negative feedback role associated with single-loop learning, thus
enabling the equifinality aspect of systems thinking (i.e., different paths to reach the
same goal) (Hammond, 1997; Montuori, 2000). The dialogic element of the
synthesized approach provides an interactive and inclusionary role with a focus on
dialogue and inquiry thus supporting Argyris’s double-loop learning process
associated with organizational learning (Argyris, 1976, 1977; Henri, 2006) and
thereby enabling the equipotentiality aspect of systems thinking (i.e., changing,
differentiating to establish systemic congruency) (Hammond, 1997). By utilizing an
accounting metrics lens that is readily available and understandable to organizational
members to enable application of a systemic perspective through the potential of
accounting metrics within a diagnostic-dialogic framework, the diagnostic-dialogic
approach provides a resolution to the dilemma of a systemic perspective.
Dilemma of Measurable Change The OD literature describes organizational change as the empirical observance of
variation in an organizational entity’s form, quality, or state over time (Ford & Ford,
23
1995; van de Ven & Poole, 1995; Weisbord, 2004). But how is empirical observance
measured? This question lies at the root of OD’s continual challenge to provide
measurable evidence of the impact of organizational change interventions on
performance (Pettigrew, Woodman, & Cameron, 2001; Woodman, 1989), and the
dilemma of providing this evidence is that change is messy, outcomes are difficult to
measure, and participants often do not see necessity of rigorous evaluation (Melnyk,
Hanson, & Calantone, 2010; Woodman, et al., 2008). This study argues that
accounting metrics, within a diagnostic-dialogic framework, can also aid in resolving
the dilemma of measurable change by serving as change measurement instruments for
both diagnostic- and dialogic-based organizational change.
Accounting Metrics as Change Measurement Instruments The measurement mechanism of accounting metrics is actually incorporated within
the previous sections on accounting metrics as diagnostic-dialogic change
instruments, as metaphorical representation of organizational systems, and as
cybernetic double-loop systems. A quick summary of elements from these arguments
supports the use of accounting metrics as change measurement instruments based on
their characteristics of being
understood as the pivotal organizational control tool (Argyris, 1952; Lawrence
& Lorsch, 1969; Marginson & Ogden, 2005);
accepted as indicative of organization performance (Burke, 2008a);
equated with an objective view of the organization (Morgan, 1988);
24
“a central element in the shaping of organizational reality” (Boland, 1989, p.
598); and
metaphoric symbols representing the whole organizational system and
members’ cognition and sensemaking activities (Argyris, 1990; Boland Jr,
1989; Boland, 1981, 1993; Morgan, 1988; Napier, 2006; Tenkasi & Boland,
1993).
Through these characteristics, accounting metrics, within a diagnostic-dialogic
framework, provide a resolution tool for the dilemma of measurable change by
serving as a control tool to deal with “messiness” (i.e., pivotal organizational control
tool), an accepted measurement mechanism (i.e., indicative of organization
performance, equated with objective view), and an aid in rigorous evaluation through
cognition and sensemaking activities (i.e., metaphoric symbols, shaping
organizational reality).
Diagnostic-Dialogic Organizational Change Model Table 2 demonstrates how the diagnostic-dialogic organizational change model,
founded on accounting metrics, functions to resolve the dilemmas of organizational
change approaches, a systemic perspective, and measurable change by:
incorporating the characteristics of both diagnostic and dialogic organizational
change,
providing organizations with a method to see and conceptualize the system,
and
providing a means to measure the impact of organizational change.
25
T a b
le 2
. D
ia g n
o st
ic -D
ia lo
g ic
O r g a n
iz a ti
o n
a l
C h
a n
g e S
c h
e m
a
S y
n th
e s is
m e c h a
n is
m s
D ia
g n o
s ti
c D
ia lo
g ic
O rg
a n iz
a ti
o n a
l C
h a
n g
e S
y s te
m ic
P e rs
p e c ti
v e
M e a
s u ra
b le
c h a
n g
e
D is
c o
ve ri n g
o rg
a n iz
a tio
na l re
a lit
y
E n su
ri n g t h a t th
e n
o rm
a tiv
e a
c c o
un tin
g
m e tr
ic a
n d
p ro
ce ss
a re
a n i m
p o
rt a n t an
d
re c u rr
in g a
g e n d
a in
d is
c u ss
io ns
D ia
g n o
si n g
d ys
fu n c tio
ns &
C h a n g in
g b
eh a v io
r
to w
a rd
n o
rm a tiv
e
g o
al s
E n su
ri n g t h a t th
e n
o rm
a tiv
e a
c c o
un tin
g
m e tr
ic a
n d
p ro
ce ss
a re
a f o
cu s
o f
o p
er a tin
g m
a n a g e rs
S h a p
in g
o rg
a n iz
a tio
na l
re a lit
y
E n su
ri n g a
c tiv
e p
ar tic
ip at
io n
in m
e e tin
g s
re la
te d
t o
t h e a
c c o
un tin
g m
e tr
ic a
n d
p ro
ce ss
C re
a tin
g c
o nt
e x t
a n d
m e a n in
g &
E n g a g in
g m
e n ta
l
fr a m
e w
o rk
s
E n c o
ur a g in
g t h e c
h a lle
n g e a
n d
d eb
at e o
f
d at
a ,
as su
m p
tio ns
a n d
a c tio
n p
la n s
re la
te d
t o
t h e a
c c o
un tin
g m
e tr
ic a
n d
p ro
ce ss
(S im
o ns
1 9
9 1
, 1 9
9 4 )
C h a
ra c te
ri s ti
c s o
f A
c c o
u n ti
n g
M e tr
ic s
in a
n O
rg a
n iz
a ti
o n a
l C
h a
n g
e P
ro c e s s
D ia
g n o
s ti
c -D
ia lo
g ic
O rg
a n iz
a ti
o n a
l C
h a
n g
e M
o d e l:
F o
u n d e d o
n A
c c o
u n ti
n g
M e tr
ic s
D ia
g n o
st ic
p ro
ce ss
e s
w ith
c o
m p
ar is
o n
to n
o rm
a tiv
e
p er
fo rm
a n c e f a c to
r.
B e h a v io
ra l fo
cu s
D ia
lo gi
c p
ro ce
ss e s
w ith
c o
nv e rs
io n
an d
d ev
e lo
p m
e n t
o f p
er fo
rm a n c e f a c to
r.
In te
ra c tiv
e f o
cu s
C y b
er n e tic
, le
a rn
in g s
y st
e m
.
D o
ub le
-l o
o p s
e rv
e s
in te
ra c tiv
e
a n d
in c lu
si o
na ry
r o
le &
e q
ui p
o te
n tia
lit y
M e ta
p ho
ri c a l, s
y m
b o
lic
re p
re se
n ta
tio n
o f o
rg a n iz
a tio
na l
sy st
e m
. S
in g le
-l o
o p s
e rv
e s
n e g a tiv
e f e e d
b ac
k r
o le
&
e q
ui fin
a lit
y
R ig
o ro
us e
v a lu
a tio
n
th ro
ug h c
o gn
iti o
n an
d
se n se
m a k
in g a
c tiv
iti e s
C o
nt ro
l t o
o l t
o d
ea l w
ith
"m e ss
in e ss
." A
c c e p
te d
m e a su
re m
e n t m
e c h a n is
m
26
Chapter 3: Methodology The diagnostic-dialogic organizational change model developed in Chapter 2 was
proposed as a conceptual model for organizational change based on a synthesis of the
characteristics of accounting metrics as
diagnostic and dialogic instruments,
metaphoric and symbolic representations of organizational systems and members’ cognition and sense-making activities,
cybernetic (double-loop) systems, and
change measurement instruments.
But it remains for this conceptual model to be empirically tested with regard to its
effectiveness as a causal agent of organizational change. The following sections
provide the methodological structure and basis for empirically testing this model in an
organizational setting.
Methodological Structure
The methodology chosen for this study is a single-case, retrospective, longitudinal
field study. Each element of this methodological structure (i.e., field study, single-
case, longitudinal, and retrospective) is both elaborated and discussed for
appropriateness in the following paragraphs.
Field Study
A field study is a nonexperimental scientific inquiry occurring in a real organizational
setting rather than in a laboratory setting. As a field research method, it serves as a
powerful research approach within behavioral science (Kerlinger & Lee, 2000). Snow
27
and Thomas (1994) state that “field research methods can, perhaps more than any
other method, realistically examine strategic processes and outcomes – that is, they
provide mechanisms for observing strategists and organizations in their natural
settings” (pp. 457-458). As a research approach, field studies are noted as comprising
five characteristics:
1. They occur ex post facto with no independent variables manipulated
by the researcher.
2. They occur in intact, naturally occurring systems.
3. Their variables are systematically measured.
4. The researcher attempts to minimize intrusion.
5. Their focus is exploration, description, or hypotheses testing. (Stone,
1978)
This study is appropriately conducted as a field study because it meets the five criteria
denoted. It is conducted after the occurrence period, it occurs within an existing
organization, variables are operationalized and measured systematically, intrusion is
minimized by limiting the research data to archival records, and the focus is to test
hypotheses developed from the diagnostic-dialogic theoretical model. Using a single
case to conduct the study is also valid because the study fits one of five scenarios
described by Yin (2003) in which a single case is appropriate. Specifically, it is a
28
longitudinal study in which the researcher can study the same case at multiple points
of time.
Longitudinal Study
Longitudinal studies can be used to provide valid measurements of development over
time for explanatory or descriptive purposes (Schaie & Hertzog, 1982). In
longitudinal studies, time actually sets a frame of reference, providing context for the
changes seen and how those changes are explained. It is a method used “to capture
reality in flight” (Pettigrew, 1990, p. 270).
The ideal goal of longitudinal research, as with many research methods, is to provide
clear and unambiguous answers to research questions through the benefit of a view of
complex and dynamic changes over time (Collins, 2006; Featherman, 1980). This
ideal is most closely met when the longitudinal research seamlessly integrates three
characteristics:
1. It is based on a well-articulated theoretical change model.
2. Its temporal design provides a clear and detailed view of the research
process.
3. The resulting data is statistically analyzed through a model that
operationalizes the theoretical model. (Collins, 2006)
29
This study fits the longitudinal research ideal since
it is based on the theoretical diagnostic-dialogic organizational change model
developed from the literature review provided in Chapter 2,
the temporal design is clearly detailed within this methods section, and
the research model is an operationalized version of the theoretical model and
is statistically analyzed.
Retrospective Study
While sometimes opportunistically chosen, retrospective longitudinal data is actually
an advantageous resource for research data. One of its benefits is its provision of
simultaneous, replicative measurements, which improves measurement consistency
and equivalency (Featherman, 1980). Retrospective data also provides a viable
manner for conducting longitudinal studies since it provides a view of complex and
dynamic changes over time, while reducing one of the primary drawbacks to
longitudinal research, namely, the time span and cost required to do the study
(Featherman, 1980). Although one weakness typically identified with retrospective
studies is the unreliability of memory recall (Featherman, 1980), this study avoids
that weakness because its design relies completely on written, archival data.
Research Setting
The following sections provide a contextual setting for this research study by briefly
describing the WholeCo organization, as well as the customer proposal process and
the process interventions that serve as the focus of this research study.
30
The Company
The research setting is a group of consumer products organizations with more than 50
years of history, referenced in this study as WholeCo. WholeCo comprises two
consumer products companies with multiple categories and operating under the
umbrella of one management group. WholeCo operates as a wholesaler to home
hardware centers, mass merchants, wholesale clubs, and the electrical contractor
market. Product breadth ranges from textile and hardware products with low
consumer brand recognition and relatively high volatility in the exogenous
environment (CommodityCo) to specialty electronics products with high consumer
brand recognition and relatively low volatility in the exogenous environment
(SpecialtyCo). Although WholeCo enjoys a substantial market position within almost
all its product lines, some of its categories compete as commodities against many
domestic providers and direct foreign importers, while other categories benefit from
high brand recognition and have few direct competitors.
WholeCo is a U.S.-based company but operates as a global sourcing, manufacturing,
and distribution company with sourcing activities throughout the Americas, Europe,
and Asia; manufacturing operations in the Americas and Asia; and distribution
occurring either directly or through distributors in the Americas, Europe, and Asia.
WholeCo seeks to maintain a competitive cost position, but it does not strategically
operate as a low-cost provider since much of the company’s market and competitive
positioning are believed to result from its long history in its product categories, its
31
substantial breadth of product line, its supply-chain flexibility, and its long
relationships within its customer channels.
The Situation
The time period in this longitudinal study occurs during a volatile, rapidly changing
environment. Retrospective, longitudinal data is drawn from June 2008 through
October 2011, a well-recognized period of domestic and global economic distress
including high unemployment, a depressed U.S. housing market, and bank and
business closures. Many of these items affected WholeCo, since performance of its
key product categories is directly tied both to retailer performance and to housing
starts. Market volatility was high and predictability low for WholeCo during this
period as retailers increased the frequency of product line reviews in an effort to both
lower their own cost structures and improve their cash flow positions.
Simultaneously, product and component suppliers increased their selling prices to
help offset rising production costs, and lenders tightened credit policies, which forced
several of WholeCo’s long-time customers to file bankruptcy, thus confronting
WholeCo with both loss in receivables and loss in continuing revenue stream.
In summary, during the study period, WholeCo operated as a complex and dynamic
entity in a volatile environment that required organizational change to be frequent,
rapid, and coordinated throughout the whole-system of WholeCo. The following
section describes a particular process, the customer proposal process, used by
WholeCo to coordinate organizational change.
32
Customer Proposal Process
As is common with the wholesaler-retailer relationship in the consumer products
industry, frequent customer proposals are submitted by WholeCo to its retail
customers. These proposals are often initiated by retailer line reviews, which typically
resemble an auction process as the retailer bids out a particular category to the
incumbent wholesaler and its competitors in order to obtain price concessions and
improve its product portfolio. WholeCo may also initiate customer proposals as it
develops new products, enters new categories, or attempts to change pricing in order
to compensate for cost changes or to potentially delay retailer line reviews through
price reductions.
WholeCo has long used a cross-functional process for developing and approving
customer proposals because of their critical importance to its revenue and profit. It is
essential for the continued relationship with its customers that WholeCo makes only
proposals that it can fully support from a product, delivery, and financial standpoint.
WholeCo seeks to ensure this support through cross-functional alignment before it
submits proposals to its retail customers.
In order to establish cross-functional alignment, customer proposals are initiated by
the WholeCo Marketing department, either in response to a customer request or as a
proactive push for new product placement or pricing, by compiling information (see
Figure 1) including product mix and estimated volumes, customer programs, product
33
costs, estimates of retail sales price and retailer margin, and aggregated sales and
profit impact for WholeCo.
Once initiated by the Marketing department, customer proposals flow through each
functional department (Operations, Finance, and various Sales channels) for review,
commentary, and approval or rejection (see Figure 2). Customer proposals might be
modified anywhere along the process but generally require unanimous approval
before submission to the CFO and CEO for final approval. A flowchart of the process
is presented in Figure 3.
The WholeCo customer proposal process is presented in this study as an example of a
whole-system process since it serves as a critical component of external and internal
systems boundaries between the organization, its customers, its suppliers, and its
parent organization. It involves essentially all functions of the organization (either
directly or through representation), it provides feedback both to and from the system
and its exogenous and endogenous environments, and the WholeCo metrics it
represents (e.g. sales, margins, and profitability) define WholeCo’s raison d’être
relative to its parent company.
34
B u
s in
e s s U
n it
C u
s to
m e
r N
a m
e C
o o p
C a s h D
is c o u n t
0 .0
0 %
T y p
e o
f P
ro p
o s a
l V
o lu
m e R
e b a te
P a y m
e n t
T e rm
s
L in
e R
e vi
e w
, P
ro m
o ,
N e w
P ro
d u c t,
N e w
C u s to
m e r
O th
e r
A ll o w
a n c e s
C o m
m is
s io
n %
0 .0
0 %
P ro
d u
c t
O th
e r
G ro
s s
S a
le s &
E s ti
m a
te d
E s ti
m a
te d
It e
m &
D e
s c ri
p ti
o n
Q u
a n
ti ty
P ri
c e
C o
s t
C o
s ts
M a
rg in
A d
m in
P ro
fi t
R e
ta il
M a
rg in
P ro
d u c t
# 1
- D
e s c ri p ti o n
X
P ro
d u c t
# 2
- D
e s c ri p ti o n
X
P ro
d u c t
# 3
- D
e s c ri p ti o n
X
T o ta
l P
a c k a g e O
ff e ri n g
X
N e
t S
a le
s $ 0 .0
P ro
d u c t
C o s t
$ 0 .0
O th
e r
C o s t
o f S
a le
s $ 0 .0
G ro
s s M
a rg
in $ 0 .0
S a
le s a
n d
A d
m in
is tr
a ti
v e
C o
s ts
$ 0 .0
P ro
fi ta
b il
it y
$ 0 .0
T o
ta l
V a
lu e
o f
P ro
p o
s a
l
P R
O P
O S
A L
F O
R M
C u
s to
m e
r P
ro g
ra m
s
0 .0
0 %
0 .0
0 %
0 .0
0 %
C u
s to
m e
r W
h o
le C
o
F ig
u r e 1
. C
u st
o m
e r P
r o p
o sa
l F
o r m
35
Figure 2. Customer Proposal Approval Form
Originator: Date:
Project Title: Number:
Project Description:
Product Category:
New Model Number:
Expected Launch/Ship Date:
Approval / Signatures
Yes No
Comments:
Originator(s) (Trade Mktg Mgr & Sales) Date
Comments:
Marketing Manager--Trade Date
Comments:
VP Marketing Manager - Retail Date
Comments:
VP Marketing Manager - Contractor Date
Comments:
VP Sales Date
Comments:
VP Sales Date
Comments:
Director of Finance Date
Comments:
VP Finance and Operations Date
Comments:
Director of Supply Chain Date
Comments:
CFO Date
Comments:
CEO Date
APPROVAL
APPROVAL FORM
36
S a
le s
1 C
o m
m e
n ta
ry &
A p
p ro
v a
l /
R e
je ct
io n
S a
le s
2 C
o m
m e
n ta
ry &
A p
p ro
v a
l /
R e
je ct
io n
C F
O /C
O O
C o
m m
e n
ta ry
&
A p
p ro
v a
l /
R e
je ct
io n
M a
rk e
ti n
g D
e p
a rt
m e
n t
S u
b m
it s
C u
st o
m e
r P
ro p
o sa
l fo
r F
u n
ct io
n a
l
A p
p ro
v a
ls
S a
le s
3 C
o m
m e
n ta
ry &
A p
p ro
v a
l /
R e
je ct
io n
U n
a n
im o
u s
C o
n se
n t
to
P ro
ce e
d ?
Y e
s
A p
p ro
v a
l?
Y e
s N
o ti
fi ca
ti o
n o
f
A p
p ro
v a
l to
F u
n ct
io n
a l
H e
a d
s
O p
e ra
ti o
n s
C o
m m
e n
ta ry
&
A p
p ro
v a
l /
R e
je ct
io n
N o
C E
O C
o m
m e
n ta
ry &
A p
p ro
v a
l /
R e
je ct
io n
N o
F in
a n
ce C
o m
m e
n ta
ry &
A p
p ro
v a
l /
R e
je ct
io n
C E
O &
C F
O m
e e
t
w it
h
M a
rk e
ti n
g /S
a le
s
Y e
s
A p
p ro
v a
l?
N o
N o
ti fi
ca ti
o n
o f
R e
je ct
io n
t o
F u
n ct
io n
a l
H e
a d
s
F ig
u r e 3
. F
lo w
o f
C u
st o m
e r P
r o p
o sa
l P
r o c e ss
37
Interventions
Because of the critical role of the WholeCo customer proposal process in generating
and managing organizational revenue, margins, and profitability, WholeCo
management initiated two distinct and successive interventions during the
longitudinal period covered by this study, in efforts to improve the efficacy of the
customer proposal process by
improving completion time in order to provide more rapid quotation responses
to customers,
improving feedback capability so that decisions were more relevant to the
current environments of both WholeCo and its customers, and
providing a WholeCo context for each proposal to help maintain alignment of
each proposal with the greater system goals.
The successive timing of these two management interventions provides three distinct
time segments within the longitudinal period that serve for hypotheses testing as
quasi-experimental groups (see Figure 4):
Laissez-faire group – experiencing no intervention
Traditional group – experiencing a diagnostic-based intervention, and
Diagnostic-Dialogic group – experiencing a diagnostic- dialogic-based
intervention.
38
The following sections frame the research question, operationalize the three quasi-
experimental groups, operationalize five outcome variables, and develop hypotheses
to respond to the research question and to test the effect of the quasi-experimental
groups on the outcome variables.
Research Question
Based on the three quasi-experimental groups (Laissez-faire, Traditional, and
Diagnostic-Dialogic), the research question is:
In the context of the WholeCo customer proposal process, how do the effects of a
Diagnostic-Dialogic intervention contrast with those of a Traditional (diagnostic)
intervention and those of a Laissez-faire approach? The research question refers
specifically to the following:
the time duration required to complete the process,
the quantity of commentary incorporated into the process,
the quality of commentary incorporated into the process,
Figure 4. Longitudinal Timeline
Diagnostic-Dialogic
From: June 10, 2008 March 3, 2010 March 3, 2011
To: February 26, 2010 February 28, 2011 October 25, 2011
Laissez-Faire Traditional
39
the alignment of WholeCo profitability goals with customer proposal
profitability metrics, and
the likelihood of proposal acceptance by WholeCo customers.
Quasi-Experimental Groups
The following sections describe the organizational changes initiated by the two
management interventions that operationalize the three quasi-experimental groups.
Laissez-faire
The Laissez-faire group of customer proposals may be characterized as opportunistic
or reactive proposals since they were generally initiated in response either to a
general growth goal or to a retailer special request. While the process flow and
approval forms were consistent with the description provided earlier, each proposal
tended to stand alone and was evaluated on somewhat ambiguous standards.
WholeCo set total business objectives during this period, but it did not typically
establish aligned and specific profit objectives by customer and product categories.
This practice essentially made general revenue growth and average profit rates the
primary benchmarks. Both the WholeCo CEO and CFO were heavily involved in
most customer proposals during this period, since each proposal was unique and was
often evaluated against overall strategic plans and management intuition.
Traditional
As WholeCo experienced growth, WholeCo management realized that the Laissez-
faire customer proposal process was becoming too burdensome to maintain, since it
40
required the CEO and CFO to allocate greater amounts of time evaluating an
increasing number of customer proposals. This realization led to a management
intervention in March 2010 designed to provide WholeCo functional areas with a
budgetary framework for proposal evaluation, as is characteristic of a traditional,
diagnostic-based process, by requiring that all customer proposals include a
diagnostic review form (see Table 3). With this intervention, WholeCo management
hoped to influence functional behavior by reducing ambiguity relative to business
objectives (i.e., financial goals), enabling comparison with a known performance
standard, constructing a comparative framework for dialogue, and encouraging
functional areas to prescribe changes to meet normative standards. These goals are
each consistent with diagnostic organizational change approaches that are
characterized by comparative and diagnostic dialogue and by a focus on gathering
data in order to diagnose problems, prescribe change, and influence behavior.
Diagnostic-Dialogic
WholeCo management initiated another organizational change to the customer
proposal process in March 2011 in response to concerns that too many proposals were
becoming stuck in a “churning” mode as the Marketing department wrestled with
incompatible WholeCo and retailer goals for product pricing, delivery timing, or
other objectives. Management realized that the Marketing department (proposal
initiating department) often dealt with these dilemmas by repeated shuffling of
potential product offerings and by procrastination in proposal routing.
41
Direct appeals by the Marketing department to the CEO, prior to routing customer
proposals, also became more commonplace as a result of concerns that the
Finance department or the CFO would reject the customer proposal because of its
inability to meet targeted WholeCo profit goals as defined by the diagnostic process
Table 3. Traditional Intervention - Diagnostic Tool
Better
Packaged (Worse)
Proposal Plan vs. Plan
Gross Sales $230 $185 $45
Customer Allowances 30 10 (20)
Net S ales 200 175 25
Product Costs 150 140 (10)
Product Margin 50 35 15
% of Net Sales 25.0% 20.0% 5.0%
Other Cost of Sales 25 10 (15)
Gross Margin 25 25 0
% of Net Sales 12.5% 14.3% -1.8%
Sales & Admin Costs 5 3 (2)
Profitability $20 $22 ($2)
% of Net Sales 10.0% 12.6% -2.6%
DIAGNOS TIC COMPARIS ON 1
1 Numbers are demonstrative and are not drawn from any sp ecific WholeCo p rop osal in
order to maintain confidentiality
42
used in the Traditional segment. In these cases, the CEO and CFO generally met to
personally structure creative proposals that would provide a competitive offering to
the retailer and provide WholeCo with a sustainable revenue and profit stream. These
restructured proposals often required contemplation of factors outside the normal
diagnostic process to alter the perception of either WholeCo or retailer goals in order
to create a previously unrecognized alternative.
With this second intervention, WholeCo management intended to
provoke the ability of functional departments to better utilize their
interrelationships for deeper dialogue;
encourage functional departments to view dilemmas and goals through a
systemic lens and multiple frameworks; and
enable the creation of new potentials and alternatives as the WholeCo
environment continued to dynamically change in relation to suppliers,
retailers, its parent company, and the overall marketplace.
Management intervened in the customer proposal process by requiring all proposals
to include a diagnostic-dialogic, accounting metrics review (see Figure 5) to
encourage organizational change, a systemic perspective, and rigorous measurement.
43
F ig
u r e 5
. D
ia g n
o st
ic -D
ia lo
g ic
, A
c c o u
n ti
n g M
e tr
ic s
T o o l
S a
le s
&
G ro
ss A
d m
in
S a
le s
M a
rg in
% E
x p
e n
se s
% P
ro fi
t %
B u
d g
e te
d W
h o
le C
o G
o a
l $
2 ,3
5 0
$ 4
7 5
2 0
.2 %
$ 2
4 0
1 0
.2 %
$ 2
3 5
1 0
.0 %
P re
-s u
p p
o se
d ,
n o
rm a
ti v
e o
b je
ct iv
e
D ia
g n
o st
ic E
le m
e n
ts S
in g
le -l
o o
p d
ia g
n o
st ic
p ro
ce ss
:
C h
a n
g e
i n
c u
st o
m e
r fo
re ca
st s
($ 5
0 0
) ($
7 5
) 1
5 .0
% ($
7 5
) 1
5 .0
% s
R e
co g
n it
io n
o f
e n
v ir
o n
m e
n ta
l ch
a n
g e
s
C h
a n
g e
i n
c u
st o
m e
r re
b a
te s
$ 5
$ 5
$ 5
fr o
m e
x te
rn a
l cu
st o
m e
r &
s u
p p
li e
r sy
st e
m s
S u
p p
li e
r p
ri ce
n e
g o
ti a
ti o
n s
$ 2
0 $
2 0
s C
o rr
e ct
iv e
m e
a su
re t
o p
u ll
r e
su lt
s b
a ck
S u
b -t
o ta
l ($
4 9
5 )
($ 5
0 )
1 0
.1 %
($ 5
0 )
1 0
.1 %
to w
a rd
n o
rm a
ti v
e o
b je
ct iv
e
D ia
g n
o st
ic -D
ia lo
g ic
E le
m e
n ts
D o
u b
le -l
o o
p d
ia g
n o
st ic
-d ia
lo g
ic p
ro ce
ss :
P o
te n
ti a
l n
e w
p ro
d u
ct p
la ce
m e
n t
$ 2
0 0
$ 8
0 4
0 .0
% $
2 5
$ 5
5 2
7 .5
% s
P ro
p o
se d
s a
le s
to e
x te
rn a
l sy
st e
m
S u
p p
ly p
u rc
h a
si n
g d
is co
u n
ts $
8 $
8 s
C o
ll a
b o
ra ti
v e
d is
cu ss
io n
s to
d e
te rm
in e
D e
p a
rt m
e n
ta l
su p
p o
rt c
o st
s $
1 5
($ 1
5 )
a lt
e rn
a ti
v e
f a
ct o
rs a
n d
a ct
io n
s
S u
b -t
o ta
l $
2 0
0 $
8 8
4 4
.0 %
$ 4
0 2
0 .0
% $
4 8
2 4
.0 %
T o
ta l
P a
ck a
g in
g O
ff e
ri n
g $
2 ,0
5 5
$ 5
1 3
2 5
.0 %
$ 2
8 0
1 3
.6 %
$ 2
3 3
1 1
.3 %
C h
a n
g e
B e
tt e
r/ (W
o rs
e )
($ 2
9 5
) $
3 8
4 .8
% ($
4 0
) -3
.4 %
($ 2
) 1
.3 %
D o
u b
le -l
o o
p d
ia g
n o
st ic
-d ia
lo g
ic p
ro ce
ss :
D ia
g n
o st
ic E
le m
e n
ts ($
4 9
5 )
($ 5
0 )
2 .7
% $
0 -2
.7 %
($ 5
0 )
0 .0
% s
R e
co g
n it
io n
o f
e n
v ir
o n
m e
n ta
l d
is tu
rb a
n ce
D ia
g n
o st
ic -D
ia lo
g ic
I m
p a
ct s
$ 2
0 0
$ 8
8 1
.9 %
$ 4
0 -0
.8 %
$ 4
8 1
.1 %
s F
o cu
s o
n n
e w
D ia
g n
o st
ic /D
ia lo
g ic
g o
a l
st a
te
T o
ta l
Im p
a ct
s ($
2 9
5 )
$ 3
8 4
.6 %
$ 4
0 -3
.5 %
($ 2
) 1
.1 %
w h
ic h
i m
p a
ct s
o v
e ra
ll o
rg a
n iz
a ti
o n
f a
v o
ra b
ly
W S
M S
u m
m a
ry
B ri
d g
e o
f Im
p a
ct s
1 N
u m
b e rs
a re
d e m
o n
s tr
a ti
v e a
n d
a re
n o t
d ra
w n
f ro
m a
n y
s p
e c if
ic W
h o
le C
o p
ro p
o s a l in
o rd
e r
to m
a in
ta in
c o
n fi
d e n
ti a li ty
44
As indicated in the conceptual diagnostic-dialogic model developed in Chapter 2, the
WholeCo diagnostic-dialogic, accounting metrics model (Figure 5) starts with an
accounting metric symbolization of WholeCo’s link with its parent company
(exogenous system). From this normative boundary (i.e., budget or plan), the
diagnostic section of the WholeCo model recognizes environmental impacts through
the feedback properties of accounting metrics (i.e., change in demand and increase in
supplier cost). It then reveals diagnostic (corrective) processes that attempt to sustain
homeostasis (stability) among the WholeCo (endogenous) system and exogenous
systems (i.e., release of customer program funds).
Following this diagnostic section, the diagnostic-dialogic section of the WholeCo
model demonstrates how accounting metrics promote and record the double-loop
processes that are in action. This characteristic is shown by the encouragement of
collaborative discussions to reach a new goal state (i.e., new sales opportunity,
make/buy conversation, and incremental overhead absorption opportunity) indicative
of a cybernetic double-loop process. The WholeCo model then completes with a
summary of both diagnostic and diagnostic-dialogic impacts of the customer
proposal.
But perhaps the best way to demonstrate the diagnostic-dialogic and systemic
characteristics of the customer proposal process, post- diagnostic-dialogic
45
intervention, is to provide a narrative account of one of its successes, which resulted
in the launch of a new million-dollar product line.
Diagnostic-Dialogic Narrative
Early in the second quarter of 2011 (after the diagnostic-dialogic intervention), the
WholeCo Marketing and Sales functions met with a major retailer to propose its
switch of a particular product line from a competitor to WholeCo. This proposal was
unsuccessful; however, during their meeting, the retailer mentioned an issue it faced
with a Christmas promotional offering that required a specialty product outside of the
product categories offered by either the competitor or WholeCo. This type of
discussion would typically have ended in pleasantries such as “maybe we can help in
the future” or “perhaps next year we can meet again,” particularly since the specialty
product mentioned was not contemplated as a WholeCo product line, the introduction
timeframe was extremely short (approximately six months), and initial profitability
levels would not likely meet WholeCo targets. Indeed, the meeting ended without
commitment but with a general understanding that WholeCo would see what it could
do.
The Marketing and Sales functions returned to WholeCo and initiated interactive
discussions, using a diagnostic-dialogic process (see example in Figure 5) to
determine whether a feasible alternative to the specialty product might be offered to
the retailer. During the next few weeks, the various organizational functions gathered
46
data on sourcing and manufacturing capabilities, estimated marketplace acceptance,
preliminary product designs, investment requirements, and estimated profitability.
This data gathering resulted in the generation of an alternative that would meet the
retailer’s objective and timeline but would not meet WholeCo profit requirements.
However, the strategic importance of the retailer was highlighted during dialogue
initiated as part of completing the diagnostic-dialogic model; and because of the
revised lens of the customer proposal as a “new product introduction” rather than as a
one-time product promotion, WholeCo committed to the project.
The subsequent proposal won the retailer’s Christmas promotion. WholeCo delivered
the product in less than six months, dramatically less than normal product
development time of a year or more, and the product was a hit on the retailer shelf.
But like any good story, there was not an immediate happy ending; the project
reduced WholeCo profitability in the category below its planned objectives.
Retailer promotions, especially seasonal promotions, are notoriously unpredictable in
their potential for recurrence, and the mix of retail product success but WholeCo
profitability reduction would normally have led to this WholeCo event being
relegated to “experience” while it moved on to the next line review or next
promotional opportunity. But WholeCo was intrigued by what it had accomplished—
something that had not been done before; and the functional areas continued running
47
the customer proposal process to determine whether potential for a better outcome
existed. Note that the process was not running in a typical “corrective” fashion that
would seek to turn around profit margins on a poorly performing product line. This
product was considered one-time, and WholeCo was content to recognize it as a
strategic concession for future growth with its customer.
But this time was different. Now the WholeCo functions had a way to surface hidden
functional assumptions through accounting metrics embedded in a diagnostic-dialogic
process and to see them through a greater lens, a systemic lens. The WholeCo
functions continued to use the customer proposal forms as a framework but also
extended discussions outside the official forms in order to discover resolutions to the
dilemmas revealed through the surfacing of underlying assumptions.
In short order, WholeCo was able to accomplish several objectives:
Remove a third-party royalty charge by converting to a related company
brand
Develop a hybrid process of sourcing and manufacturing
Extend the product concept to similar product items
Expand distribution as a multiseasonal special offering.
48
Changes occurred throughout the WholeCo system, resulting in the launch of a new
million-dollar product line for WholeCo with higher than planned profitability.
Subsequently, WholeCo management requested feedback from the functional
managers about the customer proposal process post- diagnostic-dialogic intervention.
The following quotes, obtained from that feedback, provide additional empirical
observation of the dual characteristics of the diagnostic-dialogic, accounting metrics
process.
(Marketing Head)
One of the primary differences in the [diagnostic-dialogic] process is
that it opens up discussions to new options as well as recognizing the
consequences of those options and potential alternatives. It widens the
benchmarking process from a narrowly defined goal benchmark to
discuss what is the actual, right benchmark for comparison. It forces
you to think not just of the decision of the moment but also about
likely future impact and decisions that will need to be made down the
road.
(Operational Head)
You have to anticipate how the proposal will fit with existing goals
and call together functions to determine appropriate corrective actions
or goal revisions before submitting the proposals if you see there is a
disconnect. It helps make departmental functions aware of changing
priorities and helps them know when to call together ad hoc groups for
discussion. This results in more discussion and better alignment among
the functions and enables proposals to go through more quickly.
(Finance Head)
Preliminary conversations are key to the process, and the written
commentary within the proposal forms is reflective of those
preliminary conversations. This results in proposals being accepted by
management faster and more frequently.
49
(Sales Head)
One of the differences in the [diagnostic-dialogic] process is that it is
not just a financial tool. It is a decision process to see more scenarios
and to help make strategic decisions.
Hypotheses The research model is designed to test the impact of the three quasi-experimental
variables on five outcome variables: completion time, quantity of commentary,
quality of commentary, profitability alignment, and customer acceptance. With the
exception of profitability alignment, each outcome variable was present in all quasi-
experimental groups. The outcome variable profitability alignment was present only
in the Traditional and Diagnostic-Dialogic groups since archived data was not
available for the Laissez-faire group. The following paragraphs describe the
operationalization of these outcome variables and develop a priori hypotheses
regarding their behavior under the influence of each quasi-experimental group.
Hypothesis 1: Completion Time
Each customer proposal contains an initiation date to indicate when the WholeCo
Marketing department initiated the approval process) as well as the date of final
approval or rejection (see Figure 2). The time between the final approval or rejection
date and the initiation date is captured as a total number of elapsed days to represent
completion time.
50
Because of the ambiguity of the Laissez-faire process, the greater structure of the
Traditional process, and the greater interactive nature of the Diagnostic-Dialogic
process, I hypothesized the following:
H1a Completion time will decrease in the Traditional group relative to the
Laissez-faire group.
H1b Completion time will decrease in the Diagnostic-Dialogic group relative to
both the Traditional and Laissez-faire groups.
Hypothesis 2: Quantity of Commentary As shown in Figure 2, specific and assigned space was provided on each customer
proposal, during all quasi-experimental groups, for each functional head to record any
written comments desired. While comments were not required, they were encouraged,
particularly in cases where the functional head had a strong opinion (pro or con) or
where the functional head believed additional explanation was required to explain
what might appear a strategic or counterintuitive decision. This written commentary
was captured by department for each customer proposal and maintained on the
archived proposals. The quantity of commentary for each proposal was captured by
dummy coding, with “0” being assigned for each comment field that either was blank
or contained only the notations “N/A” or “None,” and with“1” being assigned for
each comment field that contained written commentary other than “N/A” or “None.”
The aggregate number of nonzero comment fields for each record was assigned as the
value of quantity of commentary for that record. The potential value for any record
51
ranged from 0, reflecting no comments, to 9, reflecting comments by each function
(i.e., marketing, 3 sales functions, operations, finance, other, CFO, and CEO),
although no record exhibited more than 6 comments.
Since the diagnostic nature of the Traditional group should logically motivate
individuals to explain significant differences of the proposal to the normative goal
(plan), and since the combined nature of the Diagnostic-Dialogic group should
logically motivate even greater commentary because of its inclusion of both
normative comparison and generative double-loop construction of new goals, the
following hypotheses were proposed:
H2a Quantity of commentary will increase in the Traditional group relative to the
Laissez-faire group.
H2b Quantity of commentary will increase in the Diagnostic-Dialogic group
relative to both the Traditional and Laissez-faire groups.
Hypothesis 3: Quality of Commentary
Written commentary provided in the space for functional comments was maintained
verbatim in the archived forms. These comments provided the basis for the outcome
variable, quality of commentary, by representing the degree of systemic commentary
present in each proposal.
52
The archived proposals contained a total of 1,206 comments, which were coded as
functional, systemic, or unclassified. Quality of commentary is defined for each
record as the number of systemic comments divided by the total number of functional
and systemic comments, namely, Systemic / (Functional + Systemic). Descriptions of
functional and systemic comments, examples of each type of comment from the
Marketing department, and a discussion of the iterative coding process are given in
the following sections.
Commentary Types
Functional comments are those describing or justifying the decision to approve or
reject a proposal without evidence of discussions with or data retrieval from others.
Examples include the following:
Decline in GM% & EBITDA are due to the heavy amount of business [customer] does in [specific product which]…is our
highest margin sku. . . . Priced to be competitive at retail with
[competitor] & provide [customer] a min margin of [X]%
Improves bottom line $ and percentages
Low margins but we are pushing to get [customer] promos in prep for line review
Systemic comments are those describing or justifying the decision to approve or
reject a proposal, with specific reference to discussions or decision-making activities
with (or data retrieval from) other organizational or external functions. Examples
include the following:
53
After several reviews, we believe this scenario provides both [Customer] and [WholeCo] with a positive change… The
long term goal is to show a lift in POS [i.e. retailer sales rate]
with this new sku which will lead to increased distribution and
possible [sic] an innovation end cap feature. Sales team
indicates a high confidence in successfully passing …. price
adder to the customer….. This will be a good test of [product]
in market in U.S. (emphasis added)
[Customer] has approached us for… support. We are proposing this sku as a tax season promo for 2012 that should
still provide a re-order for us in 2011. This lowers GM
[slightly]. This is a small impact to keep our positive
positioning going with the Associate buyer as we were recently
notified we will have a new buyer and we wish to start our
relationship positively. This is a new buyer and we are
working to build a relationship with him. (emphasis added)
Per [Finance] these are Excess not obsolete, so no reserves needed. The plan per [Sales] is to maintain these at [X]%
EBITDA which is above plan. levels and offer them at a net-
net with no program or Price up if program is needed
(emphasis added)
Unclassified comments are those representing comments that are neither functional
nor systemic. Examples include the following:
Attached docs must be signed and returned to [customer] by [date].
One-time purchase.
Consensual Coding Process
Before proceeding to hypotheses development for the outcome variable, quality of
commentary, it is important to describe the iterative process used to code commentary
in the customer proposal forms. Through this process, I reached full consensus with
54
the WholeCo CEO as to the coding for each of the 1,206 comments as functional,
systemic, or unclassified.
Step 1
All comments were segregated by their membership in the following functional
departments: Marketing, three different Sales functions, Finance, Operations, and
Other. In order to avoid potential bias in coding, all comments from the CFO and
CEO were excluded, resulting in an exclusion of 159 comments and a final coding
population of 1,206 comments.
Step 2
All record identifiers were hidden before coding, leaving only the department name
and actual comment visible in order to ensure blind coding. The department name
was necessary in order to make contextual judgments about whether a comment was
functional or systemic based on the intradepartmental, interdepartmental, or external
nature of discussions.
Step 3
As researcher, I conducted initial coding of 483 comments made by the Marketing
department and subsequently requested that the CEO code a sample of 19 of the most
difficult comments, representing approximately 4% of the Marketing comments, for
comparison. Comments were identified as difficult whenever I vacillated between the
appropriate codes during coding. The first round of coding comparison resulted in a
31.6% agreement rate between my coding and the CEO’s coding.
55
After the first round of coding, discussion with the CEO revealed that he was relying
on memory of specific discussions and events outside the written commentary to
shape his coding. This reliance generally led to a coding of “systemic” since the CEO
was involved in discussions across all functional levels. Discussion also revealed that
this process resulted in a particularly high level of bias in the Diagnostic-Dialogic
group, since the CEO’s memories were understandably better for the more recent
records that had occurred in this group of the longitudinal study. In order to address
this potential bias, preliminary written rules were established, based on discussion
with the CEO, for evaluation of each comment (see Table 4).
Table 4: Preliminary Coding Definitions for Quality of Commentary
Code Definition
Unclassified
General statements with no opinion stated, or no description of or
attempt to influence decision
Functional
Description or justification of decision without specific reference to
discussions, requests, or data retrieval from persons in other
departments or external persons (e.g., buyer, supplier) (“not
guaranteed sale” does not count as specific reference)
Systemic
Specific reference to discussions, requests or data retrieval (e.g.,
E&O list, new cost list – but not plan and not strategy plan) from
persons in other departments or to external persons (e.g., buyer,
supplier)
56
Step 4
As the next step, I recoded the 483 Marketing comments according to the preliminary
coding rules. This second round of coding showed improvement by reaching 52.6%
agreement between the CEO’s first round of coding and my second round of coding
for the previously selected 19-record sample from step 3. I met with the CEO again to
reassess the preliminary coding rules and potential coding difficulties. To ensure as
much coding independence as possible, the CEO was advised only of the total
percentage of agreement in the comparison of coding for both round one and round
two. The CEO was not advised of which comments were coded in agreement and
which comments were not coded in agreement. Rules were then modified slightly to
generate the final coding rules (see Table 5, changes in bold type).
Table 5: Final Coding Definitions for Quality of Commentary
Code Definition
Unclassified
General statements with no opinion stated, or no description of or
attempt to influence decision
Functional
Description or justification of decision without specific reference to
discussions, requests, or data retrieval from persons (e.g., “not
guaranteed sale” does not count as specific reference) in other
departments or to external persons (e.g., buyer, supplier)
Systemic
Specific reference to discussions, requests or data retrieval (e.g.,
E&O, new cost list – but not plan and not strategy plan) from
persons in other departments or to external persons (e.g., buyer,
supplier) – also includes requests for actions related to decision
making (i.e., “Sales will need to monitor POS during test period
for validation”) but not for decision execution (i.e., “put into
forecast”)
57
Step 5
The CEO and I separately conducted a final round of test coding for the previously
selected 19-record sample, using the final written rules and with previous coding
identifiers removed. This final coding test demonstrated 73.7% agreement in code
assignment between my coding and the CEO’s coding.
Step 6
I then coded all remaining comments, using the final written rules, for a total code set
of 1,206 comments. After completion, I met again with the CEO to review and
discuss each comment.
Consensus was initially reached on all except 12 comments, approximately 1% of the
comment population, which were deemed too difficult to classify without reviewing
the actual documentation. Documentation was retrieved for these 12 comments and
was reviewed simultaneously by the CEO and me. This process resulted in 8
comments retaining the assigned code and the code for 4 comments being changed in
order to reach complete coding consensus between the CEO and me.
Table 6 provides the final stratification of coded commentary by category as well as
by quasi-experimental group.
58
Hypotheses
Having provided the evolutionary path for operationalizing the quality of commentary
outcome variable, I next turned to developing hypotheses. Because of the focus on
single-loop diagnostic processing in the Traditional group without an interactive
mechanism as described by Simon (1994), and because of the focus on double-loop
diagnostic-dialogic processing in the Diagnostic-Dialogic group, I formulated the
following hypotheses:
H3a Quality of commentary will decline (i.e. the percentage of functional
comments will increase) in the Traditional group relative to the Laissez-faire group.
H3b Quality of commentary will increase (i.e., the percentage of systemic
comments will increase) in the Diagnostic-Dialogic group relative to both the
Traditional and the Laissez-faire groups.
Hypothesis 4: Profitability Alignment
As shown in Figure 1, the final calculation in each customer proposal provides the
profitability rate that WholeCo can anticipate for a particular customer and product
line as a result of customer acceptance. This profitability metric is calculated as the
dollar value of proposed earnings divided by the dollar value of proposed net sales
Table 6. Comment Coding Matrix
Unclassified Functional Systemic Total
Laissez-faire 119 205 138 462
Traditional 78 165 90 333
Diagnostic-Dialogic 63 196 152 411
Total 260 566 380 1,206
59
and is expressed as a percentage of proposed net sales. For example, if net sales were
$500,000 and profit were $50,000, the profitability rate would be 10% (e.g., $50,000 /
$500,000 = 10%). Proposed net sales result from the deduction of proposed program
allowances for the customer (e.g., rebates, warranty allowances) from the total
amount of proposed sales to the customer.
Profitability alignment is defined as the proposal profitability rate minus the planned
profitability rate. A negative result represents profitability nonalignment, and a null or
positive result represents profitability alignment because of WholeCo’s focus on
meeting or exceeding its planned profitability rates. For example, a proposal with
$100,000 profit and $1,000,000 revenue represents a proposal profitability rate of
10%. If the planned rate is 12%, this proposal would exhibit nonalignment since 10%
minus 12% = -2%, a negative result. If the planned rate is 9%, however, this proposal
would exhibit alignment since 10% minus 9% = 1%, a positive result.
Because of the single-loop diagnostic process and the absence of an interactive
mechanism within the Traditional group, it is likely that each function will focus on
maximizing its own functional results. For example, Marketing and Sales may look
for lower pricing in order to promote acceptance of a product, and Operations may
conservatively estimate product costs in order to ensure a greater tolerance for
supplier pricing variations. Similarly, Finance may overestimate customer rebates and
60
programs in order to assure a sufficient war chest to pay for these programs in the
future. Each of these examples would lead to a proposal’s lower overall profitability
rate. However, the double-loop diagnostic-dialogic process within the Diagnostic-
Dialogic group provides an interactive mechanism with a focus on the whole system
rather than on any one function and thus should support higher profitability
alignment.
Since comparative profitability data was available only within the Traditional and
Diagnostic-Dialogic groups, I formed the following hypothesis:
H4 Profitability alignment will increase in the Diagnostic-Dialogic group relative
to the Traditional group.
Hypothesis 5: Customer Acceptance
As described in the section introducing the longitudinal dataset, data identifying
whether a proposal was accepted or rejected by the WholeCo customer (“customer
acceptance”) was available on proposals for three unique product lines constituting a
subset of 197 records. To quantify customer acceptance for hypotheses testing, each
record was dummy coded to identify whether the proposal was accepted or rejected
by the customer, with “1” representing acceptance and “0” representing rejection.
Because of the tendency of the Traditional intervention to promote greater diagnostic
controls and internal focus relative to the Laissez-faire approach, while lacking a
61
dialogic mechanism, and because of the tendency of the Diagnostic-Dialogic
intervention to promote greater systems understanding and greater generative
processes through its diagnostic-dialogic characteristics, I formulated the following
hypotheses:
H5a Odds of customer acceptance will decline in the Traditional group relative to
the Laissez-faire group.
H5b Odds of customer acceptance will increase in the Diagnostic-Dialogic group
relative to the Laissez-faire group.
Control Variables As Stone (1978) has pointed out, a researcher’s ability to exercise control over
confounding variables in a field study is relatively low; however, some control is
necessary if the research goal is to test a priori causal hypotheses. To this end, the
following variables are established and operationalized as control variables when
testing the effect of the quasi-experimental variables on the outcome variables.
General Control Variables
Company
A dummy variable, “Company,” was established to control for potential differences
among the two operating units of WholeCo by identifying the operating unit for each
record. As described earlier in the section on research setting, SpecialtyCo (coded 0)
is characterized by specialty products with high brand recognition, and
62
CommodityCo (coded 1) is characterized by commodity products with low brand
importance.
Electronic
During the longitudinal segment in the Diagnostic-Dialogic quasi-experimental
group, but post-intervention, customer proposals were converted from a manual
process to a computerized process. Although the process flow and participants did not
change with this conversion, a dummy control variable “Electronic” was established
to identify those records occurring within the manual process (coded 0) and those
records occurring within the electronic process (coded 1).
Training
On June 15, 2011 (the approximate midpoint of the Diagnostic-Dialogic segment), a
special training session occurred to increase participant understanding of diagnostic-
dialogic concepts employed through accounting metrics. A dummy control variable
“Training” was established to identify those records occurring prior to training (coded
0) and those occurring after training (coded 1) in order to control for the effect of this
training session on the outcome variables.
Specific Control Variables
Customer
Recognizing the potential for specific customers to have some impact on outcome
variables, dummy control variables were established to categorize each of WholeCo’s
top customers, as shown in Table 7. Values for each of these variables were assigned
63
by identifying the customer for each record, assigning a value of “1” to the
corresponding variable for that customer, and assigning a value of “0” to all other
customer variables for that record.
Table 7. WholeCo Customer Base
Product Group
WholeCo operates in multiple product categories, and the characteristics of any
specific product category may also impact outcome variables. In order to control for
effects of specific product groups on the outcome variables, dummy control variables
were established for each product group as shown in Table 8. Values for each of these
variables were assigned by identifying the product group for each record, assigning a
value of “1” to the corresponding product group control variable, and assigning a
value of “0” to all other product group control variables for that record.
Table 8. WholeCo Product Groups
Product A Product F
Product B Product G
Product C Product H
Product D Product I
Product E Other
Hardware 1 Home Center 1 Mass Merchant 3
Hardware 2 Home Center 2 Mass Merchant 4
Hardware 3 Internet Warehouse Club 1
Hardware 4 Mass Merchant 1 Warehouse Club 2
Hardware 5 Mass Merchant 2 Other
64
A summary of the quasi-experimental variables, outcome variables, general control
variables, and specific control variables is provided in Table 9.
Table 9: Summary of Operationalized Variables
Name Group Type Definition
Customer proposals occurring between
June 10, 2008 and February 26, 2010
without a diagnostic or diagnostic-dialogic
worksheet
Customer proposals occurring between
March 3, 2010 and February 28, 2011
with a diagnostic worksheet
Customer proposals occurring between
March 3, 2011 and October 25, 2011
with a diagnostic-dialogic worksheet
Number of days between initiation of and
final approval or rejection of customer proposal
Quantity of Aggregate number of written comments in
Commentary a customer proposal
Quality Number of systemic comments divided by
of total number of systemic and functional
Commentary comments
Profitability Proposal profitability rate minus
Alignment planned profitability rate
Customer Acceptance or rejection of WholeCo
Acceptance proposal by customer
General Identification of company as
Control SpecialtyCo or CommodityCo
General Identification of whether customer proposal
Control occurred with a manual or electronic process
General Identification of whether customer proposal
Control occurred before or after training session
Specific Identification of WholeCo customer to which
Control customer proposal was made
Specific Identification of product group for which
Control customer proposal was made Product Group Categorical
Binary
Binary
Binary
Training Binary
Customer Categorical
Company
Electronic
Outcome
Categorical
Categorical
Categorical
Interval
Ratio
Ratio
Ratio
OutcomeCompletion Time
Outcome
Outcome
Outcome
Quasi-experimentalLaissez-faire
Quasi-experimentalTraditional
Quasi-experimentalDiagnostic-Dialogic
65
Data Collection
The retrospective longitudinal data for this study was provided by the WholeCo
organization in electronic spreadsheet format in order to deidentify the data with
respect to individual participants. Data was gathered from archived records of each
customer proposal that had been initiated in WholeCo from June 10, 2008, through
October 25, 2011. Each record included data fields for each of the quasi-
experimental, outcome, and control variables as extracted from the customer proposal
forms (see Figures 2, 3, 6, and 7) and as described in the earlier sections.
The initial data file obtained from WholeCo consisted of 1,139 records, which were
reviewed for validity within the research framework. During this review, one record
was deleted for having insufficient data within the data fields listed above; five
records were deleted for having initiation dates prior to June 10, 2008; and three
records required correction of the initiation date, which was confirmed with
WholeCo. The resulting 1,133 records were used to establish the research database
and develop values for the operationalized variables discussed above.
All records were stratified into the quasi-experimental groups of Laissez-faire,
Traditional, and Diagnostic-Dialogic. But during this stratification, I determined that
some records had to be excluded from hypotheses testing because of noncompliance
within their stratified group. This noncompliance was generally found in records that
occurred during transitional periods early into a quasi-experimental group’s time
66
period and included Traditional group records that did not include the requisite
diagnostic comparative worksheet (Table 3) and Diagnostic-Dialogic group records
that did not include the requisite diagnostic-dialogic model worksheet (Figure 5).
These records were excluded in order to retain consistency among the record sets
within each quasi-experimental group. After excluding noncompliant records, the
research database consisted of 929 records, or 81.6% of the original record set
provided by WholeCo. The stratified record structure is shown in Table 10.
Table 10: Stratified Record Set
Quasi-Experimental Group Record Count
Laissez-faire 446 Traditional (diagnostic) 250
Diagnostic-Dialogic 233
Total Record Set 929
Longitudinal Dataset
The final dataset for this field study included the following data for each proposal:
Dates of initiation and completion (see Figure 2)
Written commentary by functional initiators and reviewers (see Figure 2)
Proposal profitability for WholeCo (see Figure 1).
Additionally, a subset of 197 of the archived records identified whether the retail
customer had accepted or rejected the proposal. This subset of proposals was defined
by three unique WholeCo product lines for which customer acceptance or rejection
was identifiable.
67
Testing Methods Each of the a priori hypotheses described earlier were quantitatively tested through
the application of one of three different statistical techniques: survival analysis,
generalized estimating equations, or logistic regression; each used the statistical
package Stata version 12. These statistical techniques, along with justification for
their application to specific hypotheses, are summarized below. A summary matrix
of hypotheses, techniques, and confirmatory measures appears in Table 11.
Survival Analysis
Survival analysis comprises specialized statistical methods that deal with issues
common to longitudinal data sets, including the lack of normally distributed data and
the potential for censored (incomplete) observations (Clark, Bradburn, Love, &
Altman, 2003; Hosmer, Lemeshow, & May, 2008). Applied as a regression model,
survival analysis may be used to analyze hazard rate, which in this study would be the
instantaneous potential of a customer proposal reaching completion at a specific point
in time given that it had not already completed (Clark, et al., 2003; Hosmer, et al.,
2008; Kleinbaum & Klein, 2005). However, when the outcome of interest is the
effect of covariates on multiplicatively increasing or decreasing the time duration to
an event occurrence (i.e., completion time), survival analysis is applied as an
accelerated failure time (AFT) model (Bradburn, Clark, Love, & Altman, 2003a;
Kleinbaum & Klein, 2005; Lawless, 2003).
68
Fitting AFT Models
AFT models are parametric survival models. Thus, in order to fit an AFT model, the
distribution of the baseline survival function must be specified in order to calculate
the proportional change effect in survival time resulting from a one unit change in
explanatory variable (Bradburn, et al., 2003a; Hosmer, et al., 2008; Kleinbaum &
Klein, 2005). Common distribution specifications are exponential, Weibull, log-
normal, and log-logistic (Bradburn, et al., 2003a; Hosmer, et al., 2008; Kleinbaum &
Klein, 2005); however, choosing the correct distribution can be difficult and is
generally accomplished through testing the goodness of fit of multiple, alternative
models.
Testing Goodness of Fit
Evaluating the goodness of fit of an AFT model may be accomplished by comparing
multiple AFT models using various survival function distribution assumptions. A
common test metric for this comparison is Akaike’s Information Criterion (AIC),
which tests an AFT model’s viability by trading off the model’s likelihood with its
complexity. The model demonstrating the smallest AIC is demonstrated as the best-
fitting model among the alternative models (Bradburn, Clark, Love, & Altman,
2003b; Hosmer, et al., 2008; Kleinbaum & Klein, 2005).
Method for Hypothesis Testing
Hypotheses 1a and 1b, which predict the effect of the quasi-experimental variables on
the outcome variable “completion time,” were tested by using an AFT model with
AIC testing for goodness of fit. AFT is appropriate for testing these hypotheses since
69
the outcome of interest, completion time, is a time duration to event occurrence, and
preliminary univariate testing demonstrates that the outcome variable is not normally
distributed but is highly right skewed as is typical of survival data. Multiple AFT
models are compared by using the AIC test statistic in order to control for potential
misspecification of the underlying survival function distribution and to determine the
best-fitting model.
Generalized Estimating Equations
Liang and Zeger (1986) and Zeger and Liang (1986) proposed general estimating
equations (GEEs) as an extension of general linear models (GLMs) in order to more
efficiently analyze longitudinal data sets that commonly contain both non-Gaussian
distributions and the potential for within-variable correlation of response variables
(Ballinger, 2004; Hardin & Hilbe, 2003). When applying GEE models for hypothesis
testing, the researcher must first choose whether the appropriate unit of analysis is
subject specific (SS-GEE), wherein coefficients are interpreted in terms of individual
impact (Hardin & Hilbe, 2003), or is population averaged (PA-GEE), wherein each
unit increase in a covariate indicates how much the average response is affected over
the population (Ballinger, 2004; Hardin & Hilbe, 2003). Since the outcomes of
interest in this study are based on average responses across the population of
customer proposals spread over several interventions and time segments, PA-GEE
models are the appropriate unit of analysis for this study.
70
Fitting PA-GEE Models
Fitting PA-GEE models for hypothesis testing requires selection of a link function
and outcome variable distribution, along with specification of a working correlation
structure for the outcome variable. Various link functions may be selected dependent
on an assumed distribution of the response variable and on what the researcher wishes
covariate coefficients to demonstrate (Ballinger, 2004). For example, a log link may
be used to test a count based response variable (Poisson distribution) with the intent
that the resulting covariate coefficients represent the predicted change in the log of
outcome variable mean for each one unit change in the covariate. Some typical
combinations of link functions and distribution specifications available for GEE
analysis are shown in Appendix A.
Specification of the working correlation in a PA-GEE model provides an estimate of
how response variable outcomes may be correlated within the response set. But
although specification is required, the PA-GEE method is robust with respect to
violation of the working correlation assumption and provides model consistency even
with misspecification (Ballinger, 2004; Cui & Guoqi, 2007; Hardin & Hilbe, 2003;
Liang & Zeger, 1986; Zeger & Liang, 1986). Although not an exhaustive list,
examples of working correlation specifications are independent, which assumes no
correlation among responses; exchangeable, which assumes a common but non-time
ordered correlation among responses; and unstructured, which does not impose a
71
specific correlation matrix structure (Ballinger, 2004; Hardin & Hilbe, 2003; Liang &
Zeger, 1986; Zeger & Liang, 1986).
Testing Goodness of Fit
Relatively little literature describes appropriate methods for testing the goodness-of-
fit of a PA-GEE model (Ballinger, 2004; Cui & Guoqi, 2007; Pan, 2001). However,
Pan (2001) proposes an extension of AIC, which he identifies as quasi-likelihood
under the independence model criterion (QIC), in order to test PA-GEE models for an
optimal subset of covariates as well as the best working correlation structure for
model efficiency (Cui & Guoqi, 2007). Similar to use of the AIC metric, the model
exhibiting the smallest QIC metric from among alternative models is regarded as the
best-fitting PA-GEE model (Cui & Guoqi, 2007; Pan, 2001).
Method for Hypothesis Testing
PA-GEE models were used to test hypotheses 2a, 2b, 3a, 3b, and 4, which predict the
effects of the quasi-experimental variables on the outcome variables quantity of
commentary, quality of commentary, and profitability alignment. PA-GEE analysis is
appropriate for testing these hypotheses since the underlying response variables are
longitudinal in nature with characteristics common to longitudinal data sets: non-
normal distribution, multiple observations collected over time, and the potential for
within-variable correlation arising from a common organizational data pool.
Additionally, the outcomes of interest were drawn from the average responses across
72
the customer proposal population spread over several intervention and time segments,
making population average the appropriate unit of analysis.
Multiple PA-GEE models were fitted for each response variable in order to control
for potential misspecification of distribution or link function and to determine the
most appropriate covariate subset. Once fitted, these PA-GEE models were evaluated
for goodness of fit through application of QIC posttesting (Cui & Guoqi, 2007; Pan,
2001).
Logistic Regression
Logistic regression (LR) is a regression-based statistical analysis commonly applied
to retrospective data in which the variable of interest is represented by a dichotomous
outcome (Kleinbaum & Klein, 2010). It is particularly well-suited for working with
those dichotomous outcome variables in which the desired result is outcome
probability, since LR provides results that range between the values of 0 and 1
(Kleinbaum & Klein, 2010), and it can be set to provide odds ratios (ORs) that are
most appropriate in retrospective studies, such as this one. Rather than providing the
absolute probability that an outcome will occur, ORs provide the researcher with
estimates of the comparative odds of an outcome occurring in one situation versus the
odds of that outcome occurring in another situation, as reflected by the independent
variable. In other words, the researcher can determine from the OR how many times
more or less likely a particular binary outcome is to occur given a particular predictor
73
value (e.g., a driver might be X times more likely to be stopped for speeding if
driving a red car versus a white car).
Because of the retrospective and dichotomous nature of the customer acceptance
outcome variable, hypotheses 5a and 5b were tested through LR with the reporting of
OR. Multiple LR models were fitted and evaluated for goodness of fit by comparison
with their resulting pseudo-R 2 values (Mittlbö & Schemper, 1996).
Summary This research study was conducted as a retrospective longitudinal study supporting a
single-case design. Structured from archival data collected for June 2008 through
October 2011, three quasi-experimental groups were operationalized: Laissez-faire,
Traditional (diagnostic), and Diagnostic-Dialogic. The retrospective longitudinal
nature of this study provided consistent measurement of five outcome variables:
completion time, quantity of commentary, quality of commentary, profitability
alignment, and customer acceptance. Hypotheses were developed with respect to the
predicted effect of the quasi-experimental variables on these outcome variables, as
summarized in Table 11. The appropriate methods for testing these hypotheses were
established as the AFT form of survival analysis, followed by AIC metric evaluation,
the PA-GEE form of generalized estimating equations, followed by QIC metric
evaluation, and LR, followed by pseudo R 2 evaluation.
74
Table 11. Summary of Hypotheses and Testing Methods
Outcome
Hypotheses
Test
Fit
Completion Time H1a
H1b
Traditional < L-F
D-D < Traditional
and D-D < L-F
AFT Lowest
AIC
Quantity of Commentary H2a
H2b
Traditional > L-F
D-D > L-F and
D-D > Traditional
PA-GEE Lowest
QIC
Quality of Commentary H3a
H3b
Traditional < L-F
D-D > L-F and
D-D > Traditional
PA-GEE Lowest
QIC
Profitability Alignment H4 D-D > Traditional PA-GEE Lowest
QIC
Customer Acceptance H5a
H5b
Traditional < L-F
D-D > L-F and
D-D > Traditional
LR Highest
Pseudo-
R 2
AIC = Akaike Information Criterion AFT = Accelerated Failure Time
D-D = Diagnostic – Dialogic L-F = Laissez-faire
PA-GEE = Population Averaged Generalized Estimating Equations
LR = Logistic Regression (with Odds Ratios)
QIC = Quasi-likelihood Information Criterion
75
Chapter 4: Data Analysis and Results Chapter 4 provides data analysis and results of hypotheses testing in three sections
along with a summary of findings. The first section covers the process for verifying
data accuracy. The second section provides univariate review of covariates and
outcome variables through descriptive statistics and correlations among selected
variables. The third section describes the results for each hypothesis tested. The
chapter concludes with a summary of findings.
Data Verification
The longitudinal dataset for this research study comprises 929 archived WholeCo
customer proposals for the retrospective period from June 10, 2008, through October
25, 2011. As is common in longitudinal datasets, some values are missing among
individual observations for some variables. Each record with missing values was
reviewed with WholeCo to determine that no error had occurred in transcription and
that it was reasonable for the record to experience missing values (e.g., missing
values for quality of commentary occurred if no comments existed for the record). It
was determined upon review that missing values were either missing for all records in
a time segment or had no discernible pattern and were missing completely at random
(MCAR) (Zeger & Liang, 1986). For example, all records for the Laissez-faire group
were missing values for profitability alignment since no observations of profitability
were collected during this time period; however, all other variables contained
observations within each group, and missing values were MCAR. Records with
76
missing values for a specific outcome variable were excluded from hypothesis testing
for that specific outcome variable.
After confirmation of missing values, z-scores were calculated for each of the
outcome variables: completion time, quantity of commentary, quality of commentary,
profitability alignment, and customer acceptance, in order to identify outliers at the
±3.3 z-score level. This process helped to further confirm data entry and determine
disposition of outlier records. The results of this process are shown in Appendix B.
Descriptive Statistics and Correlations
Descriptive statistics, including numbers of observations and missing records, are
shown in Table 12 for all quasi-experimental, general control, and outcome variables.
As discussed in the prior section, missing values were determined as either missing
for all records in a time segment or as MCAR.
Table 13 provides correlations among the quasi-experimental, outcome, and general
control variables, as well as selected specific control variables. Because of significant
differences in number of observations for some variables, resulting from records with
missing information as described earlier, correlations were calculated by using
pairwise deletion in order to include the maximum number of observations for each
correlation.
77
Table 12. Descriptive Statistics
Variable Obs. Missing Mean Std. Dev. Min Max
Laissez-faire 929 0 0.48009 0.49987 0 1
Traditional 929 0 0.26911 0.44373 0 1
Diagnostic-Dialogic 929 0 0.25081 0.43371 0 1
Completion Time 929 0 15.028 16.5778 2 173
Quantity of Commentary 929 0 1.46609 1.14243 0 6
Quality of Commentary 699 230 0.3037 0.39678 0 1
Profitability Alignment 363 566 0.01514 0.05732 -0.1786 0.36569
Customer Acceptance 197 732 0.39086 0.48919 0 1
Company 929 0 0.33369 0.47179 0 1
Electronic 929 0 0.22174 0.41564 0 1
Training 929 0 0.14639 0.35369 0 1
78
T a b
le 1
3 . C
o r r e la
ti o n
s fo
r S
e le
c te
d V
a r ia
b le
s
* p
< . 0 5
T im
e
Q u a n ti ty
o f
C o
m m
e n ta
ry
Q u a lit
y o
f
C o
m m
e n ta
ry
P ro
fi ta
b ili
ty
A lig
n m
e n t
A c c e p
ta n c e
L a is
se z-
fa ir e
T ra
d it io
n a l
D ia
g n o
st ic
-
D ia
lo g ic
C o
m p
a n y
E le
c tr
o n ic
T ra
in in
g
C m
d ty
P ro
d 2
C m
d ty
P ro
d 1
S p
c lt y
P ro
d 1
W h se
C lu
b 1
H o m
e
C e n te
r 1
H d
w re
2 H
d w
re 4
W h se
C lu
b 2
M a ss
M e rc
h a n t 2
T im
e 1
Q u a n ti ty
-0 .0
3 7
5 1
Q u a lit
y -0
.0 4 9
5 0 .0
5 7 8
1
P ro
fi ta
b ili
ty A
lig m
e n t
-0 .0
3 8
0 -0
.0 2 5
0 -0
.0 0 1
6 1
A c c e p
ta n c e
-0 .1
1 8
4 0 .0
2 3 9
-0 .0
5 1
3 0 .1
1 2 7
1
L a is
se z-
fa ir e
0 .1
1 5 0
* -0
.2 2 2
7 *
-0 .0
2 3
7 0 .0
0 0 0
-0 .0
3 5
1 1
T ra
d it io
n a l
-0 .0
0 1
3 -0
.0 1 1
5 -0
.0 8 9
7 *
-0 .1
6 3
5 *
-0 .1
2 7
3 -0
.5 8 6
5 *
1
D ia
g n o
st ic
-D ia
lo g ic
-0 .1
3 1
2 *
0 .2
6 9 3
* 0 .1
1 4 6
* 0 .1
6 3 5
* 0 .1
9 2 6
* -0
.5 5 1
8 *
-0 .3
5 1
8 *
1
C o
m p
a n y
0 .1
3 2 7
* -0
.2 0 9
9 *
0 .0
0 4 2
0 .0
4 2 3
-0 .0
5 7
9 -0
.0 3 6
2 0 .0
1 9 4
0 .0
2 1 8
1
E le
c tr
o n ic
-0 .1
1 6
0 *
0 .2
6 1 5
* 0 .1
2 1 0
* 0 .1
3 8 9
* 0 .2
1 2 3
* -0
.5 0 9
2 *
-0 .3
2 4
6 *
0 .9
2 2 7
* 0 .0
2 2 9
1
T ra
in in
g -0
.1 4 4
8 *
0 .1
6 6 8
* 0 .0
8 1 9
* 0 .1
3 3 1
* 0 .1
4 3 1
* -0
.3 9 5
2 *
-0 .2
5 2
0 *
0 .7
1 6 2
* 0 .0
3 5 8
0 .7
7 6 2
* 1
C m
d ty
P ro
d 2
-0 .0
2 6
9 -0
.0 0 6
6 0 .0
2 6 8
-0 .0
3 0
4 0 .0
0 3 0
-0 .0
6 5
8 *
0 .0
2 1 4
0 .0
5 4 0
0 .1
3 1 0
* 0 .0
6 2 8
0 .0
6 0 6
1
C m
d ty
P ro
d 1
-0 .0
2 5
0 -0
.1 2 8
1 *
-0 .0
6 1
7 -0
.0 6 1
6 0 .0
0 0 0
-0 .0
1 8
8 -0
.0 0 6
4 0 .0
2 8 4
0 .5
0 4 3
* 0 .0
4 6 3
0 .0
2 5 0
-0 .0
3 3
1 1
S p
c lt y P
ro d
1 -0
.0 9 5
5 *
0 .1
1 4 3
* 0 .0
0 7 6
-0 .0
9 6
6 0 .0
0 0 0
0 .0
4 5 8
-0 .0
3 4
0 -0
.0 1 7
9 -0
.5 5 4
4 *
-0 .0
1 2
0 -0
.0 1 6
7 -0
.0 7 2
6 *
-0 .2
7 9
6 *
1
W h se
C lu
b 1
0 .0
2 6 2
0 .0
2 6 5
-0 .0
1 8
1 0 .0
2 5 0
-0 .1
0 9
6 0 .0
4 0 5
0 .0
1 3 2
-0 .0
6 0
4 -0
.1 0 4
8 *
-0 .0
6 1
1 -0
.0 7 2
0 *
-0 .0
1 9
6 -0
.0 7 5
4 *
-0 .0
1 3
0 1
H o
m e C
e n te
r 1
-0 .0
3 9
6 -0
.0 2 1
9 -0
.0 1 7
4 0 .0
2 8 8
-0 .0
3 3
8 -0
.0 5 6
1 0 .0
4 4 9
0 .0
1 8 6
0 .1
2 3 7
* 0 .0
0 3 8
0 .0
3 6 7
0 .0
1 3 2
0 .0
7 1 7
* -0
.0 8 1
8 *
-0 .0
6 4
5 *
1
H a rd
w a re
2 -0
.0 0 5
7 0 .0
1 1 3
0 .0
6 3 1
0 .1
7 5 1
* 0 .1
2 7 2
-0 .0
9 6
2 *
-0 .0
0 4
4 0 .1
1 5 7
* 0 .1
7 9 2
* 0 .0
8 2 0
* 0 .0
6 9 6
* -0
.0 2 2
7 0 .0
4 3 8
-0 .0
7 7
4 *
-0 .0
5 1
7 -0
.0 7 4
8 *
1
H a rd
w a re
4 -0
.0 4 3
1 -0
.0 4 2
5 -0
.0 3 2
5 0 .0
7 4 0
0 .1
4 6 5
* -0
.0 0 6
1 -0
.0 3 7
4 0 .0
4 5 5
0 .0
7 5 5
* 0 .0
6 6 0
* 0 .0
7 3 1
* 0 .1
1 5 3
* 0 .0
9 4 7
* 0 .0
0 3 9
-0 .0
5 6
2 -0
.0 8 1
3 *
-0 .0
6 5
2 *
1
W h se
C lu
b 2
0 .0
0 1 5
0 .1
3 6 3
* -0
.0 1 8
6 -0
.0 5 3
2 -0
.1 6 0
5 *
-0 .0
2 2
5 0 .0
9 1 5
* -0
.0 6 8
2 *
-0 .0
8 4
0 *
-0 .0
6 3
0 -0
.0 4 8
9 -0
.0 1 1
0 -0
.0 4 2
4 -0
.0 9 2
8 *
-0 .0
2 5
0 -0
.0 3 6
2 -0
.0 2 9
0 -0
.0 3 1 6
1
M a ss
M e rc
h a n t 2
-0 .0
3 1
8 -0
.0 0 7
6 -0
.0 6 8
3 -0
.0 3 4
1 -0
.1 3 7
8 0 .0
0 8 1
0 .0
2 2 4
-0 .0
3 2
4 -0
.1 2 5
3 *
-0 .0
2 0
3 -0
.0 8 0
0 *
-0 .0
1 8
0 -0
.0 6 9
3 *
0 .0
5 9 8
-0 .0
4 1
0 -0
.0 5 9
3 -0
.0 4 7
5 -0
.0 5 1 7
-0 .0
2 3
0 1
79
Results of Hypotheses Testing
The following sections provide the results of hypotheses testing for each of the
hypotheses developed in Chapter 3. This chapter concludes with a summary of testing
results for each hypothesis.
Hypotheses 1a and 1b: Effect on Completion Time
H1a Completion time will decrease in the Traditional group relative to the
Laissez-faire group.
H1b Completion time will decrease in the Diagnostic-Dialogic group relative to
both the Traditional and Laissez-faire groups.
As described in Chapter 3, completion time is the time period (in days) from initiation
of a customer proposal until its completion. AFT survival models were used for
testing Hypotheses H1a and H1b because distribution is not normal but is highly right
skewed (see Figure 6), and the outcome of interest is time duration.
The final step for testing H1a and H1b is the generation of AFT models, as shown in
Table 14. These multiple AFT models were generated for comparative, goodness-of-
fit testing as described in Chapter 3. They proceed from hypothesized predictors, to
hypothesized predictors with general control covariates, to hypothesized predictors
with both general and specific control covariates.
80
Figure 6. Histogram of Completion Time
Model 3 (see Table 14) provides the best-fitting AFT model because of its lower AIC
value (2256.50). The log-form, negative coefficient for the Traditional quasi-
experimental group (log = -0.12, p = .07) indicates that completion time decreases in
the Traditional group compared with the Laissez-faire group, thus supporting
Hypothesis 1a. The greater negative log-form coefficient for the Diagnostic-Dialogic
0
.0 2
.0 4
.0 6
D e n si ty
0 50 100 150 200
Days
81
quasi-experimental group (log = -0.24, p = .10) indicates that completion time
decreases in the Diagnostic-Dialogic group by a greater amount than it decreases in
the Traditional group, thus supporting Hypothesis 1b.
Table 14. AFT Models for Completion Time
Model 1 Model 2 Model 3
Observations 929 929 929
Intercept 2.32 *** 2.28 *** 2.36 ***
Predictors
Laissez-faire base base base
Traditional -0.11 * -0.12 * -0.12 *
Diagnostic-Dialogic -0.16 ** -0.24 -0.24
General Controls
Training -0.38 *** -0.36 ***
Company 0.19 *** 0.15 **
Electronic 0.33 * 0.33 *
Specific Controls
Customers
Home Center 1 -0.20 **
Hardware 4 -0.16
Product Lines
Commodity Prod 2 -0.48
Specialty Prod 1 -0.12 *
Model Fit
AIC 2275.42 2261.48 2256.50
Model Type
Distribution
* p < .10 ** p < .05 *** p < .01
AFT
Gamma
82
Although not included in the development of Hypotheses 1a and 1b, the
unhypothesized results for two of the general controls in the selected AFT model,
Company and Electronic, are worthy of note. Results for these control variables
indicate that completion time increases (log = 0.15, p = .03) as Company increments
from SpecialtyCo (coded 0) to CommodityCo (coded 1). The results also indicate
that, counterintuitively, completion time also increases (log = 0.33, p = .07) as the
customer process increments from a manual process (coded 0) to the electronic
process (coded 1). The hypothesized and unhypothesized results are discussed in
Chapter 5.
Hypotheses 2a and 2b: Effect on Quantity of Commentary
H2a Quantity of commentary will increase in the Traditional group relative to the
Laissez-faire group.
H2b Quantity of commentary will increase in the Diagnostic-Dialogic group
relative to both the Traditional and Laissez-faire groups.
As described in Chapter 3, quantity of commentary is the aggregate of coded
comments for each customer proposal. The related hypotheses, H2a and H2b, were
tested through fitting multiple PA-GEE models, as shown in Table 15. A Poisson
distribution was selected because the underlying quantity of commentary variable
consists of counts of commentary within each customer proposal. A log link function
was applied based on preliminary testing for best-fitting distribution and link
83
functions in a noncontrolled model. Each model also includes the time-based variable
completion time and specifies an unstructured working correlation because, as
discussed in Chapter 3, PA-GEE models are robust to misspecification of the working
correlation and the specific matrix structure of potential within-variable correlation is
unknown. The best-fitting model was determined by evaluating each model’s QIC
metric, with the lowest QIC metric indicating the best model.
Table 15. PA-GEE Models for Quantity of Commentary
Model 1 Model 2 Model 3
Observations 929 929 929
Intercept 0.18 *** 0.29 *** 0.27 ***
Predictors
Laissez-faire base base base
Traditional 0.19 *** 0.20 *** 0.18 ***
Diagnostic-Dialogic 0.51 *** 0.40 *** 0.41 ***
General Controls
Training -0.12 -0.12
Company -0.39 *** -0.37 ***
Electronic 0.21 0.21 *
Specific Controls
Customers
Warehouse Club 2 0.63 ***
Model Fit
QIC 399.55 382.84 379.12
Model Type
Distribution
Link Function
Working Correlation
* p < .10 ** p < .05 *** p < .01
Log
Unstructured
GEE
Poisson
84
Model 3 (see Table 15) provides the best-fitting PA-GEE model because of its lower
QIC value (379.12). The positive, log-form coefficient shown for the Traditional
group in Model 3 (log = 0.18, p < .01) indicates that the quantity of commentary
increases in the Traditional group as compared with the Laissez-faire group, thus
supporting Hypothesis 2a. The greater positive log-form coefficient shown for the
Diagnostic-Dialogic group in Model 3 (log = 0.41, p < .001) indicates that the
quantity of commentary increases in the Diagnostic-Dialogic group by a greater
amount than it increases in the Traditional group (0.41 > 0.18), thus supporting
Hypothesis 2b.
Although not included in the development of Hypotheses 2a and 2b, the
unhypothesized results for two of the general controls in the selected PA-GEE model
are worthy of note: Company and Training. PA-GEE results indicate that the quantity
of commentary decreases (log = -0.37 p < .001) as Company increments from
SpecialtyCo (coded 0) to CommodityCo (coded 1). Counterintuitively, although not
strictly statistically significant, the results also indicate that the quantity of
commentary decreases (log -0.12, p = .20) as the customer proposal process
increments from the process prior to Training (coded 0) to the process occurring after
Training (coded 1). Hypothesized and unhypothesized results are discussed in
Chapter 5.
85
Hypotheses 3a and 3b: Effect on Quality of Commentary
H3a Quality of commentary will decline (i.e., functional comments will increase)
in the Traditional group relative to the Laissez-faire group.
H3b Quality of commentary will increase (i.e., systemic comments will increase)
in the Diagnostic-Dialogic group relative to both the Traditional and Laissez-faire
groups.
The two hypotheses for quality of commentary, H3a and H3b, were tested through
fitting multiple PA-GEE models (see Table 16) with Gamma distribution and with a
log link function, since preliminary testing revealed this distribution and link function
combination as the best-fitting combination for testing the continuous variable,
quality of commentary, in a noncontrolled model. Each model also includes the time-
based variable, completion time, and specifies an unstructured working correlation
because, as discussed in Chapter 3, PA-GEE models are robust to misspecification of
the working correlation and the specific matrix structure of potential within-variable
correlation is unknown. The best-fitting model was determined by evaluating each
model’s QIC metric, with the lowest QIC metric indicating the best model. Model 2
(see Table 16) provides the best-fitting PA-GEE model because of its lower QIC
value of (-295.32).
86
Table 16. PA-GEE Models for Quality of Commentary
Models for quality of commentary proceed from hypothesized predictors to
hypothesized predictors with specific control covariates because model convergence
was not achievable for hypothesized predictors with general control covariates. The
negative log-form coefficient for the Traditional group (log = -0.13, p = .30)
directionally indicates that the degree of systemic commentary decreases while
Model 1 Model 2 (a)(b)
Observations 699 699
Intercept -1.23 *** -1.20 ***
Predictors
Laissez-faire base base
Traditional -0.15 -0.13
Diagnostic-Dialogic 0.25 ** 0.27 **
Specific Controls
Customers
Mass Merchant 2 -0.57
Product Lines
Commodity Prod 1 -0.42 *
Model Fit
QIC -287.07 -295.32
Model Type
Distribution
Link Function
Working Correlation
* p < .10 ** p < .05 *** p < .01
(a) Electronic control not applicable
(b) General controls result in a poorer fitting model and are excluded
Log
Unstructured
GEE
Gamma
87
functional commentary increases in the Traditional group as compared with the
Laissez-faire group; however, the high p-value may also indicate no effect from the
Traditional group. Thus, Hypothesis 3a is not supported. However, the positive log-
form coefficient for the Diagnostic-Dialogic group (log = 0.27, p = .02) indicates that
systemic commentary increases in the Diagnostic-Dialogic group as compared with
the Laissez-faire group, and by default the Traditional group, thus supporting
Hypothesis 3b. Results are discussed further in Chapter 5.
Hypothesis 4: Effect on Profitability Alignment
H4 Profitability alignment will increase in the Diagnostic-Dialogic group relative
to the Traditional group.
Hypothesis H4, regarding profitability alignment, was tested through fitting multiple
PA-GEE models (see Table 17) with Gamma distribution and a log link function
because preliminary testing revealed this distribution and link function combination
as the best-fitting combination for testing the continuous variable, profitability
alignment, in a noncontrolled model. Each model also includes the time-based
variable completion time and specifies an unstructured working correlation because,
as discussed in Chapter 3, PA-GEE models are robust to misspecification of the
working correlation and the specific matrix structure of potential within-variable
correlation is unknown.
88
Model 2 (see Table 17) provides the best-fitting PA-GEE model because of its lower
QIC value of (-2559.07). The positive log-form coefficient for the Diagnostic-
Dialogic quasi-experimental group (log = 1.71, p = .10) indicates that profitability
alignment increases in the Diagnostic-Dialogic group as compared with the
Traditional group, thus supporting Hypothesis 4.
Table 17. PA-GEE Models for Profitability Alignment
Model 1 Model 2 (a)(b)
Observations 363 363
Intercept -5.69 *** -5.69 ***
Predictors
Traditional base base
Diagnostic-Dialogic 1.90 * 1.71
General Controls
Training 0.29
Model Fit
QIC -2555.06 -2559.07
Model Type
Distribution
Link Function
Working Correlation
* p < .10 ** p < .05 *** p < .01
(a) Electronic control not applicable
(b) Profit Alignment tested only for Traditional
and Diagnostic-Dialogic categories.
Convergence not achieved on company or
specific control variables
Log
Independent
GEE
Gamma
89
An unhypothesized, but logical, finding in Model 2 is the impact of the Training
general control variable on profitability alignment. Although the coefficient for
Training is not statistically significant (p = .50), its positive, log-form coefficient (log
= 0.29) directionally indicates that profitability alignment increased as the customer
proposal process incremented from the process prior to Training (coded 0) to the
process after Training (coded 1). Hypothesized and unhypothesized results are
discussed in Chapter 5.
Hypotheses 5a and 5b: Customer Acceptance
H5a Odds of customer acceptance will decline in the Traditional group relative to
the Laissez-faire group.
H5b Odds of customer acceptance will increase in the Diagnostic-Dialogic group
relative to the Laissez-faire group.
Hypotheses H5a and H5b for customer acceptance were tested through LR with the
reporting of OR (see Table 18) because the outcome represented by customer
acceptance is either acceptance or rejection of the proposal by the WholeCo
customer, thus representing a win or loss. The best-fitting model was determined by
evaluating each model’s pseudo-R 2 .
90
Model 3 (see Table 18) provides the best-fitting LR model because of its higher
pseudo-R 2 value (0.11). Unlike the remaining outcome variables in which log-based
coefficients are reported, coefficients for the LR testing of customer acceptance
Table 18. LR Models for Customer Acceptance
are provided as OR. Results for the Traditional group (OR = 1.11, p = .80) indicate
that customer acceptance is approximately 1.11 times more likely in the Traditional
Model 1 Model 2 (a)
Model 3 (a)
Observations 197 197 197
Intercept 0.59 ** 0.59 ** 0.69
Predictors
Laissez-faire base base base
Traditional 0.77 0.84 1.11
Diagnostic-Dialogic 2.27 ** 2.17 3.18 **
General Controls
Training 1.31 0.85
Company 0.68 0.38 *
Specific Controls
Customers
Warehouse Club 1 0.32 *
Warehouse Club 2 0.12 *
Hardware 2 4.22
Mass Merchant 2 0.09 **
Model Fit
R-square 0.03 0.03 0.11
Model Type
* p < .10 ** p < .05 *** p < .01
(a) Electronic control not applicable
Logistic Regression - Odds Ratio
91
group than in the Laissez-faire group; however, the high p-value demonstrates
statistical insignificance, thus indicating that there may be no causal impact from the
Traditional group. Hypothesis 5a predicted that the odds of customer acceptance
would decline in the Traditional group; therefore Hypothesis 5a is rejected. Results
for the Diagnostic-Dialogic group (OR = 3.18, p = .03) indicate that customer
acceptance is over 3 times more likely in the Diagnostic-Dialogic segment than in the
Laissez-faire segment, thus supporting Hypothesis 5b.
Although not included in the development of Hypotheses 5a and 5b, the
unhypothesized results for one of the specific controls (i.e., Customers, representing
specific WholeCo customers) are noteworthy. This control was used in the LR model
to test customer acceptance. Results indicate that one particular customer, denoted as
Hardware 2, had a significant impact on the outcome of customer acceptance. Based
on the odds ratio for Hardware 2 (OR = 4.22, p = .10), customer proposals are over 4
times more likely to be accepted if the proposal is made to the customer denoted as
Hardware 2. Hypothesized and unhypothesized results are discussed in Chapter 5.
Summary of Findings
Various hypotheses related to the impact of Traditional (diagnostic) and Diagnostic-
Dialogic interventions on completion time, quantity of commentary, quality of
commentary, profitability alignment, and customer acceptance were tested through
AFT survival analysis, PA-GEE models, and LR models. The results of hypotheses
92
testing are summarized in Table 19, and the results of statistical models testing are in
Table 20. Discussion and implications of these results are covered in Chapter 5.
Table 19. Results of Hypotheses Testing
Outcome Hypotheses Result
Completion Time H1a
H1b
Traditional < L-F
D-D < L-F &
D-D < Traditional
Supported
Supported
Quantity of Commentary H2a
H2b
Traditional > L-F
D-D > L-F &
D-D > Traditional
Supported
Supported
Quality of Commentary H3a
H3b
Traditional < L-F
D-D > L-F &
D-D > Traditional
Not supported
(directional)
Supported
Profitability Alignment H4 D-D > Traditional Supported
Customer Acceptance H5a
H5b
Traditional < L-F
D-D > L-F &
D-D > Traditional
Rejected
Supported
L-F = Laissez-faire
D-D = Diagnostic-Dialogic
93
T a b
le 2
0 . R
e su
lt s
o f
A F
T ,
P A
-G E
E a
n d
L R
T e st
in g
M o
d e l 1
M o
d e l 2
M o
d e l 3
M o
d e l 1
M o
d e l 2
M o
d e l 3
M o
d e l 1
M o
d e l 2
M o
d e l 1
M o
d e l 2
M o
d e l 1
M o
d e l 2
M o
d e l 3
O b
se rv
a ti o
n s
9 2 9
9 2 9
9 2 9
9 2 9
9 2 9
9 2 9
6 9 9
6 9 9
3 6 3
3 6 3
1 9 7
1 9 7
1 9 7
In te
rc e p
t 2 .3
2 * * *
2 .2
8 * * *
2 .3
6 * * *
0 .1
8 * * *
0 .2
9 * * *
0 .2
7 * * *
-1 .2
3 * * *
-1 .2
0 * * *
-5 .6
9 * * *
-5 .6
9 * * *
0 .5
9 * *
0 .5
9 * *
0 .6
9
P re
d ic
to rs
L a is
se z-
fa ir e
b a se
b a se
b a se
b a se
b a se
b a se
b a se
b a se
b a se
b a se
b a se
T ra
d it io
n a l
-0 .1
1 *
-0 .1
2 *
-0 .1
2 *
0 .1
9 * * *
0 .2
0 * * *
0 .1
8 * * *
-0 .1
5 -0
.1 3
b a se
b a se
0 .7
7 0 .8
4 1 .1
1
D ia
g n o
st ic
-D ia
lo g ic
-0 .1
6 * *
-0 .2
4 -0
.2 4
0 .5
1 * * *
0 .4
0 * * *
0 .4
1 * * *
0 .2
5 * *
0 .2
7 * *
1 .9
0 *
1 .7
1 2 .2
7 * *
2 .1
7 3 .1
8 * *
G e n e ra
l C
o n tr
o ls
T ra
in in
g -0
.3 8
* * *
-0 .3
6 * * *
-0 .1
2 -0
.1 2
0 .2
9 1 .3
1 0 .8
5
C o
m p
a n y
0 .1
9 * * *
0 .1
5 * *
-0 .3
9 * * *
-0 .3
7 * * *
0 .6
8 0 .3
8 *
E le
c tr
o n ic
0 .3
3 *
0 .3
3 *
0 .2
1 0 .2
1 *
S p
e c if ic
C o
n tr
o ls
C u st
o m
e rs
W a re
h o
u se
C lu
b 1
0 .3
2 *
W a re
h o
u se
C lu
b 2
0 .6
3 * * *
0 .1
2 *
H o
m e C
e n te
r 1
-0 .2
0 * *
H a rd
w a re
2 4 .2
2
H a rd
w a re
4 -0
.1 6
M a ss
M e rc
h a n t 2
-0 .5
7 0 .0
9 * *
P ro
d u c t L
in e s
C o
m m
o d
it y P
ro d
1 -0
.4 2
*
C o
m m
o d
it y P
ro d
2 -0
.4 8
S p
e c ia
lt y P
ro d
1 -0
.1 2
*
M o
d e l F
it
A IC
2 2 7 5 .4
2 2 2 6 1 .4
8 2 2 5 6 .5
0
Q IC
3 9 9 .5
5 3 8 2 .8
4 3 7 9 .1
2 -2
8 7 .0
7 -2
9 5 .3
2 -2
5 5 5
.0 6
-2 5
5 9
.0 7
R -s
q u a re
0 .0
3 0 .0
3 0 .1
1
M o
d e l T
y p
e
D is
tr ib
u ti o
n
L in
k F
u n c ti o
n
W o
rk in
g C
o rr
e la
ti o
n
* p
< . 1
0 *
* p
< .
0 5
* * *
p <
. 0
1
(a )
E le
c tr
o n
ic c
o n
tr o
l n
o t
a p
p li c a
b le
(b )
G e n
e ra
l c o
n tr
o ls
r e su
lt i n
a p
o o re
r fi
tt in
g m
o d
e l a
n d a
re e
x c lu
d e d
(c )
C u s to
m e r
A c c e p ta
n c e
(a )
G E
E
G a m
m a
L o
g
In d
e p
e n d
e n t
L o
g is
ti c R
e g re
ss io
n -
O d
d s
R a ti o
P ro
fi ta
b il it
y A
li g
n m
e n t(
a )(
c )
C o
m p le
ti o
n T
im e
Q u a
n ti
ty o
f C
o m
m e n ta
ry Q
u a
li ty
o f
C o
m m
e n ta
ry (a
)( b
)
G E
E
P o
is so
n
A F
T
G a m
m a
P ro
fi t
A li g
n m
e n
t te
st e d
o n ly
f o
r T
ra d
it io
n a l a
n d D
ia g
n o
st ic
-D ia
lo g
ic c
a te
g o
ri e s.
C
o n
v e rg
e n
c e n
o t
a c h
ie v e d
o n c
o m
p a
n y o
r sp
e c if
ic c
o n
tr o
l v a
ri a
b le
s
G E
E
G a m
m a
L o
g
U n st
ru c tu
re d
L o
g
U n st
ru c tu
re d
94
Chapter 5: Conclusion and Implications This research study began with an identification of three dilemmas in the literature on
organizational change and development: the dilemma of organizational change
approaches, the dilemma of a systemic perspective, and the dilemma of measurable
change. It then focused on synthesizing a dichotomic schema of organizational
change theories (diagnostic and dialogic) with organizational accounting metrics in
order to develop a diagnostic-dialogic model of organizational change. Founded on
the efficacy of accounting metrics as instruments of diagnostic, dialogic, systemic,
and measurable change, this model serves as a resolution tool for the three dilemmas.
The resulting diagnostic-dialogic organizational change model was tested through a
longitudinal research study in order to compare the effects of a Traditional
intervention (diagnostic) with those of a Diagnostic-Dialogic intervention within an
organizational context; the customer proposal process used in a group of consumer
products companies denoted as WholeCo. Results of the longitudinal study were
covered in Chapter 4, and discussion of those results is covered in the following
sections of this chapter.
General Findings
Chapter 4 provided the results of AFT, PA-GEE, and LR testing on five outcomes of
interest (completion time, quantity of commentary, quality of commentary,
profitability alignment, and customer acceptance) and indicated that the intervention
95
derived from a diagnostic-dialogic approach, founded on accounting metrics,
provided superior results as compared with the results of a laissez-faire approach and
with the results of a diagnostic intervention. Statistically significant improvements
were clearly shown for the Diagnostic-Dialogic group in completion time, quantity of
commentary, quality of commentary, and customer acceptance as compared with both
the Laissez-faire and Traditional groups. Profitability alignment was tested only for
the Traditional and Diagnostic-Dialogic groups, but it also demonstrated statistically
significant improvement for the Diagnostic-Dialogic group as compared with the
Traditional group. Results for each outcome of interest are discussed more fully in the
following sections.
Completion Time
AFT testing supported the hypotheses that because of the ambiguity of the Laissez-
faire process, the greater structure of the Traditional intervention, and the greater
systemic perspective of the Diagnostic-Dialogic intervention, the Traditional group
would demonstrate shorter duration for completion time than would the Laissez-faire
group, and the Diagnostic-Dialogic group would demonstrate shorter duration for
completion time than both the Laissez-faire and the Traditional groups. Results
indicate that completion time was shortened 11% (log = -0.12, p = .07) in the
Traditional group as compared with the Laissez-faire group. Completion time
decreased even further in the Diagnostic-Dialogic group. The Diagnostic-Dialogic
group demonstrates a reduction of approximately 21% (log -0.24, p = .10) as
96
compared with the Laissez-faire group, which is an incremental reduction of 11% as
compared with the Traditional group (1 – [log -0.24] / [log -0.12]). Although it may
appear counterintuitive that increased metric comparatives and greater involvement
(as indicated by results for the outcome variables quantity of commentary and quality
of commentary) would actually decrease process time, these findings are consistent
with the view that the clear provision of a normative performance objective and an
interactive mechanism help reduce uncertainty and increase integration among the
differentiated functions of an organization (Katzenbach & Smith, 2003; Lawrence &
Lorsch, 1969).
Unhypothesized findings related to the Company control variable may add additional
insight. As shown in Chapter 4, results indicate that completion time increases (log =
0.15, p = .03), or approximately 16%, as company increments from SpecialtyCo
(coded 0) to CommodityCo (coded 1). While the causes of longer completion times
for CommodityCo are not expressly examined in this study, these results may be
logically extrapolated from the previous discussion. There exists a likelihood of a
more competitive environment, greater volatility, and thus ambiguity for
CommodityCo as compared with SpecialtyCo, which may inherently cause
CommodityCo to experience longer time duration in customer proposals.
97
Another unhypothesized finding for completion time relates to the Electronic control
variable (representing a change from manual to computerized process). Completion
time increases (log = 0.33, p = .07) as the customer process increments from a manual
process (coded 0) to the Electronic process (coded 1). This impact of the Electronic
control variable is a counterintuitive increase in completion time of approximately
39%. Further discussion with organizational members revealed that when the process
was manually based, an administrative assistant acted as expeditor to track proposals
in process, send frequent reminders to pending approvers, and physically visit
pending approvers if a rapid decision was required. The removal of this expeditor
function when converting to the computerized process may explain why completion
time increased as affected by the Electronic control variable.
Faster processes are critical to organizations operating in a volatile environment,
since recognition of environmental changes and the requisite organizational
adaptations required from those changes must occur rapidly and continually in order
to support organizational success and sustainability. Completion time results indicate
that the Diagnostic-Dialogic intervention was more effective in this type of
environment than was the Traditional intervention or the Laissez-faire approach.
Quantity of Commentary
PA-GEE testing supported the hypotheses that the Traditional group would
demonstrate greater quantity of commentary than would the Laissez-faire group and
98
that the Diagnostic-Dialogic group would demonstrate greater quantity of
commentary than both the Laissez-faire and Traditional groups. These hypotheses
were based on the assumptions that the diagnostic nature of the Traditional group
should logically motivate individuals to explain significant differences of the proposal
to the normative goal and that the combined diagnostic-dialogic nature of the
Diagnostic-Dialogic group should logically motivate even greater commentary
because of its inclusion of both normative comparison with presupposed goals and
generative, double-loop construction of new goals. The PA-GEE results demonstrate
an increase of approximately 19.7% in the Traditional group (log = .18, p = < .01) as
compared with the Laissez-faire group. Results demonstrate an even greater increase
in the Diagnostic-Dialogic group with quantity of commentary approximately 50.7%
higher in the Diagnostic-Dialogic group (log .41, p < .01) than in the Laissez-faire
group.
The increase in quantity of commentary for the Traditional group is consistent with
the logic of a diagnostic process in that functional representatives would naturally
need to increase commentary to support their decisions as compared with a normative
performance criterion, whether comparatively favorable or unfavorable. And the
significant increase resulting from the Diagnostic-Dialogic intervention, in
comparison with both Laissez-faire and Traditional groups, is strong evidence that a
diagnostic-dialogic process extends beyond the benefits of normative comparison to
99
encourage commentary—an element critical to generating awareness of the systemic
and socially constructive nature of organizations and change.
PA-GEE testing also revealed an unhypothesized finding for quantity of commentary
as related to the Company control variable. Quantity of commentary decreases (log =
-0.37 p < .001) by approximately 30.9% as Company increments from SpecialtyCo
(coded 0) to CommodityCo (coded 1). While the causes of lower absolute
commentary for CommodityCo are not expressly examined in this study, a review of
the mean number of comments per record demonstrates that commentary actually
increases at a greater level for CommodityCo as compared with SpecialtyCo in both
the Traditional and Diagnostic-Dialogic quasi-experimental groups. Table 21 shows
that mean comments per CommodityCo record increased, from Laissez-faire group to
Diagnostic-Dialogic group, by over 250% as compared with an approximate 45%
increase for SpecialtyCo.
Table 21. Mean Comments per Record by Company
PA-GEE testing for quantity of commentary also revealed another unhypothesized
finding relative to the Training control variable. Quantity of commentary decreases
Mean Comments Change from Mean Comments Change from
per Record Base per Record Base
Laissez-faire (base) 1.3 0.4
Traditional 1.4 8.9% 1.1 167.4%
Diagnostic-Dialogic 1.9 44.7% 1.5 252.5%
SpecialtyCo CommodityCo
100
(log -0.12, p = .20) as the customer proposal process increments from the process
prior to Training (coded 0) to the process occurring after Training (coded 1). The
impact of the Training control variable, all other variables held constant, is a decrease
in commentary of 11%. This result is statistically insignificant; however, it
directionally indicates a potential decrease in customer proposal commentary as a
result of the accounting metrics, diagnostic-dialogic, training session. A decrease in
quantity of commentary resulting from Training might be explained by the
construction of a common social and systemic structure, which would potentially
routinize elements of the process making commentary less necessary.
Quality of Commentary
PA-GEE testing supported the hypothesis that quality of commentary would
demonstrate a greater degree of systemic perspective in the
Diagnostic-Dialogic group as compared with both the Laissez-faire and Traditional
groups. This hypothesis was developed from the focus on a double-loop dialogic
process in the Diagnostic-Dialogic group. Results for the Traditional group were
indicative but not statistically significant for a decline in systemic commentary rate
for the Traditional group; therefore, the hypothesis that quality of commentary would
decline (i.e., become more functional) in the Traditional group as compared with the
Laissez-faire group was not supported. This hypothesis was developed based on the
assumption that diagnostic processing in the Traditional group without an interactive
101
mechanism would increase the functional nature versus the systemic nature of
commentary.
Results for Hypothesis 3a demonstrate a potential decline in quality of commentary
for the Traditional group (log = -0.13, p = .30) although a high p-value indicates a
lack of statistical significance. This result may be interpreted as a Traditional
(diagnostic) intervention as having either no effect or a small negative effect on the
level of systemic commentary. However, results of PA-GEE testing for the
Diagnostic-Dialogic group (log = 0.27, p = .02) are consistent with Hypothesis 3b,
since quality of commentary, in the Diagnostic-Dialogic group, increases
approximately 31% as compared with the Laissez-faire group and perhaps a greater
amount as compared with the Traditional group results, dependent on interpretation.
Interestingly, the Company control variable was not found to be significant in the
quality of commentary analysis as contrasted with the results for both completion
time and quantity of commentary.
Overall, the findings indicate that the Diagnostic-Dialogic intervention was more
effective than the either the Laissez-faire approach or the Traditional intervention in
increasing the degree of systemic commentary in the WholeCo customer proposal
process.
102
Profitability Alignment
PA-GEE testing supported the hypothesis that profitability alignment would increase
in the Diagnostic-Dialogic group as compared with the Traditional group. Recall from
Chapter 3 that profitability alignment is defined as proposal profitability rate minus
planned profitability rate with a negative result representing nonalignment and a null
or positive result representing alignment. This hypothesis was developed from the
assumption that in the Traditional group each function was more likely to focus on
maximizing its own functional results rather than on optimizing the holistic results of
WholeCo, because of a single-loop diagnostic process. Comparatively, the
Diagnostic-Dialogic group was more likely to focus on the holistic results of
WholeCo because of its double-loop diagnostic-dialogic process.
Results for Hypothesis 4 demonstrate a significantly higher degree of Profitability
Alignment for the Diagnostic-Dialogic group (log = 1.71, p = .10) as compared with
the Traditional group. When the coefficient for the Diagnostic-Dialogic group is
converted to an exponentiated number (e 1.71
= 5.5), results show that population
averaged profitability alignment is 5.5 times higher in the Diagnostic-Dialogic group
than in the Traditional group.
PA-GEE testing also revealed an unhypothesized, but logical, finding for profitability
alignment as related to the Training control variable. Profitability alignment increases
(log = 0.29, p = .50) as the customer proposal process increments from the process
103
prior to Training (coded 0) to the process occurring after Training (coded 1). While
the high p-value indicates lack of statistical significance, there is directional
indication that profitability alignment may increase as a result of specialized
diagnostic-dialogic, accounting metrics-based training.
Customer Acceptance
LR testing did not support the hypothesis that customer acceptance would decline in
the Traditional group as compared with the Laissez-faire group; however, LR testing
did support the hypothesis that customer acceptance would increase in the
Diagnostic-Dialogic group as compared with both the Laissez-faire and the
Traditional groups. These hypotheses were based on the assumptions that the
Traditional group would promote greater diagnostic comparison and greater controls
with an inward focus while lacking a dialogic mechanism (i.e., WholeCo versus the
customer) and that the Diagnostic-Dialogic group would promote greater systems
understanding and greater generative processes through its diagnostic-dialogic
characteristics.
LR testing directionally indicates that the Traditional group may improve the odds of
customer acceptance (OR = 1.11, p = .80); however, the high p-value does not
support a definitive conclusion. The lack of statistical significance indicates that the
null hypothesis of no effect from the Traditional group on customer acceptance is
more appropriate; thus the hypothesis of lower customer acceptance in the Traditional
104
group was rejected. However, LR testing clearly indicates that the odds of customer
acceptance are higher in the Diagnostic-Dialogic group as compared with the Laissez-
faire and Traditional groups. Results for the Diagnostic-Dialogic group (OR = 3.18, p
= .03) indicate that odds of customer acceptance are over 3 times higher in the
Diagnostic-Dialogic group than in the Laissez-faire group. Even if the odds in the
Traditional group are assumed at 1.11 times, the Diagnostic-Dialogic group provides
2.5 times higher odds (i.e. 3.18 / 1.11) of customer acceptance than would the
Traditional group.
An additional, unhypothesized finding for customer acceptance relates to the specific
control variable Customers. The odds ratio for the customer denoted as Hardware 2
(OR = 4.22, p = .10) indicates that odds of customer acceptance were increased by
over 4 times when customer proposals were made to this customer, all other variables
held constant. Further review revealed that, unlike any other customer, this customer
was exclusive to WholeCo in two of the three unique categories constituting the sub-
set of records tested for customer acceptance. Being a customer’s exclusive supplier
in a specific category does not guarantee WholeCo acceptance of its customer
proposals, since factors such as timing, pricing, and product characteristics may not
be acceptable to the customer; however, exclusivity would naturally increase the odds
of customer acceptance.
105
Limitations
As is common in field studies, generalization of research conclusions to a wider scope
of organizations may be limited (Eisenhardt, 1989). This study included two
organizations of different types, CommodityCo and SpecialtyCo, which helped
provide rigor to the research; however, the study does not provide a sufficient number
of organizations and organizational types to ensure that results can be generalized to
other organizations without further research.
Research was conducted in a retrospective manner from field-based, archival data that
provided the strength of longitudinal analysis in a natural setting but precluded the
ability to conduct control based experimental research. Since not all variables could
be controlled in this study, there exists the potential for unrecognized, confounding
variables that could affect results and may be the cause of unexplained variances.
The inability to conduct controlled experiments also precluded the potential for
testing the impact of a purely dialogic intervention. While hypotheses related to the
Traditional intervention (diagnostic) and Diagnostic-Dialogic intervention were
rigorously tested through AFT, PA-GEE, and LR models, there was no occurrence of
a dialogic intervention in the archival data; thus the impact of an intervention with
dialogic but not diagnostic elements was not researched.
106
Implications for Future Research
This study provided rigorous and convincing indicators that a diagnostic-dialogic
intervention, founded on accounting metrics, was effective in integrating both
diagnostic- and dialogic- based organizational change, promoting a systemic
perspective, and providing rigorous change measurement. The Diagnostic-Dialogic
intervention was more effective than a Laissez-faire approach or a Traditional
intervention in improving process completion time and creating a higher level of
organizational commentary. The Diagnostic-Dialogic intervention was also shown
more effective at increasing the level of systemic commentary and the odds of
proposal acceptance by WholeCo’s customers. Further, this study found that the
Diagnostic-Dialogic intervention provided a greater chance of meeting or exceeding
profitability objectives than occurred in the Laissez-faire approach or the Traditional
intervention.
However, further research is needed in order to expand generalization, within OD, of
a diagnostic-dialogic organizational change model founded on accounting metrics.
Such research may replicate this study in a larger number of organizations and
organizational types, or it may conduct controlled experiments to isolate the impact of
potentially confounding variables. These types of investigations may provide
additional findings on factors that influence the degree of impact resulting from a
diagnostic-dialogic intervention, as was indicated by the findings on company type in
this study.
107
Future research is warranted on the effects of a diagnostic-dialogic, accounting
metrics-based intervention on absolute profitability and on alignment of profitability
objectives and results. Profitability metrics are often among the most important and
most visible metrics for organizations; and as is the case with generalizability,
stronger support for this type of intervention may result from replicating this study in
more organizations or from conducting controlled experiments.
Another interesting area for future research is the potential impact of a dialogic
intervention such as Appreciative Inquiry, Future Search, or World Café, as compared
with both diagnostic and diagnostic-dialogic interventions. This issue is among the
more difficult to research since the combination of all three interventions is not likely
to have naturally occurred in a longitudinal setting, and time is required for impact to
be demonstrated; however, it might be studied over a variety of organizations or
organizational departments over a period of time; and some benefit might accrue from
short-term testing in a controlled environment.
108
Appendix A: GEE Distributions and Link Functions
2
2 From “Using Generalized Estimating Equations for Longitudinal Data Analysis,” by Gary A.
Ballinger, 2004, Organizational Research Methods, Volume 7, p. 148. Copyright 2004 by Sage
Publications. Reprinted with permission
Distribution Choices and Link Functions Available
in Generalized Estimating Equations (GEEs)
The link function that is selected will vary depending on the distribution of the underlying
dependent variable. Certain dependent variables permit multiple link functions depending on
how the user wishes to interpret the coefficients (as cumulative probabilities, for example). This
appendix provides some brief guidance on the different link functions available in GEE models.
Normal Distribution
Identity link: This fits the same model as the general linear model
Power link: Any power transformation (e.g., square root, square of variable)
Reciprocal link: Links using reciprocal of dependent variable (1/µ)
Binomial Distribution (1/0 Data)
Logit link: Fits logistic regression models
Probit link: Fits cumulative probability functions
Power link: Any power transformation (e.g., square root, square of variable)
Reciprocal link: Links using reciprocal of dependent variable (1/µ)
Poisson Distribution (Counted Data)
Log link
Power link: Any power transformation (e.g., square root, square of variable)
Reciprocal link: Links using reciprocal of dependent variable (1/µ)
Negative Binomial Distribution
Power link: Any power transformation (e.g., square root, square of variable)
Gamma Distribution
Power link: Any power transformation (e.g., square root, square of variable)
Reciprocal link: Links using reciprocal of dependent variable (1/µ)
Multinomial Distribution*
Cumulative logit link
*At the time of this writing, analysis of multinomial distributed dependent variables was permitted
in SAS using only the cumulative logit link and the independence assumption regarding correla-
tion of responses.
109
Appendix B: Disposition of Outlier Records
Note: Observations for “Quality of Commentary” and “Customer Acceptance” did
not contain any z-score values > ± 3.3
Record Variable Value Z-Score Disposition
90 Completion Time 84.0000 3.4334 Retain (Last comment date = Failure date)
190 Completion Time 86.0000 3.5330 Retain (Last comment date = Failure date)
255 Completion Time 99.0000 4.1803 Retain (Last comment date = Failure date)
281 Completion Time 179.0000 8.1641 Convert to missing record (Failure date > Last Comment date)
283 Completion Time 169.0000 7.6661 Convert to missing record (Failure date > Last Comment date)
337 Completion Time 138.0000 6.1224 Convert to missing record (Failure date > Last Comment date)
338 Completion Time 138.0000 6.1224 Convert to missing record (Failure date > Last Comment date)
388 Completion Time 95.0000 3.9811 Retain (Last comment date = Failure date)
397 Completion Time 87.0000 3.5828 Convert to missing record (Failure date > Last Comment date)
672 Completion Time 90.0000 3.7322 Retain (Last comment date = Failure date)
698 Completion Time 101.0000 4.2799 Retain (Last comment date = Failure date)
721 Completion Time 82.0000 3.3338 Retain (Last comment date = Failure date)
744 Completion Time 172.0000 7.8155 Retain (Last comment date = Failure date)
745 Completion Time 172.0000 7.8155 Retain (Last comment date = Failure date)
766 Completion Time 141.0000 6.2718 Convert to missing record (Failure date > Last Comment date)
782 Completion Time 138.0000 6.1224 Convert to missing record (Failure date > Last Comment date)
816 Completion Time 82.0000 3.3338 Convert to missing record (Failure date > Last Comment date)
850 Completion Time 84.0000 3.4334 Retain (Last comment date = Failure date)
75 Quantity of Commentary 6.0000 3.9754 Retain - valid record
398 Quantity of Commentary 6.0000 3.9754 Retain - valid record
375 Profitability Alignment -0.4112 -5.8091
Convert to missing record - plan based on non-company
acquisition target - not comparable
460 Profitability Alignment 0.6382 8.5129
Convert to missing record - product grouping not comparable
between plan & proposal
489 Profitability Alignment 0.3657 4.7933 Retain - valid record
535 Profitability Alignment 0.2731 3.5294 Retain - valid record
606 Profitability Alignment 0.3612 4.7325 Retain - valid record
825 Profitability Alignment -0.4326 -6.1008
Convert to missing record - product grouping not comparable
between plan & proposal
909 Profitability Alignment 0.2768 3.5806 Retain - valid record
110
References Ackoff, R. L. (1974). Redesigning the future: A systems approach to societal
problems. New York: Wiley.
Ackoff, R. L. (1994). Systems thinking and thinking systems. System Dynamics
Review, 10(2-3), 175-188. doi: 10.1002/sdr.4260100206
Ackoff, R. L. (2006). Why few organizations adopt systems thinking. Systems
Research and Behavioral Science, 23(5), 705-708. doi: 10.1002/sres.791
Argyris, C. (1952). The impact of budgets on people. New York: Controllership
Foundation.
Argyris, C. (1953). Human problems with budgets. Harvard Business Review, 31(1),
97-110.
Argyris, C. (1976). Single-loop and double-loop models in research on decision
making. Administrative Science Quarterly, 21(3), 363-375.
Argyris, C. (1977). Double loop learning in organizations. Harvard Business Review,
55(5), 115-125.
Argyris, C. (1990). The dilemma of implementing controls: The case of managerial
accounting. Accounting, Organizations and Society, 15(6), 503-511.
Ballinger, G. A. (2004). Using generalized estimating equations for longitudinal data
analysis. Organizational Research Methods, 7(2), 127-150. doi:
10.1177/1094428104263672
Barrett, F. J., & Peplowski, K. (1998). Minimal structures within a song: An analysis
of "All of Me.". Organization Science, 9(5), 558-560.
Bogdan, R., & Ksander, M. (1980). Policy data as a social process: A qualitative
approach to quantitative data. Human Organization, 39(4), 302-309.
Boland Jr, R. J. (1989). Beyond the objectivist and the subjectivist: Learning to read
accounting as text. Accounting, Organizations and Society, 14(5-6), 591-604.
Boland Jr, R. J. (1993). Accounting and the interpretative act. Accounting,
Organizations and Society, 18(2/3), 125-146.
111
Boland, R. J. (1981). A study in system design: C. West Churchman and Chris
Argyris. Accounting, Organizations and Society, 6(2), 109-118. doi:
10.1016/0361-3682(81)90001-5
Boland, R. J. (1989). Beyond the objectivist and the subjectivist: Learning to read
accounting as text. Accounting, Organizations and Society, 14(5-6), 591-604.
Boland, R. J. (1993). Accounting and the interpretative act. Accounting,
Organizations and Society, 18(2/3), 125-146.
Bradburn, M. J., Clark, T. G., Love, S. B., & Altman, D. G. (2003a). Surival analysis
Part II: Multivariate data analysis - an introduction to concepts and methods.
British Journal of Cancer, 89(3), 431-436.
Bradburn, M. J., Clark, T. G., Love, S. B., & Altman, D. G. (2003b). Surival analysis
Part III: Multivariate data analysis - choosing a model and assessing its
adequacy and fit. British Journal of Cancer, 89(4), 605-611.
Burke, W. W. (1980). Systems theory, gestalt therapy, and organizational
development. In T. G. Cummings (Ed.), Systems theory for organization
development (pp. 209-222). Chichester, England: John Wiley & Sons.
Burke, W. W. (2008a). A contemporary view of organization development. In T. G.
Cummings (Ed.), Handbook of organization development (pp. 13-38). Los
Angeles, CA: Sage Publications.
Burke, W. W. (2008b). Organization change, theory and practice (2nd ed.). Los
Angeles , CA: Sage Publications.
Bushe, G. R., & Marshak, R. J. (2008). The postmodern turn in OD. OD Practitioner,
40(4), 9-11.
Bushe, G. R., & Marshak, R. J. (2009). Revisioning organization development:
diagnostic and dialogic premises and patterns of practice. Journal of Applied
Behavioral Science, 45(3), 348-368. doi: 10.1177/0021886309335070
Carlon, D. M., Downs, A. A., & Wert-Gray, S. (2006). Statistics as fetishes.
Organizational Research Methods, 9(4), 475-490. doi:
10.1177/1094428106291056
Carucci, R. A., & Pasmore, W. A. (2002). Relationships that enable enterprise
change: Leveraging the client-consultant connection. San Francisco, CA:
Jossey-Bass/Pfeiffer.
112
Churchman, C. W., & Ackoff, R. L. (1950). Purposive behavior and cybernetics.
Social Forces, 29(1), 32-39.
Clark, T. G., Bradburn, M. J., Love, S. B., & Altman, D. G. (2003). Survival analysis
Part I: Basic concepts and first analyses. British Journal of Cancer, 89(2),
232-238.
Collins, L. M. (2006). Analysis of longitudinal data: The integration of theoretical
model, temporal design, and statistical model. Annual Review of Psychology,
57, 505-528.
Cui, J., & Guoqi, Q. (2007). Selection of working correlation structure and best
model in GEE analyses of longitudinal data. Communications in Statistics:
Simulation & Computation, 36(5), 987-996. doi:
10.1080/03610910701539617
Cummings, T. G., & Worley, C. G. (2009). Organization development & change (9th
ed.). Mason, OH: South-Western/Cengage Learning.
Eisenhardt, K. M. (1989). Building theories from case study research. Academy of
Management Review, 14(4), 532-550. doi: 10.5465/amr.1989.4308385
Featherman, D. L. (1980). Retrospective longitudinal research: Methodological
considerations. Journal of Economics and Business, 32(2), 152-169.
Flood, R. L., & Jackson, M. C. (1991). Creative problem solving: Total systems
intervention. West Sussex, England: John Wiley & Sons, Ltd.
Ford, J. D., & Ford, L. W. (1995). The role of conversations in producing intentional
change in organizations. Academy of Management Review, 20(3), 541-570.
Gephart, R. P. (2006). Ethnostatistics and oganizational research methodologies.
Organizational Research Methods, 9(4), 417-431. doi:
10.1177/1094428106290199
Gharajedaghi, J. (2007). Systems thinking: a case for second-order-learning. The
Learning Organization, 14(6), 473-479. doi: 10.1108/09696470710825088
Hammond, D. (1997). Toward a science of synthesis: The heritage of general
systems theory. PhD dissertation, Univerity of California, Berkeley, Berkeley.
Hardin, J. W., & Hilbe, J. M. (2003). Generalized Estimating Equations. Boca Raton,
FL: Chapman & Hall / CRC.
113
Hatch, M. J., & Cunliffe, A. L. (2006). Organization theory: modern, symbolic, and
postmodern perspectives (2nd ed.). New York: Oxford University Press.
Henri, J.-F. (2006). Management control systems and strategy: A resource-based
perspective. Accounting, Organizations and Society, 31(6), 529-558. doi:
DOI: 10.1016/j.aos.2005.07.001
Hosmer, D. W., Lemeshow, S., & May, S. (2008). Applied survival analysis:
Regression modeling of time-to-event data (2nd ed.). Hoboken, NJ: Wiley.
Hung, W. (2008). Enhancing systems-thinking skills with modelling. British Journal
of Educational Technology, 39(6), 1099-1120. doi: 10.1111/j.1467-
8535.2007.00791.x
Jackson, M. C. (2006). Creative holism: A critical systems approach to complex
problem situations. Systems Research and Behavioral Science, 23(5), 647-
657. doi: 10.1002/sres.799
Katz, D., & Kahn, R. L. (1978). The social psychology of organizations (2nd ed.).
New York, NY: John Wiley & Sons.
Katzenbach, J. R., & Smith, D. K. (2003). The wisdom of teams: Creating the high-
performance organization. New York: Harper Business Essentials.
Kerlinger, F. N., & Lee, H. B. (2000). Foundations of behavioral research (4th ed.).
Belmont, CA: Cengage Learning.
Kleinbaum, D. G., & Klein, M. (2005). Survival analysis: A self-learning text (2nd
ed.). New York: Springer.
Kleinbaum, D. G., & Klein, M. (2010). Logistic regression: A self-learning text (3rd
ed.). New York: Springer.
Lawless, J. F. (2003). Statistical models and methods for lifetime data (2nd ed.).
Hoboken, NJ: Wiley.
Lawrence, P. R., & Lorsch, J. W. (1969). Organization and environment: Managing
differentiation and integration. Homewood, IL: Richard D. Irwin.
Liang, K.-Y., & Zeger, S. L. (1986). Longitudinal data analysis Using generalized
linear models. Biometrika, 73(1), 13-22.
114
Marginson, D., & Ogden, S. (2005). Coping with ambiguity through the budget: The
positive effects of budgetary targets on managers' budgeting behaviours.
Accounting, Organizations and Society, 30(5), 435-456. doi:
10.1016/j.aos.2004.05.004
Marshak, R. J., & Grant, D. (2008). Organizational discourse and new organization
development practices. British Journal of Management, 19, S7-S19.
Melnyk, S. A., Hanson, J. D., & Calantone, R. J. (2010). Hitting the Target...but
Missing the Point: Resolving the Paradox of Strategic Transition. Long Range
Planning, 43(4), 555-574. doi: DOI: 10.1016/j.lrp.2009.11.004
Melnyk, S. A., Stewart, D. M., & Swink, M. (2004). Metrics and performance
measurement in operations management: Dealing with the metrics maze.
Journal of Operations Management, 22(3), 209-217.
Mittlbö, M., & Schemper, M. (1996). Explained variance for logistic regression.
Statistics in Medicine, 15(19), 1987-1997.
Mohrman Jr., A., M., Mohrman, S. A., Ledford Jr, G. E., Cummings, T. G., & Lawler
III, E. E. (Eds.). (1989). Large-Scale Organizational Change. San Francisco:
Jossey-Bass.
Mohrman, S. A., Mohrman, A., M., & Tenkasi, R. V. (1996). The discipline of
organization design. Center for Effective Organizations Publication. Los
Angeles, CA: University of Southern California.
Montuori, L. A. (2000). Organizational longevity - Integrating systems thinking,
learning and conceptual complexity. Journal of Organizational Change
Management, 13(1), 61-73. doi: 10.1108/09534810010310249
Morgan, G. (1982). Cybernetics and organization theory: Epistemology or technique?
Human Relations, 35(7), 521-537.
Morgan, G. (1988). Accounting as reality construction: Towards a new epistemology
for accounting practice. Accounting, Organizations and Society, 13(5), 477-
485. doi: 10.1016/0361-3682(88)90018-9
Napier, C. J. (2006). Accounts of change: 30 years of historical accounting research.
Accounting, Organizations and Society, 31(4-5), 445-507. doi: DOI:
10.1016/j.aos.2005.12.004
115
Oswick, C. (2009). Revisioning or re-versioning?: A commentary on diagnostic and
dialogic forms of organization development. Journal of Applied Behavioral
Science, 45(3), 369-374. doi: 10.1177/0021886309338687
Pan, W. (2001). Akaike's information criterion in generalized estimating equations.
Biometrics, 57(1), 120-125. doi: 10.1111/j.0006-341X.2001.00120.x
Pasmore, W. A. (1998). Organizing for Jazz. Organization Science, 9(5), 562-568.
Pasmore, W. A., & Khalsa, G. S. (1993). The contributions of Eric Trist to the social
engagement of social science. Academy of Management Review, 18(3), 546-
569. doi: 10.5465/amr.1993.9309035150
Pettigrew, A. M. (1990). Longitudinal field research on change: Theory and practice.
Organization Science, 1(3), 267-292.
Pettigrew, A. M., Woodman, R. W., & Cameron, K. S. (2001). Studying
organizational change and development: Challenges for future research.
Academy of Management Journal, 44(4), 697-713.
Schaie, K. W., & Hertzog, C. (1982). Longitudinal methods. In B. Wolman (Ed.),
Handbook of developmental psychology (pp. 91-115). Englewood Cliffs, NJ:
Prentice-Hall.
Schwaninger, M. (2001). System theory and cybernetics: A solid basis for
transdisciplinarity in management education and research. Kybernetes, 30(9),
1209-1222. doi: 10.1108/EUM0000000006551
Schwaninger, M. (2006). System dynamics and the evolution of the systems
movement. Systems Research and Behavioral Science, 23(5), 583-594. doi:
10.1002/sres.800
Senge, P. M. (1990). The fifth discipline: The art and practice of the learning
organization. New York, NY: Currency Doubleday.
Simons, R. (1991). Strategic orientation and top management attention to control
systems. Strategic Management Journal, 12(1), 49-62.
Simons, R. (1994). How new top managers use control systems as levers of strategic
renewal. Strategic Management Journal, 15(3), 169-189. doi:
10.1002/smj.4250150301
116
Snow, C. C., & Thomas, J. B. (1994). Field research methods in strategic
management: Contributions to theory building and testing. Journal of
Management Studies, 31(4), 457-480.
Sterman, J. D. (1994). Learning in and about complex systems. System Dynamics
Review, 10(2-3), 291-330. doi: 10.1002/sdr.4260100214
Sterman, J. D. (2002). All models are wrong: reflections on becoming a systems
scientist. System Dynamics Review, 18(4), 501-531. doi: 10.1002/sdr.261
Stone, E. (1978). Research methods in organizational behavior. Santa Monica, CA:
Goodyear Publishing Company.
Sweeney, L. B., & Sterman, J. D. (2000). Bathtub dynamics: Preliminary results of a
systems thinking inventory. Paper presented at the International System
Dynamics Conference, Bergen, Norway.
Swieringa, R. J., & Weick, K. E. (1981). Interfaces between management accounting
and organizational behavior. Journal of Management Education, 6(3), 25-33.
doi: 10.1177/105256298100600311
Tenkasi, R. V., & Boland, R. J. (1993). Locating meaning making in organizational
learning: The narrative basis of cognition. In R. W. Woodman & W. A.
Pasmore (Eds.), Research in Organizational Change and Development (Vol.
7, pp. 77-103). Greenwich, CT: JAI Press.
Tenkasi, R. V., Mohrman, S. A., & Mohrman, J., Allan M. (2007). Making knowledge
contextually relevant: The challenge of connecting academic research with
practice. Paper presented at the Third Organization Studies Summer
Workshop, Organization Studies as Applied Science: The Generation and Use
of Academic Knowledge about Organizations, Crete, Greece.
van de Ven, A. H., & Poole, M. S. (1995). Explaining development and change in
organizations. The Academy of Management Review, 20(3), 510-540.
von Bertalanffy, L. (1972). The history and status of general systems theory.
Academy of Management Journal, 15(4), 407-426.
Weisbord, M. R. (2004). Productive workplaces revisited: Dignity, meaning, and
community in the 21st century (2nd ed.). San Francisco: Jossey-Bass.
Wolfram, S. (2002). A new kind of science. Champaign, IL: Wolfram Media, Inc.
117
Woodman, R. W. (1989). Evaluation research on organizational change: Arguments
for a "combined paradigm" approach. Greenwich, CT: JAI Press.
Woodman, R. W. (2008). Discourse, metaphor and organizational change: The wine
is new, but the bottle is old. British Journal of Management, 19, S33-S37.
Woodman, R. W., Bingham, J. B., & Yuan, F. (2008). Assessing organization
development and change interventions. In T. G. Cummings (Ed.), Handbook
of organization development (pp. 187-215). Los Angeles, CA: Sage
Publications.
Yin, R. K. (2003). Case study research: Design and method (3rd ed.). Thousand
Oaks, CA: Sage Publications.
Zeger, S. L., & Liang, K.-Y. (1986). Longitudinal data analysis for discrete and
continuous outcomes. Biometrics, 42(12), 121-130.
Zexian, Y., & Xuhui, Y. (2010). A revolution in the field of systems thinking—a
review of Checkland's system thinking. Systems Research and Behavioral
Science, 27(2), 140-155. doi: 10.1002/sres.1021
Zulauf, C. A. (2007). Learning to think systemically: What does it take? The
Learning Organization, 14(6), 489-498. doi: 10.1108/09696470710825105
- WB - cover page
- Brock (2013) - Toward a Diagnostic-Dialogic Model of Organizational Change
