Six Sigma and The University: Teaching, Research, and Meso-Analysis
Six Sigma and The University: Teaching, Research, and Meso-Analysis
Six Sigma and The University: Teaching, Research, and Meso-Analysis
DISSERTATION
By
2005
to improve any given company’s competitive position. Its acceptance by industry has
been widespread over the past two decades, yet academic research on Six Sigma has
been surprisingly limited. Further, most of the research has been focused on the tools
and statistical techniques used in Six Sigma. Its relationship with university activities
including teaching, research, and service is not clear. The purpose of this dissertation
is to explore selected aspects of the relationship between Six Sigma and universities
more fully. In doing so, there is an attempt to answer these fundamental questions:
(i) What is Six Sigma? (ii) What roles can academics usefully play in relation to Six
Sigma? and (iii) How can academia help companies to better use the new project
related data sources created by Six Sigma. Results here divide into three chapters.
First, the literature on Six Sigma is reviewed and synthesized. This includes detailed
quality management theory and topics for future research. Secondly, case base training
is examined as a method to improve Six Sigma education and increase usage on the
job among university student learners. Third, with Six Sigma’s emphasis on
ii
accumulate many large databases of “meta-data” concerning the successes or failures
making use of the project data and illustrate their application using 39 case studies
iii
ACKNOWLEDGMENTS
adequately acknowledge the incredible support I have received form so many people
I am sincerely grateful to my advisor, Dr. Allen, for all of his continuous assistance
and direction. I would like to thank my other committee members: Dr. Mount-
Campbell, and Dr. Miller as well as Chaitanya Joshi who helped in the early
formulation of some of the goals. I appreciate Lucent Technologies, Inc. and LaBarge,
Inc for the support they provided during my recent studies. I owe a debt to my Aunt
Joan for instilling in me the value and love of learning, my children and grandchildren
for the sacrifice of their time. Finally, I would like to thank my wife. This dissertation
is dedicated to her endless encouragement and support over these many years.
iv
VITA
1951 ……………………………………………… Born – Tucson, Arizona
FIELDS OF STUDY
v
TABLE OF CONTENTS
Abstract ………………………………………………………………………………. i
Acknowledgements ………………………………………………………………….. ii
VITA …………………………………………………………………………………. v
List of Tables ………………………………………………………………………. viii
List of Figures ……………………………………………………………………….. x
vi
4.5 Background Information .................................................................................... 60
4.6 In-class Exercise for PCB Study ........................................................................ 63
4.7 The Team Actions and Results........................................................................... 64
4.8 Student Feedback ............................................................................................... 68
Chapter 5 The challenge of Six Sigma Project Meso-Analysis............................... 70
5.1 Introduction ........................................................................................................ 70
5.2 Potential Use of Database................................................................................... 72
5.3 Database Example .............................................................................................. 73
5.4 General Procedure .............................................................................................. 74
5.5 Data Collection................................................................................................... 75
5.6 Statistical Process Control.................................................................................. 79
5.7 Regression .......................................................................................................... 84
5.8 Regression Data Analysis................................................................................... 87
5.9 Regression Results ............................................................................................. 94
5.10 Markov Decision Processes.............................................................................. 96
5.11 Markov Decision Process Analysis................................................................ 103
5.12 Sample MDP Example ................................................................................... 115
5.13 Conclusions .................................................................................................... 122
Chapter 6 Conclusions and Future Research ........................................................ 126
6.1 Overview .......................................................................................................... 126
6.2 Summary of Findings ....................................................................................... 127
6.3 Limitations and Future Research...................................................................... 130
Bibliography………………………………………………………………………. 143
vii
LIST OF TABLES
Table 1 List of journals or proceedings with at least one article in the study. ............. 21
Table 2 Descriptors used to classify articles. ............................................................... 24
Table 3 Tabulations of articles by focus and authorship.............................................. 30
Table 4 Literature pertinent to evaluation of Six Sigma’s effects on firm performance.
(BP – Business Performance, C – Core, SSC – Six Sigma Core, SSI – Six Sigma
Infrastructure, OP – Operational Performance, QP – Quality Performance) ....... 45
Table 5 The yields achieved for 16 weeks prior to the initial teams activities. ........... 61
Table 6 The initial team's predicted yield improvements by adjusting each factor. .... 62
Table 7 The performance after the implementation of the initial team's
recommendations.................................................................................................. 62
Table 8 The yields for the 5 weeks subsequent to the initial intervention. .................. 65
Table 9 The data for a Pareto chart of the data from week 1 to week 16..................... 66
Table 10 The data from the fractional factorial design in four factors......................... 67
Table 11 The confirmation runs establishing the process shift/improvement.............. 68
Table 12 Definition of variables................................................................................... 77
Table 13 Summary statistics for the 39 projects, duration team memberrs and profits78
Table 14 Characteristics of the Study Variables .......................................................... 79
Table 15 Summary statistics for model. ....................................................................... 89
Table 16 Coefficient estimates ..................................................................................... 89
Table 17 Regression Checklist. .................................................................................... 91
Table 18 Summary statistics for Simple model............................................................ 94
Table 19 Coefficients for Simple model. ..................................................................... 94
Table 20 Costs and rewards (in $K), rt(st,a), of applying actions (a) in different states
(st).Coefficient estimates .................................................................................... 104
Table 21 Assumed transition probabilities for applying DOE, pt(st+1= j|st= i,a = $$
DOE)................................................................................................................... 105
Table 22 Optimal decision policy for five decision periods....................................... 106
Table 23 The expected reward, ERt*(st = i), in $K as a function of period and state. 106
Table 24 Component Method of Six Sigma. .............................................................. 108
Table 25 State Description Mapped to DMAIC for Sample MDP. ........................... 115
Table 26 Actions and Rewards for Sample MDP. ..................................................... 115
Table 27 Tabulation of State Transitions for Sample MDP. ...................................... 117
Table 28 Bayesian Estimate of State Transitions for Sample MDP........................... 118
Table 29 State Transitions for Sample MDP.............................................................. 120
Table 30 Optimal Policy for Sample MDP. ............................................................... 121
viii
Table 31 Comparison of strengths and limitations for meso-analysis methods used. 124
ix
LIST OF FIGURES
Figure 1 The yearly number of Six Sigma related articles and their authorship. ......... 26
Figure 2 The percentage of articles focused on manufacturing tropics........................ 27
Figure 3 Percentages of articles sponsored by difference societies or areas................ 28
Figure 4 Pareto chart of articles by research approach ................................................ 31
Figure 5 Journal impact factors associated with articles by different types of authors.32
Figure 6 Journal impact factors of publications pertinent to business sectors. ............ 33
Figure 7 Journal impact factors associated with articles focusing on people (Pe), Tools
and Techniques (To), Systems (Sy), or a combination of these. .......................... 34
Figure 8 Journal impact factors associated with articles focusing on philosophy (Ph),
practices (Pr), Tools and Techniques (To), or a combination of these................. 35
Figure 9 Percentages of articles mentioning each of 14 success factors...................... 36
Figure 10 Extended quality performance model of Garvin.......................................... 43
Figure 11 Control chart for the entire study period...................................................... 65
Figure 12 Normal Probability Chart for Six Sigma Projects........................................ 80
Figure 13 EWMA Control Chart for first 25 Six Sigma Projects. ............................... 82
Figure 14 EWMA Control Chart for Six Sigma Projects............................................. 83
Figure 15 Main Effects Plot of Regression Model. ...................................................... 88
Figure 16 3-D plot of the results.................................................................................. 90
Figure 17 Normal sores vs. residuals. .......................................................................... 91
Figure 18 Main Effects Plot of Simple Regression Model........................................... 93
Figure 19 Flow Chart of DMAIC Six Sigma Project. ................................................ 110
Figure 20 Morkov Decision Process for Six Sigma. .................................................. 114
x
CHAPTER 1
INTRODUCTION
1.1 Overview
Six Sigma was born approximately two decades ago as a process improvement
industry and spread largely by professional consultants. Since its introduction it has
found its way into most sectors of today’s business society. Mostly led by practitioners
Six Sigma has acquired a strong prescriptive stance with practices often being
advocated as universally applicable, “one size fits all”. Yet, Six Sigma has already
spawned a large number of published articles in pier reviewed journals. One of the
While Six Sigma has made a big impact on industry, the relationship between
the university, industry and Six Sigma has not been studied. University-industry
1
The major problems addressed in this dissertation are:
2. What are the implications of Six Sigma philosophy and methods for
university education?
3. What methods should be used to mine the new databases about project
financial results? Also, what insights can be gained form studying data at a
real company?
relationship with Six Sigma. For the first question, the literature covering a fourteen
year timeframe describing the trends, sources, and findings of Six Sigma is reviewed.
The aim is to provide a description of the Six Sigma literature with an emphasis on
education and increase usage on the job among university student learners. From the
literature review, one of the major themes and “success factors” identified by authors
of Six Sigma articles is education of practitioners in the use of statistical tools. Six
Sigma differs form most quality management systems by its emphases on training
of this topic is the education of non-statisticians in the use of statistical tools. A major
2
concept than becomes how to effectively educate participants in the necessary tools
Third, with Six Sigma’s emphasis on management by data and project based
undertaken to look at this database in a way that could help management decision-
as:
This study investigates the use of Six Sigma databases in Meso-Analysis for decision-
making.
3
• Chapter 3 covers a review of the literature concerning Six Sigma from
electronic components.
4
CHAPTER 2
produced by the engineered system being improved (Shina, 2002; Tadikamalla, 1994).
It is a rating that signifies “best in class”, with only 3.4 defects per million units or
addition to specifying the USL and LSL, a target value is specified, which typically is
The symbol sigma (σ) is a letter in the Greek alphabet used by modern people
to describe variability. In Six Sigma, the common measurement index is defects per
million opportunities and can include anything from a component, piece of material,
line of code, an administrative form, time frame or distance. A sigma quality level
offers an indicator of how often defects are likely to occur, where a higher sigma
quality level indicates a process that is less likely to create defects. Consequently, as
sigma level of quality increases, product reliability improves, the need for testing and
5
inspection diminishes, work in progress declines, cycle time goes down, costs go
Specifically,
USL − LSL
Cp = (1.2)
6σ (total process range from − 3σ to + 3σ )
product specifications
Cp = (1.3)
manufacturing var iability
⎡USL − X X − LSL ⎤
Cpk = min ⎢ , ⎥ (1.4)
⎣ 3σ 3σ ⎦
set the product specifications, whereas manufacturing engineers are responsible for
6
production variability. The object of increasing the process capability to six sigma is
twofold: either increase the product specifications by widening them, or reducing the
manufacturing variability. Either effort can have a positive effect on reaching six
sigma.
An alternative and more common definition for “Six Sigma” methods, implied
by Pande and Holpp (2001) and Watson (2002a), is a series of ordered activities with
quantitative approach involving setting up a system and process for the improvement
drives the overall process of selecting the right projects based on an organization’s
business goals and selecting and training the right people to obtain the results.
phases. These component methods derive from statistics, marketing, and optimization
and are sequenced as Define, Measure, Analyze, Improve, and Control (DMAIC). In
design projects the specifics of the DMAIC steps are often modified to DMADV;
7
The phases of DMAIC are described by Rasis, Gitlow and Popovich (2003a
requirements often called “CTQs” “critical to quality”, develop a team charter and
management support.
Measure Phase: Measure the existing systems. Establish valid and reliable
metrics to help monitor progress towards the project goals. Customer expectations
problems.
8
• Perform measurement system analysis. Determine precision,
Analyze Phase: Analyze the system to identify ways to eliminate the gap
between the current performance of the system or process and the desired goal. In
this phase, project teams explore underlying reasons for defects. They use
statistical analysis to examine potential variables affecting the outcome and seek to
identify the most significant root causes. Then, they develop a prioritized list of
• Isolate and verify the critical processes. Narrow the potential list of
Identify and define the limitations of the processes. Ensure that the
Improve Phase: In this phase, project teams seek the optimal solution and
develop and test a plan of action for implementing and confirming the solution.
9
The process is modified and the outcome is measured to determine whether the
Control Phase: Control the new system. Ongoing measures are implemented
systems.
reaction plan.
belt”, “green belt”, etc. For thousands of participants at the lowest “green belt” level
of accreditation, one of the main benefits of “Six Sigma” training is that it simplifies
10
(through restriction) the sequence and choice of available techniques to apply to a
particular case. Therefore, the value of the six sigma movement derives partly from
anecdotal evidence has been rather limited and exists primarily as small-sample case
studies. Moreover, while the empirical results from these case studies have generally
been atheoretical in nature, their conduct had not been governed by rigorous, a priori
theory development.
prevention associated with a Six Sigma program can improve financial results. Their
view was to do or not to do a project. Hild, Sanders, and Cooper (2000) discussed the
Sanders and Hild (2000a and 2000b) outlined the importance of considering
organizational issues in the structuring of successful Six Sigma projects. Snee (2001a)
similarly stated that one of the keys was to understand the environment in order to
type of industry (assembly, processing, chemical, etc.). Pyzdek (2001b) pointed out
11
the importance of selecting the right individual to act as project leader even before
Pande, Neuman, and Cavanagh (2000) contributed probably the most complete
and explicit version of the Six Sigma methods. Yet, even their version of the
applications and to their own tastes. This lack of standardization of the methodologies
explains, at least in part, why the American Society for Quality did not have a
as 1798 and Eli Whitney. One of his greatest contributions to modern manufacturing
was the introduction of his revolutionary uniformity system in the mass production of
muskets. He proved it was possible to produce interchangeable parts that were similar
enough in fit and function to allow for random selection of parts in the assembly.
Throughout the next century, quality involved objective methods of measuring and
With the advent of the moving assembly line by Ford in the early 1900’s it was
stoppage of the assembly line. In addition with the increased production rates, it
became cost prohibitive to measure each part. This necessitated the development of
12
methods to monitor the part producing process for consistency and the use of sampling
(Folaron, 2003).
Western Electric. This was the beginning of statistical quality control (Small, 1956).
The role of the quality inspector changed with the statistically based control charts
form one of identifying and sorting defective product to one of monitoring the stability
of the process and identifying when it had changed. Improved product quality resulted
Dr. Shewhart kept on with his efforts and applied the fundamentals of statistical
quality control to industry. This lead to the modern attention to the use of statistical
tools for the manufacture of products and process originated prior to and during World
War II, when the United States of America geared up to a massive buildup of
machinery and arms to successfully conclude the war. The need to manage the myriad
of complex weapon systems and their varied and distributed defense contractors led to
the evolution of the system of Statistical Quality Control (SQC), a set of tools that
culminated in the military standards for subcontracting, such as MIL-Std 105 (Shina,
2002). The basis of the SQC process was the use of 3 sigma limits, which yields a rate
vertical integration (Shina, 2002). In order to maintain and manage quality, companies
had to control all of the resources used in the product. U.S. companies slowly realized
13
that quality improvements depended on the realization of two major elements. First,
they have to be quantifiable and measurable, and second all elements that make the
because it was the breeding ground for many quality leaders, not only Shewhart but
Joseph Juran, Edwards Deming and Kaoru Ishikawa all worked there at some time
(Dimock, 1977). After World War II, numerous manufacturing experts where involved
with the rebuilding of the Japanese business infrastructure. Two prominent individuals
were Deming and Juran. Deming promoted the use of the plan-do-check-act (PDCA)
Any discussion on quality today will most likely cite at least one from the
group of Deming, Juran, Crosby, Feigenbaum, and Ishikawa, if not all. They certainly
14
• Eliminate non-conformance, Appraise conformance to standards. Have
Adding to this group, Bill Smith, Motorola Vice President and Senior Quality
Assurance Manager, is widely regarded as the father of Six Sigma, Shina (2002),
although several have played key roles in promoting this phrase including Harry
According to Shina (2002) before, January 15, 1987, Six Sigma was solely a
statistical term. Since then, the Six Sigma crusade, which began at Motorola, has
spread to other companies which are continually striving for excellence. While it is
quality strategy and ultimately into a sophisticated quality philosophy. However, this
unique philosophy only became well known after GE’s Jack Welch made it a central
focus of his business strategy in 1995. Today, Six Sigma is considered one of the
fastest growing business management system in industry (Cook, 1990; Gill, 1990;
Rayner, 1990; Behara et al.,1995; and Maguire, 1999b). Since its initial development
and deployment at Motorola, Six Sigma has influenced virtually every sector of the
15
economy, from manufacturing to service and from the largest to the smallest
organization. Today, Six Sigma processes are being executed in a vast array of
The evolution of what is known as Six Sigma began in the late 1970s, when a
Japanese firm took over a Motorola factory that manufactured television sets in the
United States and the Japanese promptly set about making drastic changes to the way
the factory operated (Folaron, 2003). Under Japanese management, the factory was
soon producing TV sets with 1/20th the number of defects they had produced under
Motorola management. Finally, Motorola recognized it quality was awful. Since then
Motorola management decided to take quality seriously (Main, 1994; Pyzdek 2000).
When Bob Galvin became Motorola’s CEO in 1981, he challenged his company to
During 1985, Bill Smith wrote an internal quality research report which caught
the attention of Bob Galvin, Smith discovered the correlation between how well a
product did in its field life and how much rework had been required during the
manufacturing process. He also found that products that were built with fewer
nonconformities were the ones that performed the best after delivery to the customer
(Harry, 1998; Maguire, 1999b). Although Motorola executives agreed with Smith’s
supposition, the challenge then became how to create practical ways to eliminate the
Improve, Control (MAIC). Later, the MAIC discipline became the roadmap for
16
On January 15, 1987, Galvin launched a long term quality program, called
“The Six Sigma Quality Program”. The program was a corporate program which
established Six Sigma as the required capability level standard. This new standard was
After implementing Six Sigma, in 1988, Motorola was among the first recipients of
as Texas Instruments, began a similar pursuit, But, it wasn’t until late 1993 that Six
Sigma really began to transform business. That’s the year Allied Signal (Honeywell)
By adequately selecting the right Six Sigma projects and promptly providing
the right support for them, Bossidy suggested that high level executives should also
understand Six Sigma tools. At Allied Signal, an entire system of leadership and
support systems began to form around the statistical problem solving tools of Six
Sigma.
Not long after Allied Signal began its pursuit of Six Sigma quality, Jack
Some argue that many of the tools used with Six Sigma are not new
(Tadikamalla, 1994: Hahn et. Al, 1999; Watson, 2000). However, while Six Sigma
stresses the impotence of searching for a new way of thinking and doing (Maguire,
17
1999b). In fact, Six Sigma defines a clear roadmap to achieve Total Quality
(Balkeslee, 1999; Breyfogle, 1999; Pyzdek, 2000; Harry, 1998; Young, 2001).
quality initiatives. The Six Sigma approach involves the use of statistical tools within
a structured methodology for gaining the knowledge needed to achieve better, faster,
and less expensive products and services than the competition. The repeated,
disciplined application of the master strategy on project after project, where the
projects are selected based on key business objectives, is what drives dollars to the
bottom line, resulting in impressive profits. Moreover, fueled by the bottom line
work culture will be constantly nurtured, then customers will definitely be satisfied
18
CHAPTER 3
LITERATURE REVIEW
3.1 Introduction
Although Six Sigma originated in industry, it has inspired a
sources, and findings. The chapter also seeks to synthesize the literature, with an
emphasis on establishing its relationship to quality management theory and topics for
future research. In doing so, there is an attempt to answer the fundamental questions:
(i) What is Six Sigma? (ii) What are its impacts on operational performance? and (iii)
search spanning the time period from 1990 through 2003. Five descriptors were used:
Six Sigma, quality systems, quality improvement, quality management, and quality
19
meta-model. The test of each article was reviewed in order to eliminate those that were
clearly not related to “Six Sigma” improvement strategies. For example, articles were
removed that focused on detailed synthesis of chemicals and used the term Six Sigma
in character. The list of journals, proceedings, and magazines that provided at least
one relevant article is shown in Table 1. Two hundred and one (201) articles were
identified and covered in the review. The terms used to categorize these articles are
defined in section 3.3. Overall, it is not claimed that the list of articles is exhaustive,
only that the associated database serves as a reasonably comprehensive list for
The focus of this review was in three areas. First, what is the definition of Six
Sigma? A solid definitional foundation must exist before rigorous analysis can be
undertaken. Second, what is the impact of Six Sigma on a firm’s performance? Six
Sigma reportedly has saved millions of dollars for such companies as GE, but can its
benefits be quantified in the literature? Third, what has been reported on how to
implement Six Sigma in a real business setting? Is there guidance for the practitioner
20
Accreditation and Quality Assurance Journal of Evaluation in Clinical Practice
AIAA-2002-1471 Journal of Healthcare Management
Annual Quality Congress Transactions Journal of Management Engineering
Annual Reliability and Maintainability Journal of Manufacturing Science and
----Symposium Proceedings ----Engineering Transactions of the
Archives of Pathology & Laboratory Medicine ----ASME
Assembly Automation Journal of Mechanical Design
Aviation Week and Space Technology Journal of Operations Management
Building Research and Information Journal of Quality and Participation
Business Management Journal of Quality Technology
Business Month Journal of The IES
Cancer Journal Lecture Notes In Computer Science
Chemical Engineering Communications Manufacturing Engineering
Chemical Engineering Progress Milbank Quarterly
Chemical Week Proceedings of the 2001 Winter Simulation
Clinical Chemistry ----Conference
Computers & Industrial Engineering Proceedings of the 2002 Winter Simulation
Computers In Industry ----Conference
Control Engineering Proceedings of the ASME Design
Electronic Business ----Engineering Technical Conference
Fortune Professional Engineering
Genetic Engineering News Quality and Reliability Engineering
Hospitals & Health Networks ----International
Hydrocarbon Processing Quality Digest
IEEE Engineering Management Review Quality Engineering
IEEE Software Quality Management in Health Care
IEEE Transactions on Neural Networks Quality Progress
IEEE Transactions on Semiconductor R&D Magazine
----Manufacturing Radiology
IIE Solutions Research-Technology Management
Industrial Management & Data Systems Six Sigma Forum Magazine
International Journal of Production Research Technometrics
International Journal of Quality & Reliability The American Statistician
----Management The Physics Teacher
International Journal of Quality Science Therapeutic Apheresis
Journal of American Geriatrics Society Total Quality Management
Journal of Applied Statistics Total Quality Management & Business
Journal of Engineering Design ---- Excellence
Training & Development
Table 1 List of journals or proceedings with at least one article in the study.
21
3.3 The Classification Scheme
Articles were classified using the eleven descriptors in Table 2. Authors
with representatives from both. Many articles contain definitions of the phases
Define, Measure, Analyze, Improve, and Control (DMAIC) but most did not.
Two schemes were used to evaluate the primary topic(s) of each article.
Oakland (1989) divided quality issues roughly into systems, practices, people, or other
focused, without providing precise definitions of these terms. Zain, Dale and Kehoe
(2001) followed Oakland in using this division to classify articles (version 1). Sousa
and Voss (2002) developed a modified scheme based on philosophies, practices, tools
and techniques, and other (version 2). Sousa and Voss (2002) defined “philosophy” as
Those authors also described “tools and techniques” as “core elements” with examples
The Science Citation Index (SCI) provides a number called “journal impact
factor” that is a ratio between the citations to articles in a journal to the average
number of citations to journals in that field. The impact factor can be viewed as a
rough evaluation of the academic quality of the journal. Many articles made explicit
reference to either the manufacturing or service sector issues, while others offered
general contributions. A common feature of articles was mentioning 3.4 defects per
22
The articles were each affiliated with one of the following sponsoring societies
including publications sponsored by the American Society of Quality (ASQ) and the
American Statistical Association (ASA), the Institute of Industrial Engineers (IIE), the
(INFORMS) and other related journals, or the medical area in general including the
Following Zain, Dale, and Kehoe (2001), articles were classified as focused on
case studies, survey results, literature review, comparative analysis, or theoretical with
the success of Six Sigma implementations. For those articles, the specific success
factors mentioned were tabulated. The terminology used to describe the success
Finally, articles that recommended the usage of one or more practices without
clarifying conditions in which this practice has provable properties were classified as
optimization justifications.
23
Descriptor Source Levels
Authorship Brady and Allen Industrial (I), Academic (A), or Both (I A)
Define DMAIC Brady and Allen Yes (Y) or No (N)
Topics Version 1 Oakland (1989) Systems (Sy), Tools and Techniques (To), and
People (Pe)
Topics Version 2 Sousa et al. (2002) Philosophy (Ph), Practices (Pr), Tools and
Techniques (To), and Other
Industrial Sector Zain et al. (2001) Manufacturing (M), Service (Se), or General
(G)
Journal Impact Science Citation Index 0.13 to 4.76
Factor
Mention of 3.4 ppm Brady and Allen Yes (Y) or No (N)
Research Approach Zain et al. (2001) Case Study (Ca), Comparative (Co), Survey
(Su), Literature Review (R), or Theoretical with
Application (TA)
Society or Area Brady and Allen AIChe, ASME, ASQ, IIE, INFORMS, or
Medical
Success factors Brady and Allen All combinations of 13 possible factors
Speculative in Brady and Allen Yes (Y) or No (N)
Nature
derived from the classifiers described in the last section is presented. Goals include
the identification of trends including those that relate to the authorship of articles and
with research focus and a tabulation of the associated sponsoring societies or areas of
study. Finally, results relating to success factors including a tabulation of the success
Figure 1 plots the number of articles verses the year. Based on the number of
papers there is little doubt that the subject is actively reported. The plot suggests two
24
findings. First, the number of articles by industrial authors peaked in 2000. It is
hypothesize that this declining trend was influenced by condemnations of Six Sigma
in the popular press such as Clifford (2001) in Fortune Magazine. Second, at the same
time, interest among academics continued to grow in 2003. Over the entire search
period, 69.2% of the authors had industry affiliations and 30.8% had academic
affiliations. These proportions have been changing to the point where 53% of the
authors reviewed in 2003 were associated with a university or college. This trend in
considering the industrial origins of Six Sigma. It would be anticipated that the
growing interest in Six Sigma from the academic arena would add rigor and
25
45
40
Academia
35
Industry
30
Number of Articles
25
20
15
10
0
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
Year
Figure 1 The yearly number of Six Sigma related articles and their authorship.
control (SPC) in the context of Six Sigma, e.g., Mason and Yong (2000), Coleman et.
al. (2001), McCarthy and Stauffer (2001), Koch (2002) and Goh (2002). Figure 3
charts the percentages of articles associated with different areas. The fact that the
26
earliest medical related publication in the database is Buck (1998) supports the finding
that the medical area is playing an important role in increasing the topic diversity. Six
all. Rigorous academic studies are needed to question the universal validity of Six
100%
90%
% of Manufacturing Articles
80%
70%
60%
50%
40%
30%
20%
10%
0%
1990 1992 1994 1996 1998 2000 2002
Figure 3 also indicates that the applied statistics journals such as the Journal of
27
multidisciplinary nature of Six Sigma as noted by Hahn, Doganaksoy, and Hoerl
IEEE
OR+MS
Medical
ASME
AICHE
Applied Statistics
examined. We begin by focusing on the topics covered and the dependence of the
number of articles and scholarly impact on the authorship. Then, the methods used in
supports three findings. First, the majority of articles focused on either philosophy or
28
systems topics. The percentages on these topics were 54.7% and 66.2% respectively.
In general, papers in these categories provided a general description of Six Sigma and
advocated its use, e.g., Rayner (1990), Harry (1998), Snee (2001) and Does et. al.
(2002). It was found that 32% of the total articles are in this category are introductory
in nature. Overall, only 6% of the articles were written at the practices level which
Sousa and Voss (2002) argued are most useful for stimulating actual organizational
improvements. Academic authors wrote about practices with higher frequency (10%).
Second, Table 3 also shows that academics were more likely to choose topics
amenable to theoretical study such as tools and techniques and practitioners were more
likely to present philosophical or systems level contributions. For example, only 17%
The literature has focused primarily in the two areas of philosophy or tools and
techniques. These areas of study are not helpful for development of usable models to
aid decision makers in the implementation of Six Sigma. More research in needed into
29
Topics Version 1 from
Oakland (1989) Academic Industrial Mixed Totals
Systems 23 83 4 110
Tools and Techniques 20 28 5 53
People and Systems 3 18 2 23
People and Tools and
Techniques 4 5 0 9
Systems and Tools and
Techniques 0 5 1 6
Figure 4 is a Pareto chart of the number of articles associated with the different
research methods. The papers containing case studies constituted a sizable fraction of
papers on all topics. For example, 40% of the papers classified as philosophy focused
contained case studies with no new tools and techniques and 50% of the papers
exclusively on tools and techniques contained a case study with no new tools and
techniques. Of the articles with case studies the majority contained only a single case.
These articles for the most part have been descriptive in nature lacking rigor. Those
academics. It was noted that they offered more solid foundation but mostly dealt with
30
tools and techniques and where not helpful in developing a model on how and why six
Sigma works.
# of Articles
0 10 20 30 40 50 60 70 80
Case Study
Research Approach
Survey
Comparitive
Literature Review
Other
Science Citation Index (SCI) to provide a rough measure of journal quality or impact.
Figure 5 is a box and whisker plot of the journal impact factors associated with the
articles in the database. It shows that, not surprisingly, academic authors tended to
publish in journals with higher scholarly impact. Again, these dealt mostly at the
technique level and not the level of practices, which is most helpful to those decision
makers looking for guidance. Figure 6 is a box and whisker plot of the journal impact
31
factors associated with articles associated with either manufacturing or service sectors
of business or of generic interest. The plot shows that service related publications
have the highest scholarly impact. This can be attributed to the relatively greater
impact associated with the specific journals Clinical Chemistry, Radiology, and the
Journal of the American Geriatrics Society. The associated articles covered topics
4
Impact factor
0
Academic Industry Mixed
Figure 5 Journal impact factors associated with articles by different types of authors.
32
5
Impact factor 4
0
General Manufacturing Service
concerned the tools that Six Sigma black belts “should” know, e.g., the discussion of
Hoerl (2001a) and Hoerl (2001b) and Montgomery, Lawson, and Molnau (2001).
These articles were classified into people combined with tools and techniques in the
Oakland (1989) scheme and philosophy and tools using the Sousa and Voss (2002)
scheme. These articles caused both categories to be associated with the highest
median journal impact factors. Figure 7 and Figure 8 provide box and whisker plots
of the impact factors associated with the research topics. Surprisingly, the topic
33
associated with the least impact is “practices.” Sousa and Voss (2002) argued that this
Two other important themes related to people topics. First, articles focusing
and Wiklund (2002). Second, leadership and training are also popular themes (e.g.,
see Hahn, 1999, and Hoerl, 2001). Many of these articles examined of success factors
as we describe next.
4
Impact factor
0
Pe Sy Pe To Sy Sy To To
Figure 7 Journal impact factors associated with articles focusing on people (Pe), Tools
and Techniques (To), Systems (Sy), or a combination of these.
34
5
4
Impact factor
3
0
Ph Ph Pr Ph To Pr Pr To To
Figure 8 Journal impact factors associated with articles focusing on philosophy (Ph),
practices (Pr), Tools and Techniques (To), or a combination of these.
necessary for Six Sigma to succeed. Fourteen separate factors can be found among the
literature, some being in conflict with each other, Figure 9. Half would be classified as
infrastructure, half would be classified as core as defined by Sousa and Voss (2002).
The most commonly cited factor is “Top Management Commitment”, which could be
said of any organization initiative. These success factors are offered for the most part
without rigorous proof. The second most common factor was training. This high lights
35
Critical thinking
Adaptable system
Change management
Customer focused
Project selection
Team involvement
Structured approach
Data system
Team Training
primarily interested in academic contributions. It is not surprising, then, that the role
Linderman et. al. (2002), only a small fraction of the Six Sigma literature has been
devoted to theory. In this section, the description of Six Sigma to bring its
contribution into clearer focus is synthesized. Then, modifications appropriate for the
36
evaluation of Six Sigma to the quality performance model of Garvin (1987) are
suggested.
Sigma. First, it can be debated whether or not the principle of establishing monetary
justification for applying the Six Sigma method belongs in the definition. Yet,
monetary justification of projects assuredly is associated with Six Sigma. Second, Six
management practices. This fact is established by the definition in Linderman et. al.
(2002) of Six Sigma as a “method”. Also, 24% of articles defined the DMAIC phases
and many of the most popular books on Six Sigma associate specific core statistical
methods with phases (e.g., Breyfogle, 2003, Harry 1998a, and Pande et al., 2001).
Third, the books and training materials associated with Six Sigma are relatively
vocational in nature. For example, Hahn, Doganaksoy, and Standard (2001) wrote
that the aim is not to train “statistical experts”. Fourth, the most important success
factors associated with six sigma were believed to be (1) top management
top management commitment. As noted by Hahn and Hoerl (1998), money is the
37
Next, we connect greater specificity and relatively vocational materials with training
specificity about what “should” be used and when it “should” be used in the context of
a project, combined with the omission of complicated theory, would seem appropriate
usage of statistical techniques is the implied goal of Six Sigma and its main
First, the bottom line and multi-phase nature of Six Sigma has likely increased the
scope of research to embrace total projects and not just the portions associated with
the application of a single statistical method. This explains the interest on modeling
quality savings in Bisgaard and Feriesleben (2000) and why over one third of the
Second, the relatively greater emphasis on specific core (SC) methods and
specific infrastructure (SI) has spawned considerable academic discussion with greater
specificity. For example, there is a substantial academic thread focused on what tools
“should” be learned and used by Six Sigma trained participants or “black belts” (e.g.,
see Hoerl 2001a and the related discussion in the Journal of Quality Technology).
While the discussion of training materials is not new to the quality literature, the
Third, Six Sigma has caused many people from multiple disciplines to become
aware of and apply statistical methods. It is perhaps remarkable that 69% of authors
38
of academically relevant publications had industry affiliations. While, in general, Six
Sigma practitioners have learned only standard methods, they constitute a large
potential market for research and, perhaps, new methods. Distinguishing features of
this market include that participants: (1) are relatively practical and focused on
business results, (2) need techniques for predicting the bottom line impacts of projects
before they embark upon them, and (3) apply statistical methods without, in general,
In Section 3.10, we discuss the implications of these findings on the roles that
academics can most usefully play in relation to Six Sigma. Next, where the specific
infrastructure and core elements of Six Sigma fit into the quality performance model is
discussed.
process improvement and new product and service development that relies
Those authors further acknowledged that “the name Six Sigma suggests a goal” of
less than 3.4 defects per million opportunities (DPMO) for every process. However,
39
Linderman et. al. (2003) did not include this principle in the definition because, “Six
One concern with the Linderman et. al. (2003) definition of Six Sigma as a
“method” is that the definition leaves out philosophy and principles. For example,
Dean and Bowen (1994) defined quality management to include techniques and a set
(1998a), Hahn et. al. (1999), Bisgaard and Freiesleben (2000), and other seminal
literature warrants the following addition: “The Six Sigma method only fully
Montgomery (2001) argues that it is this focus on the bottom line that keeps
management interested while its predecessors like Total Quality Management (TQM)
are “dead”.
Virtually all popular books and training materials describe statistical methods
much more vocationally than standard statistical texts (Breyfogle, 2003, Harry, 1998a,
and Pande et al., 2001). Specifically, these books and training materials omit much of
the associated theory and include, in some cases, simplified versions of standard
statistical methods. Further, Hahn, Doganaksoy, and Standard (2001) wrote that the
related education goals are not to train “statistics experts” but only to give the
the following principle to the definition of Six Sigma: “Practitioners applying Six
Sigma can and should benefit from applying statistical methods without the aid of
statistical experts.”
40
Another concern with the Linderman et. al. (2003) definition is that it may be
unnecessarily vague. It can be argued from a review of the literature this vagueness in
the definition of Six Sigma is partly in an attempt to avoid controversy as most authors
have historically been practitioners advocating its use. We submit that there is
sufficient consensus within the Six Sigma literature to offer the following additional
The Six Sigma method for completed projects usually but not always
Design, and Verify (DMADV) for new product and service development,
Widely read books such as Harry (1998a) and Pande et al. (2001) clearly imply
that this refinement is part of the definition of Six Sigma. Again, Linderman et. al.
(2003) points out that this definition may only be useful for improvement efforts on
complex challenging projects. For simple tasks; the so called “low hanging fruit” this
ridged approach may not create substantial benefits. The use of an “organized and
methods … to make dramatic reductions in customer defined defect rates” would still
apply.
Harry (1998a), Pande et al. (2001), and others also imply that multiple
techniques are often used in applying Six Sigma. Therefore, the definition of Six
41
Sigma as “a method” complicates reference to the techniques used in its application.
scope to that of Six Sigma. The existence of “sub-methods” helps to connote the idea
that Six Sigma is broader than its definition as a method might imply. Also, Six
Sigma then becomes more like a “practice” than a “core method” as defined by Sousa
and Voss (2002). Sousa and Voss (2002) also defined “infrastructure practices” as
those that create “an environment supportive of the use of core practices”. With these
definitions in mind, it becomes apparent that both of the principles above are
associated with what might be called specific “Six Sigma infrastructure” (SSI)
practices.
It is also unmistakable from reading the most popular books on six sigma
(Breyfogle, 2003, Harry, 1998a, and Pande et al., 2001), and others that there is a
strong attempt to associate sub-methods with specific phases of the application of Six
Sigma. For example, the application of gauge R&R would generally not be
definition of Six Sigma. Also, any specific set of associations could justifiably be
practices with the various aspects of firm performance (Voss and Sousa, 2002).
42
Garvin (1984) introduced a quality performance model to set up an empirical
quality and product quality performance (QP) and their effects on operational
performance (OP) and business performance (BP). In reviewing the literature on Six
methods associated with “Six Sigma infrastructure” (SSI) and “Six Sigma core” (SSC)
quality management practices. It was argued that the method of Six Sigma is itself a
quality practice while sharing some characteristics with a core method. Figure 10
shows the placement of these specific core practices and infrastructure in the Garvin
(1984) model.
Six Sigma
Practice
Six Sigma
Infrastructure Business
Performance
Six Sigma
Product Quality
Performance
43
3.11 Review of Empirical Evaluations of Six Sigma
In this section, the literature relating to the performance evaluation of Six
Sigma is briefly reviewed. Only a small fraction of articles in the database pertain to
an empirical model or evaluation with scope greater than estimating the savings
associated with a single case study. Table 4 below lists five of these articles with
reference to Six Sigma core (SSC) and Six Sigma infrastructure (SSI) practices. The
other acronyms used are referenced in the extended quality performance model in
Figure 10. The fourth article by Gautreau et al. (1997) does not make specific
reference to Six Sigma but was included because it addresses issues related to the
inclusion of specific methods in the context of quality projects. This seems relevant
given Six Sigma’s emphasis on specific core methods and business outcomes.
Table 4 contains a description of the roles each article plays for empirical
validation in relation to Six Sigma. Goh et al. (2003) examined stock performance
associated with announcements of Six Sigma programs and dates of quality awards.
They found hints of short-lived abnormal returns but no significant evidence of short
or long term returns. Another data driven meta-analysis that we found was Lee
(2002), which was based on survey data. The associated surveys indicated positive
self assessments of the value of the company’ own Six Sigma efforts. Also, to our
knowledge, the impacts of specific core sub-method selection on bottom line impacts
has not been studied empirically, with Gautreau, Yacout, and Hall (1997) providing
44
Study Quality Study Main findings
performance method
model
Bisgaard and SSI→BP Economic Fraction
Feriesleben brake-even nonconforming and
(2000) analysis unnecessary activities can
model significantly influence cost
and reduce profit
Gautreau, Yacout, C→QP→BP Partially Decision based model
and Hall (1997) Observed of process improvement
Markov activities, e.g., do nothing
Decision or inspect and separate can
Process each be optimal depending
on assumptions
Goh et al. (2003) SSI/SSC→BP Hypothesis The majority of firms
(stocks) testing show positive returns after
announcing Six Sigma
programs but no statistical
significance was
established
Lee (2002) SSI/SSC→BP Survey of 106 Top management
firms commitment, project
selection, team leader,
training and the specific
tools used effect business
results
Linderman, SSC→QP/OP Goal-theoretic Types of goals effect
Schroeder, Zaheer →BP model quality and operational
and Choo (2003) performance that effect
business results
45
Considering the emphasis on modeling profits to justify each project, it is not
surprising at some attempts to provide meta-modeling tools, e.g, see Bisgaard and
product value under simple, generic assumptions. As they themselves suggest, more
research and its relationship to quality management theory was clarified. Reflecting
on the findings, we next return to the questions: (i) What is Six Sigma? (ii) What are
its impacts on operational performance? and (iii) What roles can academics usefully
Analyze, Improve, and Control (DMAIC) or Define, Measure, Analyze, Design, and
Verify (DMADV) as phases. This definition of Six Sigma as a method builds on the
one proposed by Linderman et. al. (2003). The inclusion of DMAIC and DMADV in
the definition is supported by the fact that 75% of introductory articles on Six Sigma
reference these structures. Two principles are included in this definition. The first
emphasizes attention to the bottom line in initiating projects. This was supported by
46
comments of seminal writers relating to how Six Sigma differs from Total Quality
Management (TQM), e.g., Harry (2001a) and Montgomery (2001). Also, bottom line
focus was mentioned by 24% of relevant articles as a critical success factor. The
This inclusion is based on remarks by Hahn, Doganaksoy, and Standard (2001) and
others about the goals of Six Sigma training. In addition, relatively frequent mention
relevant articles) is found. Finally, popular books on Six Sigma such as Breyfogle,
(2003), Harry (1998a), and Pande et al. (2001) noticeably de-emphasized theory.
Sigma is that such justifications are needed for each Six Sigma project. For example,
Dupont reported an expected profit from Six Sigma practices of over $1 billion
between 1999 and 2003 (Noble, 2001). Also, Goh et al. (2003) found hints of short
lived abnormal stock performance associated with the decisions to start Six Sigma
claims nor evidence of long term effects. Those authors also included specific caveats
about the ability to connect Six Sigma programming effects at divisions with overall
parent company performance. Considering that failure to find significant effects does
not constitute proof, the conclusion is that more work is needed for a thorough
47
In Section 3.8, we identified three main types of contributions of Six Sigma to
academia embodied in the literature: (1) increased emphasis on complete case studies
compared with single sub-method applications, (2) new, relatively specific core and
infrastructure practices, and (3) the development of a large new market of industrial
non-experts who might be interested in practically oriented research and new methods.
Many authors have proposed areas for further research building on these contributions.
Also, Cooper and Noonan (2003), Linderman et. al. (2003), and Snee (1999) suggest
that, in general, too much research has been focused on descriptions of practice rather
three groups. First, Sousa and Voss (2002) highlighted the need for empirical
context of Six Sigma, statements abound that are unsupported by objective evidence.
Examples include self reported profits, the effects of success factors, and advocacy for
Six Sigma in general. For example, as noted above, the impacts on stock performance
investigated by Goh et al. (2003) are not fully resolved. This suggests a need for
additional data collection and analysis to answer the important question of long term
Second, while over 50% of the articles in the database either explicitly or
implementing Six Sigma in specific business contexts has, apparently, not been
investigated. Related, largely unanswered questions include: How can data about any
48
specific company’s management, training programs, or environment be useful in
Third, Snee (1999 and 2000a) calls for research to help practitioners identify a
robust set of improvement tools to be used in conjunction with the DMAIC process.
3.8, new techniques might be relevant to Six Sigma practitioners who are often not
experts in statistics.
impacts of projects are needed. This follows because of the central importance of
profit related justifications in Six Sigma for initiating decisions on projects. Also,
Bisgaard and Feriesleben (2000) admit that the assumptions associated with their
models have limited scope and may ignore indirect savings. New models that are also
accuracy.
models to aid practitioners from different disciplines select the most advantageous
techniques. This could build on research related to the most appropriate methods for
training black belts, e.g., in Hoerl (2001a), by associating the methods more
49
In conclusion, it is proposed that Six Sigma is both a method and two
support and to fostering usage of methods among practitioners who are not experts in
broader focus than solely on manufacturing. Only partial consensus about the factors
making Six Sigma effective is found. Opportunities for new research on Six Sigma
including creating more realistic project payback models, clarifying which techniques
are most applicable in which situations, and developing new statistical methods with
disciplines, the relevant material is scattered across various journals. The search
resulted in the identification of 201 articles published between 1990 and 2003.
insights into the state-of-the-art. It is felt that the results presented in this chapter have
and many companies have gained the profits and advantages from this disciplined
approach, the literature is limited and the research of the impacts of Six Sigma
implantation and factors contributed to Six Sigma success remain unclear. Many
50
articles on the impact analysis of operations performance do not mention the detailed
improvements in the operating areas such as scrap rate, rework rate and so on, but
focus on the overall bottom line impact [Breyfogle (2001), Noble (2001) and Lucas
(2002)]. Therefore, it is necessary to do a deeper and more detailed study in this area.
In addition, only a few articles were found that dealt with factors in the area of
success factor analysis to Six Sigma implantation. Even the existent studies are not
well integrated and the research is mostly anecdotal. Current concepts in the field of
Six Sigma are largely based upon case studies, anecdotal evidence and the
factors are critical to the success of the approach. Most of the articles reported that top
management leadership is the main factor to Six Sigma success [Blakeslee (1999) and
Scalise (2001)]. However, many other factors affecting Six Sigma’s success are
Even so, there is substantial evidence that Six Sigma and other quality
management systems have a positive effect on the value of a company. Goh (2003),
Hendricks (2000) and Adams (1997) all found improvement of companies financial
authors have produced broad frameworks for implementing and sustaining competitive
Although some authors have called for theoretic research (see Cooper and
Noonan 2003, Linderman et. al. 2003 and Snee 1999), too much research is focused
51
managers and scholars. The attempt to build a theory of how and why Six Sigma
works is aimed at building a prescriptive model. From this, managers would be able to
identify which activities from which programs are more or less likely to be useful in
their situations, as well as which of their goals would be most affected. With the
future success of corporations riding on the outcome, there has been little theory to
52
CHAPTER 4
This chapter discusses one of the major themes of Six Sigma: training of
practitioners. From the literature, Six Sigma differs from most quality management
statisticians in the use of statistical tools. A major concept then becomes how to
effectively train participants in the necessary tools and techniques of Six Sigma. In
this chapter, we describe one case study and associated exercises used in senior and
introductory graduate level engineering courses on SPC and DOE at The Ohio State
University. The role of the case study exercises in the context of material covered in
the lectures is described. Section 4.2 provides an overview of the case study structure.
In Sections 4.3 a case study and associated in-class exercises is presented. Finally, a
description of the student feedback to the case based teaching approach is reviewed.
53
4.1 Introduction
One constant tread in the literature dealing with Six Sigma is the training and
the limited chose that non-statisticians have to make. For thousands of participants at
the lowest “green belt” level of accreditation, one of the main benefits of “Six Sigma”
training is that it simplifies (through restriction) the sequence and choice of available
techniques to apply to a particular case. Therefore, the value of the Six Sigma
movement derives partly from standardization of problem solving methods and partly
improvement project.
successful methods that is used to teach skills: breaking a task into small bits and
presenting it to the student with practical, logical steps that he or she can understand
and apply. By simplifying the decision-making model and helping our students to
define and operate within a well-defined “box” of their limitations, we have a practical
decision making is to reach the right decision at the right time. We can make the
Now, rather than trying to teach an ambiguous “good judgment” concept, we are
54
Drawing a box for our students involves expanding on the familiar concept of
personal knowledge and capabilities. The first step in the model is the ability of a
student to operate within his or her own capability. The second step involves looking
at a given situation, evaluating current skills, and setting boundaries for that specific
situation. In essence, you’re drawing a box of operating parameters for your student in
that situation. The third and final step is to insist that he or she always operate within
that box.
By drawing the box before the situation a good decision was made before a
crisis forced the student’s hand. Everyone’s box will be different, since we all have
different levels of experience and ability. As proficiency is gained, the size of the box
will change.
Teaching decision making is a daunting- some would say impossible task, but
one that is imperative. By giving students the tools and ability to draw their own box,
Six Sigma has caused many people from multiple disciplines to become aware
Sigma practitioners have learned only standard methods, they constitute a large
potential market for research and, perhaps, new methods. Distinguishing features of
this market include that participants: (1) are relatively practical and focused on
business results, (2) need techniques for predicting the bottom line impacts of projects
55
before they embark upon them, and (3) apply statistical methods without, in general,
literatures that exercises based on case studies can play an important role in
process control (SPC) and design of experiments (DOE) techniques, e.g., see Alloway
(1993), Barton and Nowack (1998), Cobb (1992), Nolan and Speed (1999),
Petruccelli, Nandram, and Chen (1995). There is also abundant evidence, e.g., Czitrom
apply the technology they have learned and in general see their class experiences as
they are exposed to the realistic engineering contexts in which the related methods are
engineering, Howell (1996), McKeachie (1993,) and Wankat (1993) gives us insight
into restructuring traditional class outlines into a more productive approach. For these
in-class case study based exercises that emphasize the other forms of knowledge and
reinforce the lecture materials. These exercises also seem to aid in establishing
A major challenge in using case studies in classes has been that the examples
in standard textbooks generally seem highly contrived to the students and divorced
56
from real life contexts. As described by Bisgaard (1998) and in our own experience, it
is important that the case study examples used in the course are impressive to the
students in their relevance to their own possible careers. Therefore, we have found that
it is helpful to use exercises based closely on case studies from local companies in
which the application of the techniques taught in the course played a key role in
exercises is that the students can visualize themselves in the role of facilitating real
world successes.
The case study based exercises were introduced into the two combined senior
and first year graduate level courses at The Ohio State University. The classes meet
twice a week for 10 weeks, each period is for one hour and twenty minutes. Many of
the students have had two quarters in statistics. Topics addressed in the SPC courses
can be divided into four categories: (1) a survey of quality techniques and a review of
probability and inference relating to continuous random variables, (2) control charts
for continuous random variables, (3) review of probability relating to discrete random
variables and attribute control charts, and (4) acceptance sampling and assorted topics
including ISO9000 and “six sigma” methodology. The topics in the DOE class are: (1)
t-testing and ANOVA, (2) screening experiments, (3) response surface experiments,
approximately five class periods and one homework assignment. The case study
exercises are timed on the days after each of four assignment due dates, so that the
57
The case studies are true but for proprietary concerns some are modified in
order to maintain the confidentiality of the companies involved. Each case study is
selected from area industries to provide relevant experience and important evidence of
the practicality of the statistical techniques described in the lecture and textbook.
problem, 4) questions for the students, and 5) the team implementation of course
techniques and results from the actual study. The exercise begins by handing out
description of the industry in question and some of the standards, issues and
terminology used so the students have some familiarity with the subject. The
introduction also serves the important role of establishing that the students might
realistically find themselves in similar situations in their future careers. In the problem
statement, we use 5 minutes of lecture time to describe the goal, e.g., 5% defective
products from a particular process, and the associated business objective, e.g., avoid a
$2.5 million capital expenditure to meet increased product demand. In our oral
description of the problems, we encourage creative analyses and emphasize that there
background information which includes the relevant data that was available to the
58
company at the time of the team intervention involving class methods and sufficient
information such that the students can be expected to provide their own analyses and
recommendations. Our goal is to present the problem such that the students should be
able to easily identify themselves as playing the role of consultant and advocate of
questions. These questions involve both critiquing the results of earlier investigations
and developing recommendations for future study. The students are divided into
groups of 4-6 and given a 30-minute exercise period to answer the questions. Students
can ask technical questions and when relevant we sketch results on the blackboard
including, e.g., formulas for sample size estimation and additional details requested by
the students. From group input, we use 15 minutes class time to construct composite
answers to questions in class and write these answers on the blackboard. During this
proposed approaches while preserving the students intent. After the answers from the
class are on the board, we use the remaining class time to describe the actual
illustrates how with a small number of quality technology tools highly valuable
59
information can be learned. This case study involves the use of Pareto diagrams, p
charting, hypothesis testing, and fractional factorial designs. The study also illustrates
new advanced product that quickly captured 83% of the market in North America.
repair) had stabilized in the 70% range. In early 1999 the product was selected as a
demand, the company needed to purchase additional test and repair equipment at the
cost of $2.5 million or the first test yield had to increase to above 90%. The latter was
the preferred situation due to the substantial savings in capital and production labor
cost thus, the problem was how to increase the yield in a cost-effective manner.
manufacturing areas throughout the company. Their task was to recommend ways to
improve the production yield based on their prior knowledge and experience. Table 5
gives the weekly first test yield results for the 16 weeks prior to the team’s activities
60
Week Yield Week Yield Week Yield Week Yield
1 71% 5 87% 9 66% 13 63%
2 58% 6 68% 10 70% 14 68%
3 69% 7 71% 11 76% 15 76%
4 77% 8 59% 12 82% 16 67%
Table 5 The yields achieved for 16 weeks prior to the initial teams activities.
improvement team reviewed the design and production process. They created a list of
15 potential process and design changes for improvement based on their engineering
judgment and anecdotal evidence. With this list in hand, they proceeded to run various
single factor experiments to prove the validity of their plan. Due to perceived time and
cost constraints, only one run of each factor was completed using a sample of 30 units.
Results were compared with the yield from the previous 13 weeks of production. Each
unit could only succeed or fail to meet specifications. Factors that showed a yield
decrease below the 13 week average were discarded along with the experimental
results. Table 6 shows the results of the experiments with yield improvements
Based on their analysis of the circuit, the above experimental results and past
experience the improvement team predicted that a yield improvement of 18% would
result from their proposed changes. All of their recommendations were implemented
at the end of week 17. Table 7 gives the weekly first test yields results for the six
61
FACTOR YIELD IMPROVEMENT
Replace vendor of oscillator 8.5%
Add capacitor to transistor 8.5%
Add RF absorption material 5.5%
New power feed layout 5.5%
Increase size of ground plane 2.5%
Lower residue flux 2.5%
Change bonding heat sink 2.5%
Solder reflow in air vs. N2 2.5%
Temperature of solder tips 2.5%
Table 6 The initial team's predicted yield improvements by adjusting each factor.
Reviewing the data we see that the yield actually dropped 29%. On week 22 it
was apparent that the proposed process changes were not achieving the desired
62
4.6 In-class Exercise for PCB Study
In our SPC classes, we then give the students the following questions based on
the premise that the company is hiring them as a quality and manufacturing consultant.
(1) Critique the methods used by the engineers for predicting improvements, e.g., is
there any evidence that they increased the yield? Roughly how many samples would
they need for the standard deviation of the estimated yield to approximately equal 5%,
i.e., they could begin to resolve differences of approximately 5%? (2) What additional
procedures would you recommend to measure the process capability? Include in your
answer what specific types of data or information would you request or collect. (3)
What addition procedures would you recommend to aid in improving the process?
In DOE classes, we ask the students: (1) What size of experiment should they
start with? (2) How should they determine which runs to perform? A representative
from each team presents their group’s answers in class, which are written on the
board. The entire class then comments on the reviews and coherent consulting answers
to the questions are created. In the real study, Pareto Charts were used to direct the
investigation, p-charts were used to measure the common cause variation, and formal
approaches were used to improve the process. Therefore, a wide variety of class
techniques are relevant, presenting life-like ambiguities for the students. If we do not
homework assignments. After the class discussion we relate in detail what actually
63
4.7 The Team Actions and Results
The second team’s first step was to construct a yield attribute control chart (a
yield chart or 1- defective chart “1-p”) with the knowledge of the process change date
(Figure 1). The engineers were able to tell that most of the fluctuations in yield
observed before the team implemented their changes were, as Deming calls it,
common cause variation or random noise. The engineers’ first decision was to revert
back to the original, documented process in place during week 16 since the evidence
that had supported these changes was probably due to random noise within the
process. Table 8 gives the weekly test yields for the five weeks after this occurrence.
This had the effect of restoring the process to its previous in control state with yields
around 75%. The increase in yield shown on the control chart (Figure 11) during this
time frame was discounted as random noise or the “Hawthorn” effect since no known
attention to and study of the process. Next the procedure of Pareto charting was
64
100%
90%
UCL
80%
70%
Y ie ld CL
60%
50%
LCL
40%
30%
0 10 20 30 40
Subgroup
Week 24 25 26 27 28
Yield 62% 78% 77% 75% 77%
Table 8 The yields for the 5 weeks subsequent to the initial intervention.
65
ACP –30 kHz 6.7%
ACP +30 kHz 5.0%
ACP –60 kHz 2.9%
ACP +60 kHz 2.8%
ACP –90 kHz 2.7%
VDET 1950 mHz 2.4%
VDET 1990 mHz 2.0%
ACP +90 kHz 1.4%
Power supply 1.2%
Gain 1.0%
Output return .8%
Bias voltage .5%
Max current .2%
Other .4%
Table 9 The data for a Pareto chart of the data from week 1 to week 16.
It can clearly be seen from a Pareto graphs that 21.5% of the 30% of total
defects was attributed to one parameter called “ACP”. The engineers concentrated
their efforts on this dominant defect code. The team of experts was reassembled with
the addition of representation from the production workers to identify what variables
might cause this defect. Four factors identified were: (1) transistor performance
distribution (high end of spec or low end of spec), (2) transistor mounting (socket or
solder), (3) input circuit tuning (centered or low end of spec), and (4) transistor heat
sink type (current or new configuration). This last factor was added at the request of a
representative from production. This factor was not considered important by most of
the engineering team. The two lead engineers decided to include this factor as the
66
marginal cost of adding it was small. An eight run fractional factorial experiment was
Table 10 The data from the fractional factorial design in four factors.
Using ANOVA and main effects plots, the heat sink type emerged as the main
effect. The two engineers went forward based on the DOE results and recommended
that the process change to the new heat sink. This was implemented during week 29.
Table 11 gives the weekly yield results for the period of time after the recommended
change was implemented. Using the yield charting procedure, the engineers were able
to confirm that the newly designed process produced a stable first pass yield in excess
of 90% thus avoiding the equipment purchase and saving the company $2.5 million.
The main points that we emphasize in our wrap-up of the exercise are the
fourth factor to be added without additional runs being needed. The importance of this
factor was controversial because it had been suggested the operators. If fractional
67
factorial had not been used, then the additional costs would likely have precluded its
inclusion and the important subsequent discovery. Second, the case also illustrates the
experiments are at each level of each factor providing the maximum power to identify
small effects.
33 91% 38 92%
traditional lectures. Before implementation of the case study exercises, end-of quarter
evaluations by students for the two classes had been below the average end-of-quarter
class evaluations in the department. After the implementation of the case study
approach and with minimal other changes, the evaluations climbed well above
average. We attribute this dramatic turn-around mainly to the introduction of the case
68
based method. Further, we conjecture that presenting applications of course methods
concerned about job security and advancement in the highly competitive industrial
marketplace.
It is our belief that these results should generalize to other statistics, operations
research, and engineering design theory related courses. This follows because all of
these courses are fundamentally concerned with the students' usually voluntary
discipline and motivation, we feel that both practicing the application of techniques in
69
CHAPTER 5
MESO-ANALYSIS
5.1 Introduction
One outcome of the wide spread acceptance of Six Sigma and with its
systematic program is the fact that many databases describing the performance of
improvement projects and the methods used have been generated. Even with its
apparent highly visible success, one area of investigation is the use of this new source
of data, the industry Six Sigma database, for additional insight into the management of
70
Martin (1982) pointed out that the availability of certain type of data might
conclusions drawn. Reviewing the literature with historical data sources reveals a
Six Sigma methodology at a company, Bisgaard and Freiesleben (2000), Chan and
Spedding (2001), Gautreau, Yacout and Hall (1997), Yacout and Gautreau (2000), and
Yu and Popplewell (1994). This is mostly based on single case studies and anecdotal
component tools and techniques for green belts and black belts. These are the terms
used for the individuals working on project implementation. Little work is published
that relates to the meso level of managing and decision-making of Six Sigma, the mid-
system, Linderman et. al. (2002). The uses of these databases are likely being ignored
at most companies for at least two reasons. First, there is little assistance form
academics in how to make sense of them. Second, the people with the most statistical
expertise are involved in the individual projects and not in cross project evaluation.
Most managers are not statisticians and need help in making sense of the data now
available to them.
According to Juran and Gryna (1980) the activities in companies that assure
quality can be grouped in three processes: quality planning, quality control and quality
71
Typically, quality improvement activities are conducted in projects. This proactive and
project wise nature distinguishes improvement form quality control, which is an on-
line process that is reactive in nature. In Harry (1994) all things are a process. A
central belief of Six Sigma is that the product is a function of the design and the
manufacturing process which must produce it. This is symbolized as Y=f(X), where Y
monitored”. The X is described as independent, input, cause, problem, and its role as
“to be controlled”. The view is that the emphasis should shift from monitoring Y to
With Juran and Harry in mind, Six Sigma can be viewed as a process and
subject to the same controls and improvement objectives of other processes. Against
this background, the purpose of this study was to look at this growing database in a
activities could be useful in the empirical study of some important research questions.
Potential research topics include: the health of a given company’s quality system,
modeling Six Sigma, or the optimality of selection and ordering component methods
associated with Six Sigma. Researchers focus on what they have data and tools for.
Now, new data sources and the associated ability to ask and answer new types of
questions are more readily available. For example, “Is my quality system
72
out-of-control?” “Which method would lead to greatest expected profits in my case?”
these kinds of questions can be systematically explored in the Six Sigma discourse
then important lessons can be learned regarding investment decisions. Moreover, there
will be an increase in critical writings and inquiry on this subject, which will add
depth and meaning to Six Sigma in organizations. To help in this study empirical data
company was collected over a 30-month period starting in January 2002. A total of 39
company, which manufactures components for the aerospace, industrial and defense
industries. It has approximately 1000 employees, annual sales of $170 million, with
six factories located in five states. The data is all derived from one of its six
manufacturing sites. This one site has 250 employees with sales of $40 million.
for this company. The ability to predict project savings and how best to manage
project activities would be advantages to future competitive posture and the long-term
73
5.4 General Procedure
Over the course of this study data was collected on 20 variables. Two
additional dependent variables are tabulated in the data, which are functions of some
of the other variables. The two variables are Profit, which is Actual Savings minus
cost and Formal Methods (FM) which is any combination of Charter, Process
Mapping, Cause & Effect, Gage R&R, DOE or SPC. Table 12 lists and describes the
Data was collected on each project by direct observation and interviews with
team members to determine the use of a variable such as DOE or Team Forming. No
attempt was made to measure the degree of use or the successfulness of the use of any
variable. We only were interested if the variable activity took place during the project.
A count was maintained if an activity was used multiple times such as multiple DOE
runs i.e. a screening DOE and an optimization DOE would be recorded as 2 under the
variable heading.
Expected Savings and Actual Savings are based on an 18 month period after
implementation. The products and processes change fairly rapidly in this industry and
Costs were tracked with existing company accounting procedures. All projects
where assigned a work order for the charging of direct and non-direct time spent on a
specific improvement activity. Direct and non-direct labor where charged at the
74
average loaded rate. All direct materials and out side fees (example, laboratory
analysis) where charged to the same work order to capture total cost.
at one manufacturing site over a 30 month time frame. Not all programs followed
DMAIC nor was it the intent. Our intent was to observe unbiased projects in a real
interviews determined the use or non use of a tool. No measure of the degree of use
was attempted. We only indicated whether or not a tool was used and how often. The
number of times a tool was used during a project was recorded and listed in the data
summary, Appendix B. These projects did not include all activity but where judged to
be project unbiased by the author. As a manager at the facility where the study was
conducted one author could influence the activities. The studied projects where
75
Field Description
Expected savings An estimate of the projects saving over an 18
month period based on the current business
forecast.
Expected time An estimate made at the start of a project as to the
time needed to complete the project
s-short less than 3 months
m-medium between 3 and 9 months
l- long over 9 months
M/I management or self initiated Whether the project was initiated by management
or initiated by team members
Assigned or participative Whether the project was assigned to a team by
management or the members actively chose to
participate
# people Number of team members
EC Economic analysis A formal economic analysis was preformed with
the aid of accounting to identify cost and cost
brake allocations
CH Charter Formally define project scope, define goals and
obtain management support
PM Process Mapping Identify the major process steps, process inputs,
outputs, end and intermediate customers and
requirements; compare the process you think
exists to the process that is actually in place
CE Cause & Effect Fishbone diagram to identify, explore and display
possible causes related to a problem
GR Gage R&R Gage repeatability and reproducibility study
DOE A multifactor Screening or optimization design of
experiment
SPC Any statistical process control charting and
analysis
76
Table 12 Definition of variables (Continued from last page)
As stated in section 3.11 one of the main principles of Six Sigma Harry
(2001a) and Montgomery (2001) is the emphasis placed on the attention to the bottom
line results in initiating projects. In the literature reviewed, bottom line focus was
combination of cost and actual savings, than becomes the dependent variable. The
remaining eighteen variables are the independent variables. Of these variables, Process
Mapping, Cause & Effect, Gage R&R, DOE, and SPC are related to Formal Methods.
77
These 39 improvement projects generated a total of $4,385,099 in net savings
(profit). The projected savings form these projects ranged rorm $1600 to $2,200,000
with actual net savings of -$220,000 to $3,874,500 over the first 18 moths after
implentation. Cost to implement the projects ranged from $1,000 to $325,500. Fifteen
of the thirty nine projects resulted in negative net savings as reported. Table 13
presents descriptive statistics for project duration, number of team members and
profits.
Table 13 Summary statistics for the 39 projects, duration team memberrs and profits
the literature was team participation. Of the projects, 72% were initiated by
management directive. Management assigned the team members in 48.7% of the tasks.
The other teams were self-initiated by the team members. Additional characteristics of
78
Item Number in Percent of the Study
Study Sample
Economic Analysis 15 38.5%
Formal Charter 28 71.8%
Team Forming Exercise 12 30.8%
Process Mapping 21 53.8%
Cause & Effect 11 28.2%
Gage R&R 7 17.9%
Design of Experiment 11 28.2%
Statistical Process Control 7 17.9%
Formal Documentation 28 71.8%
Engineering Analysis 29 74.4%
One Factor Experiment 10 25.6%
regression, Markov Decision Process, and SPC. Each analyses tool was used to look at
(SPC). The control chart is a very useful process monitoring technique; when unusual
sources of variability are present, sample averages will plot outside the control limits.
79
methodology (Six Sigma) or changes to a process such as did training have a positive
The first step in deriving a process control chart is to check the assumption of
normality. Figure 12 in a normal probability plot of the project data. It can be seen
from an examination that the data is comprised of two populations and two outliers.
2.5
2
1.5
1
0.5
z-score
0
-2000000 -0.5 0 2000000 4000000
-1
-1.5
-2
-2.5
Profit ($)
With the data base in the example the logical subgroup size is n=1. With only
one measurement per subgroup (a project) a subgroup range can not be calculated. The
80
weighted moving-average (EWMA) control chart is typically used with individual
defined as:
Z i = λx i + (1 − λ ) Z i −1 (5.1)
where 0 < λ ≤ 1 is a constant and the starting value is the process target so that
Z0 = μ0 (5.2)
Z0 = x (5.3)
UCL = μ 0 + Lσ
λ
(2 − λ )
[1 − (1 − λ ) ]
2i
(5.4)
CL = μ 0 (5.5)
LCL = μ 0 − Lσ
λ
(2 − λ )
[1 − (1 − λ ) ]
2i
(5.6)
work well, with λ=0.05, λ=0.10, and λ=0.25 being popular. L values between 2.6 and
3.0 also work reasonably well. Hunter (1989) has suggested values of λ=0.40 and
health of a quality improvement methodology like Six Sigma is savings per project.
81
From the database under investigation and the above equations a control chart in can
be generated. We start with plotting the first 25 points to obtain the control limits as
shown in Figure 13. One out of limit point was discarded after the derivation of this
chart. This one project was a DFSS (Design for Six Sigma) vs. a DMAIC project and
as such was unique. This chart was constructed with based on Hunter (1989) with
2000000
1500000
1000000
Profit ($)
500000 UCL
zi
0 LCL
1 3 5 7 9 11 13 15 17 19 21 23 25
-500000
-1000000
-1500000
Project
82
EWMA Chart for Profit
150000
100000
50000
Profit $
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37
-50000
-100000
Six Sigma Project
The projects are listed in rough chronological order therefore Figure 14 can be
considered a time series graph. One major outlier was removed from the plot.
Traditionally since their introduction by Walter Shewart, control charts have been
unpredictable behavior, then the underlying process which gives rise to the time series
83
through the use of past experience, we can predict, at least within limits, how the
because prediction is the essence of doing business. Predictability is a great asset for
any process because it makes the manager’s job that much easier. The fact that a time
series remains within the computed limits, and that there is no obvious trend, not any
long sequences of points above or below the central line, suggests that a process may
display a reasonable degree of statistical control. In this case, the process has an
average profit per team of $112,438 and values have varied form a low of -$220,000
to a high of $3,874,500
Of special interest are the last seven projects that took place after Six Sigma
training provided by Honeywell and Raytheon. This would tend to indicate “proof”
that training helps in project improvement. This is limited data but future study with
5.7 Regression
The popularity of ordinary linear regression models is attributable to its low
its support by a broad and sophisticated body of statistical inference. Given the data,
the tool of regression can be employed on at least three separate conceptual levels.
84
First, it can be applied mechanically, or descriptively, merely as a means of curve
fitting. Second, it provides a vehicle for hypothesis testing. Third, and most generally,
data may be brought together to increase our understanding of complex physical and
social phenomena. Simply put regression can be used to derive a model that will help
In most regression analysis the coefficients of the independent variables are the
major point of interest. The slope of the regression line measures the relationship of
coefficients are these slopes. The most interesting parameter in a linear model is
usually the slope. In this equation the slope of Y on X1 has a constant value across the
range of X2. For complex systems researcher may notice that regressions based on
different subsets of the data produce very different results, raising questions of model
caused by interaction between the independent variables and the response. Terms
involving products e.g. f1,7(x) = x1x7, are called interaction terms, a relationship
85
The effect of this first order interaction is to shift the regression line based on
the level of the independent variables producing a family of results. The X1X2
interaction signifies that the regression of Y on X1 depends upon the specific value of
has the effect of a buffer. That is while X1 is positively related to Y the strength of this
have been published suggesting key Six Sigma elements and ways to improve the
management of the total quality of the product, process, corporate and customer
supplier chain.
independent entity affecting the Six Sigma environment. But the extent to which one
factor is present may affect the other factor. The estimate of the net effect of these
interacting factors is responsible for the success of the Six Sigma philosophy.
Quantification of Six Sigma factors and their interdependencies will lead to estimating
the net effect of the Six Sigma environment. The authors are not aware of any
86
5.8 Regression Data Analysis
The data was analyzed using a commercial software regression modeling
EXCEL spreadsheets and has the capability of simple regression models, multiple
regression models or custom modeling. The software regression model features permit
the user to select models automatically or manually. They facilitate fitting high order
The motivation for this study was to evaluate the use of an emerging database
business environment for the analyses. The data was analyzed using a commercial
polynomial software regression modeling program. One of the models that was
derived from the data was an interaction model between the use of Formal Methods
and Engineering Analysis. This model illustrates the possible use of these techniques
and appears helpful in validating the importance of estimating expected savings and
then deciding how formal to be. The model can be represented by equation 5.9:
Where:
87
FT is the number of Formal Methods used,
Figure 15 is the main effects plot produced by the software for this model. It
can be seen that as the expected savings increases the model would suggest a greater
reliance on the use of formal methods. Likewise, based on this limited database, as the
The results of the regression model are presented in Table 15, with the
summary statistics given in Table 16. A 3-D plot of the results is included in Figure
16. Last, normal scores vs. residuals are found in Figure 17.
12000000
10000000
8000000
Predicted Response
6000000
4000000
2000000
0
Exp Savings:1600
Savings:2.2e+006
Savings:1.1008e+
TF:0.5
FT:3.5
Time:15.5
k2:I
k1:L
DOE:0
SPC:0
SPC:1
SPC:2
DC:0
DC:1
DC:2
EA:0
EA:2
EA:4
OF:0.5
#people:1
#people:5
#people:9
EC:0.5
CH:0.5
PM:0
PM:1
PM:2
CE:1.5
GR:0
GR:1
GR:2
DOE:2
DOE:4
k1:M
k2:M
k1:S
k3:A
k3:P
-2000000
TF:0
TF:1
FT:0
FT:7
Time:30
EC:0
EC:1
OF:0
OF:1
CH:0
CH:1
CE:0
CE:3
Time:1
-4000000
Exp
-6000000
-8000000
-10000000
Factors
88
Summary statistics from the model are presented in Table 15. The model
produced an R2 value (observed vs. predicted) of 0.934 and an SSE (Sum of Squared
Errors) of 0.883.
Summary Statistics
Criterion Value
R^2 0.934
R^2 adj 0.924
R^2 predict 0
R^1 0.743
PRESS 4.54E+15
s (est. err.) 171190.53
SSE(LSE)/SSE(LAD) 0.833
Coefficient Estimates
Standard p-
Coefficients Error t Stat value Lower 95% Upper 95% VIF
const 79930 62285 1.28 0.208 -46790 206652
Exp Savings 0.85 0.80 1.06 0.297 -0.78 2.47 101.65
FT -76708 19028 -4.03 0.000 -115422 -37994 2.04
EA 1258 56503 0.02 0.982 -113699 116216 2.22
Exp Savings*FT 2.20 0.17 13.09 0.000 1.86 2.54 222.24
Exp Savings*EA -4.04 0.22 -18.37 0.000 -4.48 -3.59 123.45
89
Profit
20000000
15000000
10000000
15000000-20000000
5000000
10000000-15000000
Profit 0 5000000-10000000
-5000000 0-5000000
-10000000 -5000000-0
-15000000 6.2 -10000000--5000000
-20000000 3.1 FT -15000000--10000000
1600.0
-20000000--15000000
490133.3
978666.7
0.0
1467200.0
1955733.3
Exp Savings
90
Normal Scores vs. Residuals
2.5
2
1.5
1
Normal Scores
0.5
0
-800000 -600000 -400000 -200000-0.5 0 200000 400000
-1
-1.5
-2
-2.5
Residuals
This model is intuitive and appears to provide a good fit. With higher expected
savings it seems logical to apply more formal methods and obtain higher profit as the
model predicts. With any regression questions should be asked as to the goodness of
91
An examination of the variance inflation factors (VIF) associated with this
model would indicate that they are unacceptable and this model should not be used for
predictions. Variance inflation factor measures the impact of collinearity among the
regression model we expect a high variance explained (r-square). The higher the
variance explained is, the better the model is. However, if collinearity exists, probably
the variance, standard error, parameter estimates are all inflated. The high variance
might not be a result of good independent predictors, but a mis-specified model that
carries mutually dependent and thus redundant predictors. Scaling the inputs does not
always cause improvements in the VIFs, as in this example problem. Some model
Where:
92
The main effects plot for this model is shown in Figure 18, with the results of
the regression model presented in Table 19, and the summary statistics given in Table
18.
60000
Predicted Response
40000
20000
0
k1:M k1:I Training:-1 Training:0 Training:1
93
Summary Statistics
Criterion Value
R^2 0.194
R^2 adj 0.148
R^2 predict 0.0251
R^1 0.102
PRESS 1.33E+11
s (est. err.) 56133.32
SSE(LSE)/SSE(LAD) 0.827
Coefficient Estimates
Standard Lower Upper
Coefficients Error t Stat p-value 95% 95% VIF
-
const 13509.53 12750.01 1.06 0.297 12374.36 39393.43
M/I_I 38856.45 21245.49 1.83 0.076 -4274.20 81987.09 1.06
Training 19566.38 11002.61 1.78 0.084 -2770.10 41902.86 1.06
while most of the current literature is based on only a single case study. Specifically,
our results indicate the importance of estimating expected savings with an economic
study prior to the project start. With this information management can more
94
effectively decide how formal to be in the approach to the proposed project. From our
data the brake even point is $12,500. If a project is projected to save less than $12,500,
formal Six Sigma methods are not advised. Instead relying on quick engineering input
On projects over $12,500, the model states formal quality improvement tools
increasing time with statistical tool to increase the probability of increased profits.
study demonstrated that utilizing existing data analysis tools to this new management
data source provides useful knowledge that could be applied to help guide in project
management. In this study we compared results of various sized projects and the use
of formal tools. In our case study we found determining the estimate of the economical
value to be important to guide the degree of use of formal tools. Based on the results
engineering input is best. As projects expand more statistical data improves outcome.
The simple model also tends to show a strong benefit to training. This model
has good VIF values and supports the findings from the SPC findings. Of interest is
results. For example it is not know if people worked harder on projects they initiated
95
However, replication of these results for other projects at other companies and
extensions of this study would be to expand the scope of the analysis to include other
and analysis method, including Six Sigma related methods. For example, Bisgaard and
investment for any process improvement method for a particular context. Yu and
support model for Six Sigma quality improvement efforts, taking into consideration
yield and cost. Chan and Spedding (2001) expanded on this work. Eid, Moghrabi and
(MDP, Puterman, 1994) and Partially Observed Markov Decision Processes (POMDP,
available actions which can be chosen by the system in each state. The use of MDP
96
models has encompassed a wide range of applications. They have been widely applied
to inventory control problems which represent one of the earliest areas of application.
MDP models have been used to study equipment maintenance and replacement
problems, (Rust, 1987) and (Golabi, 1993) as well as computer, manufacturing and
used in a wide range of contests to gain insight into factors influencing animal
behavior. Examples include models of social and hunting behavior of lions (Clark,
1987: Mangel and Clark, 1988), site selection and number of eggs laid by apple
maggots and medflys (Mangel, 1987), daily vertical migration of sockeye salmon and
plankton (Levy and Clark, 1988: Mangel and Clark, 1988), changed in mobility of
spiders in different habitats (Gallespie and Caraco, 1987), and singing versus garaging
tradeoffs in birds (Huston and McNamara, 1986) as well as Games of Chance (Dubins
and Savage, 1965). Kelly and Kennedy (1993) used MPD models in their study of
The Markov decision processes describe a model for sequential decision making under
uncertainty which takes into account both the outcomes of current decisions and future
decision making opportunities. At each decision epoch, the system state provides the
decision maker with all necessary information for choosing an action from the set of
available actions in that state. As a result of choosing an action in a state, two things
happen; the decision maker receives a reward (or cost), and the system evolves to a
possibly different state at the next decision epoch. Both the rewards and transition
97
probabilities depend on the state and the choice of action. As this process evolves
At each decision epoch, the decision maker chooses an action in the state
occupied by the system at that time. A policy provides the decision maker with a
prescription for choosing this action in any possible future state. A decision rule
specifies the action to be chosen at a particular time. It may depend on the present
state alone or together with all previous states and actions. A policy is a sequence of
decision problem is to choose prior to the first decision epoch a policy to maximize a
The key ingredients of this sequential decision model are the following:
(actions) available. Choosing the best action requires a consideration of more than just
the immediate effects of the action. The immediate effects are often easier to calculate
than the long-term effects. Sometimes actions with poor immediate effects can have
better long-term ramifications. To maximize the total expected value function, the
98
right tradeoffs between the immediate rewards and the future gains are needed to yield
When making a decision, we need to consider how actions will affect the
system. The state is the way the system currently exists and an action will have the
effect of changing the state. The actions are the set of possible alternatives that can be
chosen. The problem is to know which of these actions to take when in a particular
state of the system. When deciding between different actions, we need to consider how
they will affect the current state. The transitions specify how each of the actions
changes the state. Since an action could have different effects, depending upon the
state, we need to specify the action’s effect for each state. The most powerful aspect of
the Markov decision process is that the effects of an action can be probabilistic. We
could specify the effects of doing action ‘a1’ in state ‘s1’ if there is no question about
how ‘a1’ changes the system. However, many decision processes have actions that are
not this simple. Sometimes an action usually results in state ‘s2’ but occasionally it
might result in state ‘s3’. MDPs allow specifying a set of resulting states and the
received from a number of sources including sensory input and memories of previous
inputs. This information tells the learner something about the state of the world.
An agent in some state at time t executes an action and receives a reward from
the environment. This is called a decision process. If the next state is dependent only
99
on the current state and action the decision process is said to obey the Markov
If the set of states and actions are finite, then the problem is called a finite
MDP. We can also distinguish between finite and indefinite horizon problems, where
the task has a natural endpoint, and infinite horizon problems, where the task
continues forever.
o A reward distribution P (r t s t , a t ), s t ∈ ℜ, a t ∈ A.
In the above, t indexes the time step, which ranges over a discrete set of points
in time. The transition probability is denoted by Pij (a), Pr( s t +1 = j s t = i, a t = a). The
{
ri (a) = E r t s t = i, a t = a } (5.11)
The solution to an MDP is called a policy and it simply specifies the best
action to take for each of the states. The goal of solving an MDP is to find a policy
that maximizes the total expected reward received over the course of the task. A policy
tells the learning agent what action to take for each possible state. It is a (possibly
stationary, in which case a different mapping from states to actions can be used at each
100
point in time. Alternatively, it can be stationary, in which case the same mapping is
The expected return for a policy π is defined as the total reward that is
⎧n ⎫
Eπ {R t } = Eπ {r t + γ r t +1 + ... + γ n r t + n } = Eπ ⎨∑ γ k r t + k ⎬ (5.12)
⎩ k =0 ⎭
where t is current time. Notice that the expectation is taken with respect to the
policy π .
By assuming that the problem is Markov, we know that an optimal policy need
only be a function of the current state: no other information is required to act optimally
(Howard 1960).
assuming that a problem is Markov, we can ignore the history of the process, and
thereby prevent an exponential increase in the size of the domain of the policy. The
101
In the case of MDPs, we must find a policy that produces the greatest expected
return. With knowledge of transition probabilities Pij (a) and expected immediate
rewards ri (a), and given a stochastic policy π , we can calculate the expected
⎧n ⎫
V π ( s ) = Eπ ⎨∑ γ k r t + k s t = s ⎬ (5.13)
⎩ k =0 ⎭
Here t denotes the current time. The function V π is called the value function
for policy π . The value function gives the expected return that can be achieved by
selections in which (1) the outcomes are random with probabilities that depend upon
the current state (st) at time t, and action taken, a, and (2) the rewards, rt, depend on
the current decision and results form a sequence of future decisions. With a finite
number of decision periods, e.g., days in a project, the optimal policy can be derived
system state, st. Depending upon how the states are defined; a team in a process
improvement project might not have complete knowledge of the current state of their
project. Partially Observed Markov Decision Processes (POMDP) permit the user the
flexibility to supply only probabilities that the system is in specific states. Intuitively,
102
the fact that project improvers have only partial knowledge of the system state might
type methods.
Yet, while MDP models permit the derivation of globally optimal decision
policies, POMDP methods do not generally lead to guarantees that the derived policies
are optimal. Lovejoy (1991) and White (1991) both survey solution methods for these
problems including work in Smallwood and Sondik (1973) and White and Scherer
(1989). Yacout and Gautreau (2000) used POMDP models to compare three quality
(1997), they had already used POMDP directly as a tool for quality improvement. In
our study we only considered unconstrained, discrete time, finite horizon MDPs.
process but knows a competitor is making $150K/year from running a similar process.
improvement project with a three month time limit. Assume that the resident quality
expert, a “green belt” six sigma practitioner, has already declared that the process state
(st) must be one of those listed in the left-hand column of Table 20. Also, assume that
the possible actions (a) are listed in the top first row of the table.
103
These actions include applying statistical process control (SPC) p charting, a
screening method using fractional factorials), and a relatively expensive ($) DOE
method (e.g., a response surface method application). The quality expert estimates the
annualized cost of applying each action for processes in the various states, including
the cost of measurement equipment, record keeping, and training. If either a control
plan is applied or the production action is taken, then the system improvement ends
action (a)
SPC p Control
State (st) Meet Charting $ DOE $$ DOE Optimize Plan Produce
1. Settings lose
money and
-20 -50 -70 -70 -10 -150 -500
capability
unknown
2. Settings lose
money and -20 -30 -40 -70 -10 -10 -200
capability known
3. Measured but
input-output func. -20 -30 -40 -70 -10 -50 -100
not known
4. Input-outputs
func. known but -20 -30 -40 -70 -10 -30 80
not fully opt.
5. Improved
settings found but -20 -30 -40 -70 -10 250 150
not validated
6. System
-20 -30 -40 -70 -10 0 50
improvement ends
Table 20 Costs and rewards (in $K), rt(st,a), of applying actions (a) in different states
(st).Coefficient estimates
104
Assume further that the quality expert has documented that each of the m = 6
actions will cause systems in each of the q = 6 states to transition to each state (st+1 =
1,…,6). Table 21 shows the probabilities for an example action, which is applying an
methods.
To State (st= i)
1 2 3 4 5 6
1 0.9 0.1 0.0 0.0 0.0 0.0
From 2 0.0 0.2 0.8 0.0 0.0 0.0
State 3 0.0 0.00001 0.1 0.79999 0.1 0.0
(st+1= j) 4 0.0 0.0 0.0 0.5 0.5 0.0
5 0.0 0.0 0.0 0.01 0.99 0.0
6 0.0 0.0 0.0 0.0 0.0 1.0
Table 21 Assumed transition probabilities for applying DOE, pt(st+1= j|st= i,a = $$
DOE).
globally maximize the expected reward at every decision period for every possible
state. This calculation is based on the following recursion for t = 5,4,…,1 and i =
1,…,q:
ERt*(st = i) = max {rt(st,a) + Σj=1,…,q pt(st = j|st = i,a) ERt+1*(st+1= j)} (5.14)
a
where ERt=6*(s6 = i) = r6(s6 = i,a = Produce) for i = 1,…,q. The results are
shown in Table 22. The quality expert inputs the time period and the state and the
table indicates which method is recommended. A second table also gives the expected
105
reward or return on investment shown in Table 23. Note that if the current state
cannot be observed and one has only the probability of being in a certain state,
Lovejoy, 1991a).
StateWeeks 1-2 Weeks 3-4 Weeks 5-6 Weeks 7-8 Weeks 9-10 Weeks 11-12
1 Meet Meet Meet Meet Meet Produce
2 SPC SPC SPC SPC Produce Produce
3 $ DOE $ DOE $$ DOE $ DOE $$ DOE Produce
4 Optimize Optimize Optimize Optimize Produce Produce
5 Control Plan Control Plan Control Plan Control Plan Control Plan Produce
6 Produce Produce Produce Produce Produce Produce x
StateWeeks 1-2 Weeks 3-4 Weeks 5-6 Weeks 7-8 Weeks 9-10 Weeks 11-12
1 65.0 -27.6 -101.5 -179.6 -272.2 -500.0
2 175.0 96.8 -5.3 -75.7 -150.0 -200.0
3 307.8 251.3 183.4 67.7 -1.0 -100.0
4 429.4 379.1 327.2 264.5 130.0 80.0
5 500.0 450.0 400.0 350 300.0 150.0
6 500.0 250.0 200.0 150.0 100.0 50.0 x
Table 23 The expected reward, ERt*(st = i), in $K as a function of period and state.
The results are subjective in the sense that the transition probabilities associated with
each action, pt(st = j| st = i,a), e.g., the numbers in Table 18 were subjectively
for the choice of methods with explicit assumptions. Another possible benefit of using
106
the above method for planning improvement projects is the possibility that the
probabilities can capture knowledge of experts and make this knowledge available to
A more detailed model might be useful as it relates to the data used in the
regression model and based on the component methods definition of Six Sigma.
Component methods can be associated with a phase of Six Sigma as shown in Table
24.
107
Phase Deliverables Component methods
DEFINE- (Finding the • Establish process Surveys
Ys in the model) responsibilities Focus Groups
• Define system Interviews
(inputs, outputs, Process MAP
customers) Pareto Analysis
• Identify customer
requirements
MEASURE – (Finding • Gap between Process mapping
the Ys in the model) customer Pareto analysis
requirements and Process capability
process capability Gage R&R
Measurement system
analysis
ANALYZE (find the • Cause and effect Cause and effect
factors or x’s in the matrix analysis
model) • Prioritized list of all Process mapping
x’s Benchmarking
• Few vital x’s Histograms
• Statistical analysis Cater diagrams
for significance Run charts
Multivariate charts
Box and whisker
Pareto chart
Regression
IMPROVE - (Define • Target setting for x’s Process mapping
Y=(x,z) and move Simulation
toward Max Y) DOE
CONTROL – Maintain • Control plan Risk assessment
improvements to the SPC
system Process capability
Checklists
Documentation
108
A Six Sigma type method is an arrangement of component methods such that
there is at least one selection of the associated activities that is consistent with a
DMAIC ordering of activities. Also, in Six Sigma type methods multiple components
methods can be used in the contest of the same activity and no method needs to be
The Figure 19 below shows an example of Six Sigma type method that the
109
Define:
Hold formal meeting
Measure:
Perform gage R&R
Create SPC charts
Perform
benchmarking
Analyze:
Plan and perform DOE
Fit regression model
Apply FMEA
Improve:
Apply formal optimization
Control:
Develop control plan
Continue SPC charting
110
The optimality of any particular Six Sigma type method in question or Six
Sigma type method for a particular application can be investigated using Markov
improvement projects and the actions as defined in Table 21. The rewards or costs of
these actions are easy to estimate based on the data. The transition probabilities are not
as easy to estimate based on this limited data. One method for this estimate is a
analyses are often unique. The situation in which one is making the decision may
probabilities by repeated sampling. People in the same state might take different action
may be used to compute the probabilities needed to make decision. Because these
111
probabilities cannot be measured by repeated sampling, they are called subjective and
combined with a prior belief to end up with a posterior belief. In short, it’s a way to
update a belief with new data. A positive aspect of the Bayesian approach is that it
distribution, then the analyst may simply assume a particular prior distribution. Most
multinomial estimation is via the use of the Dirichlet distribution. The Dirichlet
inference. According to Friedman and Signer (1999), estimates derived using Dirichlet
priors are consistent, the estimate converges with probability one to the true
beliefs about the parameters. As data is gathered, these beliefs do not play a significant
role anymore. More specifically, if the prior is well-behaved, does not assign 0
probability to feasible parameter values, the approach will converge in the limit to the
112
Then the posterior will have the same form, with parameter βi + Ni:
The property that the posterior distribution follows the same parametric form
as the prior is called conjugacy. The Dirichlet prior is a conjugate family for the
multinomial likelihood. Conjugate families are useful because they can be represented
in closed form, incremental updates to the parameters can be done as data is gathered,
With this background a MDP model for Six Sigma based on the previous
collected data and action definitions could take the form of Figure 20.
113
D ev elo pe d "r ec om m e nd atio ns "
C on fir m e d inp or ta nt fac to rs
C on fir m e d ga ug e c a pa bility
Initia l ba s e line es ta blis h ed
"A c c u ra te " Yr ( x ) m o de ls
E s ta blis h n ew ba s e line
C ha r ter is es ta blis h ed
F o un d "r igh t" p ro jec t
U pd ated S O P s
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
0 0 0 0 0 0 0 0 0 0 0 TF TF EC
1 0 0 0 0 0 0 0 0 0 0 EC EC EC TF
1 1 0 0 0 0 0 0 0 0 0 TF TF TF TF SP
1 1 1 0 0 0 0 0 0 0 0 SP SP SP SP SP
1 1 1 .5 0 0 0 0 0 0 0 SP SP SP SP GR
1 1 1 1 0 0 0 0 0 0 0 GR GR GR GR CE
1 1 1 1 1 0 0 0 0 0 0 CE CE CE CE CE
1 1 1 1 1 .5 0 0 0 0 0 CE CE CE CE SD
1 1 1 1 1 1 0 0 0 0 0 SD SD SD SD SD
1 1 1 1 1 1 .2 0 0 0 0 SD SD SD SD SD
1 1 1 1 1 1 .4 0 0 0 0 SD SD SD SD SD
1 1 1 1 1 1 .6 0 0 0 0 SD SD SD SD SD
1 1 1 1 1 1 .8 0 0 0 0 SD SD SD SD RD
1 1 1 1 1 1 1 0 0 0 0 RD RD RD RD RD
1 1 1 1 1 1 1 .3 0 0 0 RD RD RD RD RD
1 1 1 1 1 1 1 .5 0 0 0 RD RD RD RD RD
1 1 1 1 1 1 1 .8 0 0 0 RD RD RD RD OP
1 1 1 1 1 1 1 1 0 0 0 OP OP OP OP SP
1 1 1 1 1 1 1 1 1 0 0 SP SP SP SP SP
1 1 1 1 1 1 1 1 1 0 0 SP SP SP SP CP
1 1 1 1 1 1 1 1 1 .5 0 CP CP CP CP TE
1 1 1 1 1 1 1 1 1 1 .5 CP CP CP TE
1 1 1 1 1 1 1 1 1 1 1 TE TE TE
The actual actions taken might be along the diagonal as indicated. The amount
of data obtained to date has not allowed for the construction of an adequate model and
114
5.12 Sample MDP Example
The data associated with the 30 improvement projects can be used in the
simplified case six states are defined which map to the standard DMAIC as follows in
Table 25.
Ten actions were possible with average cost calculated form the collected data.
These action-rewards are presented in Table 26 with the tabulation of the number of
115
Action State to State 1 2 3 4 5 6
Charter 1 12 8 6 2 - -
Charter 2 - - - - - -
Charter 3 - - - - - -
Charter 4 - - - - - -
Charter 5 - - - - - -
Charter 6 - - - - - -
Team Forming 1 1 11
Team Forming 2 - - - - - -
Team Forming 3 - - - - - -
Team Forming 4 - - - - - -
Team Forming 5 - - - - - -
Team Forming 6 - - - - - -
Process Mapping 1 - - 3 3 - -
Process Mapping 2 - 3 5 2 5 -
Process Mapping 3 - - - - - -
Process Mapping 4 - - - - - -
Process Mapping 5 - - - - - -
Process Mapping 6 - - - - - -
GR&R 1 - - 1 - - -
GR&R 2 - 1 5 - - -
GR&R 3 - - - - - -
GR&R 4 - - - - - -
GR&R 5 - - - - - -
GR&R 6 - - - - - -
116
Tabulation of State Transitions (continued from previous page)
SPC 1 - - - 1 - -
SPC 2 - - 3 - - -
SPC 3 - - - - - -
SPC 4 - - - - - -
SPC 5 - - - - - -
SPC 6 - - - - - -
- - - - - -
Documentation 1 - - - - - -
Documentation 2 - - - - - -
Documentation 3 - - - - - -
Documentation 4 - - - - - -
Documentation 5 - - - - 3 26
Documentation 6 - - - - - -
Engineering Analysis 1 - - - 1 - 3
Engineering Analysis 2 - - - - - -
Engineering Analysis 3 - - - 15 1 -
Engineering Analysis 4 - - - 2 10 4
Engineering Analysis 5 - - - - - -
Engineering Analysis 6 - - - - - -
117
An examination of Table 27 illustrates the limited data available from the
improvement projects and the need for Bayesian estimation. Table 28 gives an
example of the tabulated actual data and Alpha priors which when added together
results in a usable transition matrix for the action Engineering Analysis. Continuing
with the remaining actions, (estimating Alpha priors and adding to the tabulated data)
results in a matrix that can be used to solve for a MDP policy. One possible Bayesian
distribution based on the data from the 39 projects and Alpha priors as described
Alpha Priors
No Charter 1 67 3 5 9 5 7
No Base line 2 7 80 6 4 2 1
Define input-output 3 2 3 50 19 6 4
No Recommendation 4 2 2 4 10 50 16
Not Controlled 5 2 5 20 30 40 3
Production 6 0 0 0 0 0 10
118
Action State to State 1 2 3 4 5 6
Charter 1 .01 .61 .189 .071 .07 .05
Charter 2 .3 .5 .1 .05 .03 .02
Charter 3 .2 .3 .4 .05 .03 .02
Charter 4 .02 .2 .3 .4 .05 .03
Charter 5 .02 .02 .2 .3 .41 .05
Charter 6 0 0 0 0 0 1
Process Mapping 1 .6 .2 .2 .1 0 0
Process Mapping 2 .07 .16 .3 .13 .3 .04
Process Mapping 3 .1 .1 .5 .2 .05 .05
Process Mapping 4 .05 .1 .1 .5 .2 .05
Process Mapping 5 .05 .05 .1 .1 .5 .2
Process Mapping 6 0 0 0 0 0 1
119
State Transitions (continued from previous page)
120
The above Bayesian probabilities and the rewards given in Table 26 where
used in the development of an optimal MDP policy for the six listed states. It was also
assumed that the production state had two different rewards. This was based on real
data gathered from the improvement project. The average reward for a project that
terminated early was $4,700. The average reward for a project that entered production
at epoch 9 was $180,875. The optimal policy calculated with these assumptions is
This optimal policy has the structure on the standard DMAIC but does not
follow the five phases explicitly. This could arise for numerous reasons. The Alphas
selected may not be reasonable. The validity of the selection would become more
apparent with added real data which would allow an update to the Bayesian estimates.
121
DMAIC may not be optimal in all cases. Again more data is needed to investigate this
possibility. Both Linderman et. al. (2003) and Harry and Crawford (2005) have
indicated that DMAIC may not be optimal for small project, the “low hanging fruit”
and that DMAIC should be focused on complex challenging problems. For simple
tasks such a system may decrease performance. The 39 improvement projects included
in this study included a wide array of programs with estimated savings ranging from a
low of $1,600 to a high of $2.2 million. It is possible that the data should be split into
simple and complex projects. This is an area for future work, relating the size of a
5.13 Conclusions
Much of the literature related to six sigma focuses on knowledge that
participants at various levels “should know” (Hoerl 2001). Other literature has
Hild 2000, and Hoerl 2001). Goh (2002) has discussed some strategic perspectives of
Six Sigma based on organization needs. The literature on supporting method selection
section in the context of Six Sigma has focused on decisions at a high level such as
whether to perform projects at all including Bisgaard and Freiesleben (2000), Chan
and Spedding (2001), Gautreau, Yacout and Hall (1997), Yacout and Gautreau (2000),
and Yu and Popplewell (1994). In this study some methods for the evaluation of the
expanding Six Sigma database are investigated. These approaches help to express Six
122
Sigma in more quantitative terms, which has more often been expressed in qualitative
terms.
Clearly the list of methods considered here for meso-analysis is incomplete and
123
Tool Strength Limitation
Regression Simple and many diagnostics Necessary sample size can become
available large as
n = 4 × [1 + m pre + m act + (m pre × m act )]
MPDs Can address sequencing effects There is strong pressure to define
of method applications and only a small number of states for
interactions between them. simplicity.
124
To our knowledge no research has focused on the use of this database in this
help industry to identify, analyze and evaluate factors and their interdependency,
which would help to understand and unveil the complexity of Six Sigma. This could
open the door to research that more clearly clarifies and capitalizes on the primary
value of the Six Sigma movement which lies in providing a relatively detailed
algorithms that can be used to generate software for green belts, the research may lead
to substantially more valuable methods than those of the Six Sigma type.
125
CHAPTER 6
A summary and conclusion of this study is presented in this chapter. The first
section provides an overview of the research while the second section summarizes the
major research findings. The third section addresses the limitations of the study and
potential areas for future research. Finally, the contributions of this study are presented
6.1 Overview
The main objective of this dissertation was to review the published literature
on Six Sigma and identify opportunities for additional contributions for the academy.
1. With reference to past academic contributions, what is Six Sigma? Also, what
2. What are the implications of Six Sigma philosophy and methods for university
education?
126
3. What methods should be used to mine the new databases about project
financial results? Also, what insights can be gained form studying data at a
real company?
with Six Sigma as follows: For the first question a literature review covering a
fourteen year timeframe was undertaken to describe the trends, sources and findings in
the publications on Six Sigma. Secondly, case base training was examined as a method
to improve Six Sigma education and increase usage on the job among university
quality improvement projects. The study explored methods to utilize this new data
quality policies.
between 1990 and 2003. Although this review cannot claim to be exhaustive, it does
provide reasonable insights into the state-of-the-art. As the nature of research on Six
across various journals. It is felt that the results have several important implications.
127
To a great extent, the financial impact of Six Sigma on operational
Sigma is that such justifications are needed for each Six Sigma project. The impacts
associated with the decisions to start Six Sigma programs. Yet, those authors found no
This suggests a need for additional data collection and analysis to answer the
It is proposed from the study that Six Sigma can be defined as a component
based method involving Define, Measure, Analyze, Improve and Control (DMAIC) or
(DMADV) and two principles. These principles relate both to building and
practitioners who are not experts in statistics. The first principle emphasizes attention
to the bottom line in initiating projects. The second principle emphasized the training
embodied in the literature: (1) increased emphasis on complete case studies compared
with single sub-method applications, (2) new, relatively specific core and
infrastructure practices, and (3) the development of a large new market of industrial
non-experts who might be interested in practically oriented research and new methods.
128
While over 50% of the articles in the database either explicitly or implicitly
implementing Six Sigma in specific business contexts has, apparently, not been
investigated. Related, largely unanswered questions include: How can data about any
Focusing on the second principle covered in the definition of Six Sigma which
graduate level engineering courses on statistical process control (SPC) and design of
experiments (DOE). Students seem to identify with the case study "stories" much
better than the traditional lectures. Before implementation of the case study exercises,
end-of quarter evaluations by students for the two classes had been below the average
case study approach and with minimal other changes, the evaluations climbed well
above average. This dramatic turn-around can mainly be attributed to the introduction
of the case based method. Further, presenting applications of course methods in which
concerned about job security and advancement in the highly competitive industrial
marketplace.
One outcome of the wide spread acceptance of Six Sigma and with its
systematic program is the growing database that now exist within industry and specific
129
companies on individual project improvement activities. The purpose of this study was
to look at the database in a way that could help management better run improvement
projects. Three possible analysis methods investigated for this task were regression,
SPC, and Markov Decision Processes (MDP). The study was based on 39 quality/cost
collected over a 30 month period. It was found that by viewing the quality
improvement process in the same light as other processes and applying such
techniques to the accumulating database could provide better insight for operational
each project. Yet, the literature review found no published work showing statistical
significance in relation to the decisions to start Six Sigma programs and stock
performance. Considering that failure to find a significant effect does not constitute
proof, more work is needed for a thorough evaluation of the bottom line impacts of
Six Sigma.
objective evidence. Examples include self reported profits, the effects of success
factors, and advocacy for Six Sigma in general. For example, as noted above, the
impacts on stock performance investigated by Goh et al. (2003) are not fully resolved.
130
This suggests a need for additional data collection and analysis to answer the
important question of long term impacts of decisions to adopt Six Sigma programs.
Snee (1999 and 2000a) calls for research to help practitioners identify a robust
set of improvement tools to be used in conjunction with the DMAIC process. The
relevant to Six Sigma practitioners who are often not experts in statistics.
impacts of projects are needed. This follows because of the central importance of
profit related justifications in Six Sigma for initiating decisions on projects. Our
research did indicate a correlation with training and the bottom-line impact of projects.
New models that are also easy-to-use could be developed with broader applicability
models to aid practitioners from different disciplines select the most advantageous
techniques. This could build on research related to the most appropriate methods for
training black belts, e.g., in Hoerl (2001a), by associating the methods more
and many companies have gained the profits and advantages from this disciplined
131
approach, the research of the impacts of Six Sigma implantation and factors
contributed to Six Sigma success remain unclear. Even the existent studies are not
well integrated and the research is mostly anecdotal. Current concepts in the field of
Six Sigma are largely based upon case studies, anecdotal evidence and the
factors are critical to the success of the approach. Most of the articles reported that top
management leadership is the main factor to Six Sigma success [Blakeslee (1999) and
Scalise (2001)]. However, many other factors affecting Six Sigma’s success are
frame. This is limited data, collected at one site and no definitive conclusions should
be drawn. The data does provide a start into the use and modeling of Six Sigma,
and the methods used that is being generated as a result of it implementation. The
attempt to build a theory of how and why Six Sigma works is aimed at building a
prescriptive model. From this, managers would be able to identify which activities
from which programs are more or less likely to be useful in their situations, as well as
which of their goals would be most affected. An example is the correlation between
training and increased profit seen in this study. With the future success of corporations
riding on the outcome, there is a need for more theory to explain the differences
132
APPENDIX A: THE ARTICLE DATABASE
The descriptors used in the table below are defined in Section 3. The complete
Management committed
Speculative in Nature?
Change Management
Structured approach
Research Approach
Team involvement
Customer focused
Project leadership
Adaptable system
Project selection
Industrial Sector
Success Factors
Define DMAIC
Impact factor
Data system
Goals based
Bottom line
Authorship
Right team
Define 3.4
Training
Year
Author(s)
Pe
Abraham et.al. 2001 A 0 0 To Ph To G 1.5 Co N N - - - - - - - - - - - - -
Pe
Ackermann 1993 I 0 0 To Ph To G 0.7 C N N - - - - - - - - - - - - -
Ackermann et.al. 1993 I 0 1 To To M 0.7 C N N - - - - - - - - - - - - -
Ali et.al. 1999 I 0 0 To To M 1.5 C N N - - - - - - - - - - - - -
Antony et.al. 2002 I 0 1 Sy Ph M 0.3 C Y N - - - - - - - - - - - - -
Arvidsson 2003 A 0 0 Sy Ph M 0.3 Su N N - - - - - - - - - - - - -
Pe
Bailey 2001 I 0 0 Sy Ph M 1.5 C N N - - - - - - - - - - - - -
Bartos 1999 I 0 0 Sy Ph M 0.3 TA Y Y - Y - - - - - - - - - - -
Basu 2001 I 0 1 Sy Ph M 0.2 Co N N - - - - - - - - - - - - -
N
Behara et.al. 1995 I A 0 1 Sy Ph M A C N N - - - - - - - - - - - - -
133
N
Benedetto 2003 I 1 0 Sy Ph Se A C N Y Y - Y Y - - - - - - - - -
Berlowitz 2003 A 0 1 To To Se 2.9 C R N Y - - - - Y - - - - - - - -
Binder 1997 I 1 1 Sy Ph M 0.8 Co N N - - - - - - - - - - - - -
N
Bisgaard et.al. 2000 A 0 0 Sy Ph M A TA Y N - - - - - - - - - - - - -
Blakeslee 1999 I 0 0 Sy Ph Se 0.2 C Y Y Y - Y - Y Y - Y - - - - -
N
Blanton 2002 A 0 0 Sy Ph Se A C Y N - - - - - - - - - - - - -
Bossert 2003 I 0 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Breyfogle 2002 I 0 0 Sy Ph Se 0.2 C N N - - - - - - - - - - - - -
Breyfogle et.al. 2003 I 0 0 Sy Ph M 0.7 TA Y N - - - - - - - - - - - - -
N
Breyfogle et.al. 2002 I 1 1 Sy Ph Se A TA Y N - - - - - - - - - - - - -
Breyfogle et.al. Pe
(a) 2001 I 0 0 Sy Ph G 1.5 TA Y N - - - - - - - - - - - - -
Breyfogle et.al.
(b) 2001 I 0 0 To To G 0.2 C Y N - - - - - - - - - - - - -
Broderick et.al. 2002 A 0 0 Sy Ph Se 4.8 C N N - - - - - - - - - - - - -
N
Buck et.al. 2001 I 1 0 Sy Ph Se A C N Y Y - - - - - - - - - - - -
C Su
Buck 1998 I 1 1 Sy Ph Se 1.9 R N N - - - - - - - - - - - - -
N
Buck 2001 I 1 1 Sy Ph Se A C N N - - - - - - - - - - - - -
Buggie 2000 I 0 0 Sy Ph M 0.1 TA Y N - - - - - - - - - - - - -
Card 2000 I 0 0 To To M 0.8 Co N N - - - - - - - - - - - - -
Sy
Caulcutt 2001 I 1 1 To Ph Pr M 0.3 C N Y - - - - Y - - - - - - - -
Chan et.al. 2001 A 0 1 To To M 0.4 C Y N - - - - - - - - - - - - -
Chassin 1998 A 0 1 Sy Ph Se 1.9 R N N - - - - - - - - - - - - -
Chowdhury 2000 I 1 0 Sy Ph G 0.3 TA Y Y - - Y - - - - - - - - - -
N
Clifford 2001 I 0 1 Sy Ph M A C CoN Y Y - - - - - - - - - - - -
Coleman et.al. 2001 I A 0 0 To To M 0.3 TA Y N - - - - - - - - - - - - -
Connolly 2003 I 0 0 Sy Ph M 0.7 C N N - - - - - - - - - - - - -
Conner 2003 I 0 0 Sy Ph M 0.3 C N N - - - - - - - - - - - - -
Cooper 1992 I 0 1 Sy Ph M 0.1 C N N - - - - - - - - - - - - -
Pe
Cooper 2003 I 0 0 Sy Ph Se 0.2 Su N Y - - - - - Y - - - - - - -
Pe
Crom 2000 I 0 0 Sy Ph G 0.2 Co Y Y Y - - - - - - - - - - - -
Pe N
Dasgupta 2003 A 1 1 Sy Ph Se A C Y Y Y - - - - - - - - - - - -
Davies 2001 A 0 0 Sy Ph Se 0.8 R N N - - - - - - - - - - - - -
Davig et.al. 2003 A 0 0 Sy Ph M 0.2 Su N Y Y - - - - - - - - - - - -
De Mast 2003 A 0 0 Sy Ph G 0.2 Co Y N - - - - - - - - - - - - -
De Mast et.al. 2000 A 1 0 Sy Ph G 0.2 Co N N - - - - - - - - - - - - -
134
N
Dedhia 1995 I 0 0 Sy Ph Se A TA Y Y Y - Y - - - - - - - - - -
Pe N
DeFeo 2000 I 0 1 Sy Ph G A Su N N - - - - - - - - - - - - -
Deshpande 1998 A 0 0 To To M 0.2 C Y N - - - - - - - - - - - - -
Deshpande et.al. 1999 I A 1 1 Sy Ph G 0.4 C N N - - - - - - - - - - - - -
N
Does et.al. 2002 A 1 0 Sy Ph Se A Co N N - - - - - - - - - - - - -
Doganaksoy et.al. 2000 I A 0 0 To To M 0.2 C Y N - - - - - - - - - - - - -
Dornheim 2001 I 0 0 To To M 0.3 TA Y N - - - - - - - - - - - - -
N
Douglas 2000 I 0 0 Sy Ph G A C Su N N - - - - - - - - - - - - -
Du et.al. 2000 A 0 1 To Pr To M 0.5 Co Y N - - - - - - - - - - - - -
Duguesaoy et.al. 2002 I 0 0 To To G 0.4 C N N - - - - - - - - - - - - -
Eid et.al. 1997 A 0 0 To To M 0.4 TA Y N - - - - - - - - - - - - -
Farntz 2001 I 1 0 Sy Ph M 0.3 C N N - - - - - - - - - - - - -
Feng et.al. 1997 A 0 0 To To M 0.4 C N N - - - - - - - - - - - - -
N
Ferrin et.al. 2002 I 1 1 To To M A Su N N - - - - - - - - - - - - -
Finn 1999 I 1 0 Sy Ph M 0.3 TA Y N - - - - - - - - - - - - -
Fontenot et.al. 1994 I A 0 1 To To M 0.2 Su N N - - - - - - - - - - - - -
N
Fuller 2000 I 0 0 Sy Ph M A TA Y Y - - - - - - - - - - Y - -
N
Gano 2001 A 0 0 To To G A C N N - - - - - - - - - - - - -
Gautreau et.al. 1997 A 0 0 To To M 0.4 TA Y N - - - - - - - - - - - - -
N
Gill 1990 I 0 1 Sy Ph M A Su N N - - - - - - - - - - - - -
Gnibus 2000 I 0 0 To To G 0.2 C Y N - - - - - - - - - - - - -
N
Goh (a) 2002 A 1 1 Sy Ph G A Co N N - - - - - - - - - - - - -
Pe
Goh (b) 2002 A 1 0 Sy Ph G 0.2 R N Y - - Y - - - Y - - - - - -
Goh 2001 A 0 0 To To M 0.3 TA Y Y Y - - - - - - - - - - - -
N
Goh et.al. (a) 2003 A 1 0 Sy Ph M A R N N - - - - - - - - - - - - -
N
Goh et.al. (b) 2003 A 1 1 To To G A TA Y N - - - - - - - - - - - - -
Gordon 2002 I 0 0 Sy Ph G 0.2 Co Y N - - - - - - - - - - - - -
N
Grandzol et.al. 1998 A 0 0 To Pr M A Su N N - - - - - - - - - - - - -
Greek 2000 I 0 1 Sy Ph M 0.3 Su N Y Y - - - - - - - - - - - -
Gross 2001 I 0 0 Sy Ph G 0.2 TA Y Y Y - - - - - - - - - - - -
Pe
Hahn et.al. 1998 I 0 0 Sy Ph M 0.2 TA N N - - - - - - - - - - - - -
Pe
Hahn et.al. 1999 I 1 1 Sy Ph M 1.2 TA N Y - - - - - - Y - - - - - -
Pe
Hahn 2002 I 0 0 Sy Ph G 1.2 TA Y N - - - - - - - - - - - - -
135
N
Hahn et.al. 2000 I 1 1 Sy Ph G A TA Y Y Y - - - Y - Y - - - Y - -
Hahn et.al. 2001 I 0 0 To To G 0.2 TA Y Y - - Y - - - - - - - - - -
N
Hammer 2002 I 1 0 Sy Ph M A TA N Y - - - - - - Y - - - - - -
Pe
Harrold 1999 I 1 1 Sy Ph M 0.3 C N Y - - - - - Y - - - - - - -
Harrold et.al. 1999 I 1 0 Sy Ph M 0.3 TA Y N - - - - - - - - - - - - -
Pe
Harry 1998 I 1 1 Sy Ph M 0.2 C N Y - - Y - - - - - - - - - -
Sy
Harry (a) 2000 I 0 0 To Ph To G 0.2 Su N Y - - - - Y - - Y - - - - -
Harry (b) 2000 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Harry (c) 2000 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Harry (d) 2000 I 0 1 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Harry (e) 2000 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Harry (f) 2000 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Henretta et.al. 2003 I 1 1 Sy Ph Se0.2 C N N - - - - - - - - - - - - -
N
Hild et.al. 2000 I 0 0 Sy Ph M A Co N N - - - - - - - - - - - - -
Pe
Hill 2001 I 0 0 To Ph To M 1.5 C N N - - - - - - - - - - - - -
Hoerl 1998 I 0 0 Sy Ph G 0.2 C N Y - Y - - - - - - - - - - -
Pe
Hoerl (a) 2001 I 0 0 Sy Ph G 1.5 TA N Y - - - - - - - - Y - - - -
Pe
Hoerl (b) 2001 I 0 0 Sy Ph G 1.5 TA Y N - - - - - - - - - - - - -
Horst 1999 I 0 0 Sy Ph Se 0.3 C CoN N - - - - - - - - - - - - -
Howell 2000 I 0 0 Sy Ph M 0.3 Su N N - - - - - - - - - - - - -
Howell 2001 I 0 0 Sy Ph M 0.3 Su N Y Y - - - - - - - - - - - -
Hunter 1999 I 0 0 Sy Ph Se 0.3 C N N - - - - - - - - - - - - -
Hunter 2000 I 0 0 Sy Ph G 0.3 TA Y N - - - - - - - - - - - - -
Hunter et.al. 1999 I 0 0 Sy Ph M 0.3 C N Y Y - - - - - - - - - - - -
Pe
Hutchins 2000 I 0 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Pe N
Ingle et.al. 2001 I A 1 1 Sy Ph M A Co N N - - - - - - - - - - - - -
Pe
Johnson 2002 I 0 1 Sy Ph G 0.3 TA Y N - - - - - - - - - - - - -
Johnson et.al. 2003 I 1 1 Sy Ph M 0.3 Su N N - - - - - - - - - - - - -
Johnstone et.al. 2002 I 0 1 Sy Ph Se 2.9 C N N - - - - - - - - - - - - -
Sy
Johnstone et.al. 2003 I 0 1 To Ph Pr Se 1.2 C N Y Y - - - - - - - - - - - -
Kandebo 1999 I 0 0 Sy Ph M 0.3 TA N N - - - - - - - - - - - - -
Kane 1998 I 0 1 Sy Ph M 0.2 C N N - - - - - - - - - - - - -
N
Kazmer et.al. 2002 A 0 1 To To M A C Y N - - - - - - - - - - - - -
Kazmierczak 2003 A 0 1 To To Se 1.6 R N N - - - - - - - - - - - - -
136
Kendall et.al. 2000 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Kenett et.al. 2003 I A 0 0 To To M 0.2 C N N - - - - - - - - - - - - -
Knowles et.al. 2003 A 0 0 To To Se 0.2 C N N - - - - - - - - - - - - -
Koch 2002 I 0 1 To To M 0.8 C N N - - - - - - - - - - - - -
Koonce et al. 2003 A 0 0 To To M 0.4 C Y N - - - - - - - - - - - - -
Krouwer 2002 I 0 1 To To Se 0.8 C N N - - - - - - - - - - - - -
Kunes 2002 I 1 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Landin et.al. 2001 A 0 0 Sy Ph G 0.3 Su N Y - - - - - - - - - - Y - -
Sy
Leffew et.al. 2001 I A 1 1 To Ph To M 0.4 C N N - - - - - - - - - - - - -
Linderman et.al. 2003 A 1 1 Sy Ph G 1.5 TA Y Y - - - - - - - - - Y - - -
Sy
Lucas (a) 2002 I 1 1 To Pr To G 0.2 TA Y Y Y - - - - - - - - - - - -
Lucas (b) 2002 I 0 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Mader 2002 I 0 0 Sy Ph M 0.2 TA Y N - - - - - - - - - - - - -
Maguire (a) 1999 I 0 1 To To M 0.2 TA N N - - - - - - - - - - - - -
Magure (b) 1999 I 0 1 Sy Ph M 0.2 C N Y - - - - - - - - - - Y - -
Pe N
Mandal et.al. 1998 A 0 0 To Ph Pr G A Su R N Y Y Y Y - - - - - - - Y - -
Mason et.al. 2000 I 1 0 To To G 0.2 C N Y - - - - - - - - - - - - -
N
McCarthy et.al. 2001 I 1 1 To To M A C N N - - - - - - - - - - - - -
McFadde 1993 A 1 1 Sy Ph M 0.2 TA Y N - - - - - - - - - - - - -
Pe
Montgomery 2000 A 0 0 To Ph Pr G 0.2 TA Y N - - - - - - - - - - - - -
Montgomery 2001 A 0 0 Sy Ph G 0.2 TA Y Y Y - - - - - - - - Y - - -
Pe
Montgomery 2002 A 0 0 To Ph Pr G 0.2 TA Y N - - - - - - - - - - - - -
Montgomery et Pe
al. 2001 I A 0 0 Sy Ph G 1.5 TA Y N - - - - - - - - - - - - -
Mukesh 2003 I 1 1 Sy Ph M 0.4 C N N - - - - - - - - - - - - -
Munro 2000 I 0 1 Sy Ph M 0.2 Co N N - - - - - - - - - - - - -
Murugappan Sy
et.al. 2003 I 0 0 To Ph Pr Se 0.8 C N N - - - - - - - - - - - - -
Nave 2002 I 1 0 Sy Ph G 0.2 Co Y N - - - - - - - - - - - - -
Neuscheler et.al. 2001 I 1 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
Nevalainen et.al.
(a) 2000 I 0 0 Sy Ph Se 1.3 TA Y N - - - - - - - - - - - - -
Nevalainen et.al.
(b) 2000 I 1 1 Sy Ph Se1.3 Co N N - - - - - - - - - - - - -
Nielsen et.al. 1999 I 0 0 Sy Ph Se0.3 C N N - - - - - - - - - - - - -
Noble 2001 I 0 1 Sy Ph M 0.4 Su N N - - - - - - - - - - - - -
Olexa 2003 I 0 0 To Pr M 0.3 C N Y - - - - Y - - Y - - - - -
Pearson 2001 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
N
Plotkin et.al. 1999 I 0 1 Sy Ph M A C N N - - - - - - - - - - - - -
N
Pyzdek (a) 2001 I 0 0 Sy Ph M A Co N N - - - - - - - - - - - - -
137
Pe
Pyzdek (b) 2001 I 0 0 To Ph To G 1.5 TA Y N - - - - - - - - - - - - -
N
Ramberg 2000 A 0 1 Sy Ph G A TA Y Y - - - - - - - - Y - - - -
N
Rasis et.al. (a) 2003 I A 1 1 Sy Ph M A C N N - - - - - - - - - - - - -
N
Rasis et.al. (b) 2003 I A 1 0 Sy Ph M A C N N - - - - - - - - - - - - -
N
Rayner 1990 I 0 0 Sy Ph G A C N N - - - - - - - - - - - - -
Ribardo et.al. 2003 I A 0 0 To To M 0.2 C N N - - - - - - - - - - - - -
N
Riley et al. 2002 A 1 0 Sy Ph Se A C N N - - - - - - - - - - - - -
Rowlands et.al. 2003 A 0 0 To To M 0.6 C N N - - - - - - - - - - - - -
N
Sanders et.al. (a) 2000 I 0 0 Sy Ph G A TA Y Y Y - Y - - - - Y - - - Y -
N
Sanders et.al. (b) 2000 I 0 0 Sy Ph Se A C N N - - - - - - - - - - - - -
N
Sanders et.al. 2001 I 0 0 Sy Ph G A TA Y N - - - - - - - - - - - - -
Sarewitz 2000 I 0 0 Sy Ph Se 1.3 TA Y N - - - - - - - - - - - - -
N
Scalise 2001 I 0 0 Sy Ph M A Su N N - - - - - - - - - - - - -
N
Scalise 2003 I 0 0 Sy Ph Se A C N N - - - - - - - - - - - - -
Schmitt 2000 I 0 1 Sy Ph M 0.3 Su N N - - - - - - - - - - - - -
Schmitt 2001 I 0 1 Sy Ph M 0.3 Su N N - - - - - - - - - - - - -
Schmitt 2002 I 0 1 Sy Ph G 0.3 Su N N - - - - - - - - - - - - -
Sigal et al. 2001 A 1 0 Sy Ph Se 4.8 C N N - - - - - - - - - - - - -
Smith 2003 I 0 0 Sy Ph M 0.2 C N Y Y - - - - - - - - - - - -
Pe
Snee 1999 I 1 1 Sy Ph G 0.2 TA Y Y - - - - - Y - Y - - Y - -
Pe N
Snee (a) 2000 I 0 1 Sy Ph G A R Y N - - - - - - - - - - - - -
Pe
Snee (b) 2000 I 0 0 To Ph To G 0.2 TA Y Y - - Y - - - - - - - - - -
Snee (a) 2001 I 0 0 To To G 0.2 TA Y Y Y - - - - - Y - - - - - -
Pe
Snee (b) 2001 I 0 0 To Ph To G 1.5 TA Y N - - - - - - - - - - - - -
Snee 2003 I 1 0 Sy Ph G 0.2 TA Y N - - - - - - - - - - - - -
N
Stamatis 2000 I 0 0 Sy Ph M A C CoY N - - - - - - - - - - - - -
Stein 2001 I 0 0 To To G 0.2 TA Y N - - - - - - - - - - - - -
Studt 2002 I 1 0 Sy Ph G 0.7 TA Y Y - Y - - - - - - - - - - -
Takikamalla 1994 A 0 1 To To M 0.2 TA Y N - - - - - - - - - - - - -
Tang et.al. 1997 A 0 0 To To M 0.2 Co Y N - - - - - - - - - - - - -
Treichler et.al. 2002 I 0 0 To Pr G 0.2 C N Y Y Y Y - Y Y - - - - Y Y -
Trivedi 2002 I 0 0 Sy Ph M 0.4 C N Y - - Y - - - - - - - - - -
Tylutki et.al. 2002 A 0 0 Sy Ph Se 0.2 C N N - - - - - - - - - - - - -
138
Vandenbrande 1998 I 0 0 To To M 0.2 TA Y N - - - - - - - - - - - - -
N
Vaugham 1998 A 0 1 To To G A C N N - - - - - - - - - - - - -
Velocci (a) 1998 I 0 1 Sy Ph M 0.3 Co N N - - - - - - - - - - - - -
Velocci (b) 1998 I 0 0 Sy Ph M 0.3 C N N - - - - - - - - - - - - -
Velocci (c) 1998 I 0 0 Sy Ph M 0.3 C N N - - - - - - - - - - - - -
Velocci 2000 I 0 1 Sy Ph M 0.3 C N N - - - - - - - - - - - - -
Velocci 2002 I 0 0 Sy Ph M 0.3 Su N Y - Y - - Y - - Y - - - Y Y
Voelkel 2002 A 0 0 To To G 0.2 C N Y - - Y - - - - - - - - - -
Walsh et.al. 2000 I 0 0 Sy Ph G 0.3 Su N N - - - - - - - - - - - - -
Pe
Watson 2000 I 0 0 Sy Ph G 0.2 Co N N - - - - - - - - - - - - -
Watson (a) 2002 I 0 0 To To Se 0.2 C CoN N - - - - - - - - - - - - -
Watson (b) 2002 I 1 0 Sy Ph Se 0.5 TA Y N - - - - - - - - - - - - -
Waurayniak 2002 I 0 0 To To M 0.3 Su N Y - - - - Y - - - - - - - -
Pe
Weinstein et.al. 1998 A 0 0 Sy Ph G 0.2 Su N N - - - - - - - - - - - - -
Westgard 2002 I 0 0 To Pr To Se 4.4 TA N N - - - - - - - - - - - - -
Wheeler 2002 I 0 0 Sy Ph M 0.4 Co N Y - - - - Y - - - - - - - -
N
Wiklund et.al. 2002 A 0 0 Sy Ph Se A C N Y Y - Y - Y - - - - - Y - -
Wood 2001 I 0 0 Sy Ph G 0.3 C N Y Y Y - - - - - - - - - - -
Pe N
Wyper et.al. 2000 I 1 1 Sy Ph Se A C N N - - - - - - - - - - - - -
Yeung et.al. 2003 A 0 0 Sy Ph M 1.5 Su N Y Y - - - - - - - - - - - -
Yu et.al. 1994 A 0 0 Sy Ph M 0.4 R N N - - - - - - - - - - - - -
139
APPENDIX B: PROJECT DATABASE
Tabulated data from the 39 quality improvement and cost reduction projects
140
Exp Exp
Project Savings Time M/I A/P #people EC CH TF PM CE GR
1 $35000 L M A 7 0 1 1 2 1 0
2 $70000 L M A 1 1 1 0 0 0 0
3 $81315 M M A 2 1 1 1 1 0 0
4 $40000 M M A 1 0 0 0 1 0 0
5 $250000 L I P 6 1 1 1 0 2 2
6 $150000 L M P 4 0 1 1 1 0 0
7 $125000 L I P 3 0 1 1 0 0 1
8 $2200000 L M P 9 0 1 0 0 3 0
9 $50000 M M P 5 1 1 1 1 1 1
10 $39195 M M P 1 1 1 0 0 0 0
11 $34500 L M A 1 1 0 0 1 1 1
12 $21000 L M A 1 0 1 0 0 0 0
13 $25000 M M A 1 0 0 0 1 0 0
14 $20000 M M A 1 0 0 0 1 0 0
15 $10000 M M A 1 0 1 0 0 0 0
16 $20000 S M A 1 0 0 0 0 0 0
17 $28000 M I P 1 0 0 0 1 0 0
18 $20000 S M P 5 0 1 1 0 2 0
19 $20000 S M P 1 0 0 0 1 0 0
20 $4350 S M A 1 0 1 0 1 0 0
21 $13750 S M A 1 0 1 0 1 0 0
22 $8500 S M A 1 0 1 0 1 0 0
23 $1600 S M A 1 1 0 0 0 0 0
24 $12500 S M A 1 0 1 0 1 0 0
25 $4000 S M A 1 0 0 0 0 0 0
26 $13000 S M A 1 0 0 0 0 0 0
27 $15000 L I P 1 1 1 0 0 0 0
28 $6000 M I P 1 1 1 0 1 0 0
29 $11500 M I P 2 0 1 1 1 0 0
30 $4500 M I P 1 1 1 0 1 0 0
31 $11000 S M P 5 0 1 1 0 1 0
32 $5400 S M P 5 0 1 1 1 1 0
33 $150000 S I P 4 0 1 0 1 1 1
34 $8600 S I P 2 1 1 0 0 0 0
35 $90000 M M A 5 1 1 1 1 1 1
36 $30000 M M P 7 1 1 1 0 1 0
37 $45000 S M A 3 0 1 0 0 0 1
38 $240000 S I P 3 1 0 0 0 0 0
39 $50000 S I P 4 1 1 0 1 0 0
141
Project DOE SPC DC FT EA OF Time Cost Act Savings Profit
1 0 0 1 2 0 1 13 $48700 $36000 $-12700
2 1 0 0 1 1 1 18 $7590 $0 $-7590
3 0 0 1 1 1 0 25 $35300 $31500 $-3800
4 0 0 0 0 1 0 20 $2900 $0 $-2900
5 2 0 1 7 0 1 16 $325500 $4E+06 $3874500
6 0 0 1 1 1 0 9 $76000 $170000 $94000
7 1 0 0 2 1 0 7 $17725 $130500 $112775
8 4 0 0 7 4 0 30 $220000 $0 $-220000
9 2 2 1 7 2 1 5.5 $31125 $97800 $66675
10 0 0 1 1 1 1 14 $12350 $19575 $7225
11 0 0 1 3 2 0 18 $22800 $13500 $-9300
12 0 0 0 0 1 0 18 $2600 $0 $-2600
13 0 0 0 0 1 0 18 $2000 $0 $-2000
14 0 0 0 0 1 0 20 $7500 $21740 $14240
15 0 0 1 1 1 1 8 $30800 $17200 $-13600
16 0 0 0 0 1 0 9 $2000 $0 $-2000
17 0 0 2 2 1 0 4 $12000 $7000 $-5000
18 2 1 1 6 0 0 1.5 $5300 $23220 $17920
19 0 0 1 1 1 0 3 $1900 $8050 $6150
20 0 0 1 1 0 0 3 $1000 $4025 $3025
21 0 0 1 1 0 0 3 $1000 $4025 $3025
22 0 0 1 1 0 0 3 $1000 $4025 $3025
23 0 0 1 1 1 1 3 $3525 $3125 $-400
24 0 0 1 1 0 0 3 $3000 $8400 $5400
25 0 0 0 0 1 0 18 $1900 $0 $-1900
26 0 0 0 0 1 0 8 $1900 $0 $-1900
27 1 0 1 2 1 0 19 $12125 $14985 $2860
28 0 0 1 1 1 0 2.5 $1700 $6500 $4800
29 0 1 0 1 1 1 8 $12880 $11700 $-1180
30 0 0 1 1 1 0 4.5 $3060 $6300 $3240
31 1 2 1 5 0 0 3 $4250 $10900 $6650
32 0 1 1 3 0 0 1.5 $2400 $5375 $2975
33 2 0 1 5 1 0 6 $38900 $165440 $126540
34 0 0 1 1 1 0 1 $1500 $10750 $9250
35 1 1 1 5 1 0 3 $12640 $66100 $53460
36 0 0 1 2 1 1 10 $18780 $34056 $15276
37 1 0 1 3 1 1 13 $38584 $46300 $7716
38 0 1 1 2 1 0 12 $15690 $236280 $220590
39 0 0 1 1 0 0 1.5 $1275 $11927 $10652
142
LIST OF REFERENCES
Abraham and Mackay (2001) Discussion - Six Sigma Black Belts: What do
they need to know? Journal of Quality Technology, 33 (4): pp. 410-413, Oct.
Ackermann C.S., and Fabia J.M. (1993) Monitoring supplier quality at ppm
levels, IEEE Transactions on Semiconductor Manufacturing, 6 (2): pp.189-195,
May
Ahire S., Landeros R., and Golhar D. (1995) Total quality management: a
literature review and an agenda for future research, Production and Operations
Management, pp. 277-307
Ali O.G., Chen Y.T. (1999) Design quality and robustness with neural
networks, IEEE Transactions on Neural Networks, 10 (6): pp. 1518-1527, Nov.
Antony J. and Coronado R.B. (2002) Design for Six Sigma, Manufacturing
Engineer, 81 (1): pp. 24-26
Arvidsson M., Gremyr I., and Johansson P. (2003) Use and knowledge of
robust design methodology: a survey of Swedish industry, Journal of Engineering
Design, 14(2): pp. 129-143, Jun.
143
Bailey S.P. (2001) Discussion - Six Sigma Black Belts: What do they need to
know? Journal of Quality Technology, 33 (4): pp. 426-431, Oct.
Bartos, F.J. (1999) Six sigma for complex systems, Control Engineering,
46(3): pp. 90, Mar.
Basu R. (2001) Six sigma to fit sigma, IIE Solutions, 33 (7): pp. 28-33, Jul.
Berlowitz, D.R. (2003) Striving for Six Sigma I pressure ulcer care, Journal of
American Geriatrics Society, 51(9): pp. 1320-1321, Sept.
Binder R.V. (1997) Can a manufacturing quality model work for software?
IEEE Software, 14 (5): pp. 101, Sep-Oct.
Blakeslee J.A. (1999) Implementing the Six Sigma Solution – How to achieve
quantum leaps in quality and competitiveness, Quality Progress, 32 (7): pp. 77-85,
Jul.
144
Blanton P. (2002) Quality tools in science education. The Physics Teacher,
40: pp.188-189, Mar.
Bossert J. (2003) Lean and Six Sigma – Synergy made in heaven, Quality
Progress, 36 (7): pp. 31-32 Jul.
Brady, J. and T. Allen (2002) Case Study Based Instruction of SPC and DOE,
The American Statistician, 56, 4, 1-4.
Breyfogle F.W. (2002) Golf and Six Sigma – Use Six Sigma metrics to drive
proper process behavior, Quality Progress, 35 (11): 83-85, Nov.
Breyfogle F.W., Enck D., and Meadows B. (2001) Discussion - Six Sigma
Black Belts: What do they need to know? Journal of Quality Technology. 33 (4):
pp. 424-425, Oct.
Broderick L.S., Knuteso, H.L., Rankin R.J. et. al. (2002) Use of Six Sigma
methodology to enhance capacity management in an academic center-first year’s
experience, Radiology 225: 1223 Suppl. S Nov.
Buck C., Miller R., and Desmarais J. (2001) Six Sigma – The quest for
quality, Hospitals & Health Networks, 75 (12): pp. 41-48, Dec.
Buck C.R. (1998) Health care through a Six Sigma lens, Milbank Quarterly,
76 (4): pp. 749+
Buck C.R. (2001) What Hospital leaders say about Six Sigma, Hospitals &
Health Networks, 75(12): pp. 43, Dec.
145
Buggie F.D. (2000) Beyond ‘Six Sigma’, Journal of Management
Engineering, 16 (4): pp. 28-31, Jul.-Aug.
Card D.N. (2000) Sorting out Six Sigma and the CMM, IEEE Software, 17
(3): pp. 11-13, May-Jun.
Chassin M.R. (1998) Is healthcare ready for Six Sigma quality?, Milbank
Quarterly, 76 (4): pp. 565+
Clifford L. (2001) Trend spotting – Why you can safely ignore Six Sigma,
Fortune, 143 (2): pp. 140, Jan.
Coleman S.Y., Arunakumar G., Foldvary F., et al. (2001) SPC as a tool for
creating a successful business measurement framework, Journal of Applied
Statistics, 28 (3-4): pp. 325-334, Mar. May
Cook, B.M., (1990) In Search of Six Sigma: 99.9997 per cent Defect-free,
Industry Week, 1 October, pp. 60-65
146
Cooper N. P. and Noonan P. (2003) Do teams and Six Sigma go together?
Quality Progress, 36 (6): pp. 25-28.
Crosby, P.B. (1980) Quality is Free: The Art of Making Quality Certain, New
York; McGraw-Hill
Dasgupta T. (2003) Using the six-sigma metric to measure and improve the
performance of a supply chain, Total Quality Management & Business Excellence,
14 (3): pp. 355-366, May.
Davig W., Brown S., Friel T., and Tabibzadeh K. (2003) Quality management
in small manufacturing, Industrial Management & Data Systems, 103 (1-2): pp.
68-77
Dean, J., Bowen, D. (1994). Managing theory and total quality: improving
research and practice through theory development. Academy of Management
Review 19 (3), pp. 392–418.
Dedhia N.S. (1995) Survive Business challenges with the total quality
management approach, Total Quality Management, 6 (3): pp. 265-272, Jul.
DeFeo J.A. (2000) An ROI story, Training & Development, 54 (7): pp. 25+,
Jul.
De Mast J., Schippers WAJ, Does RJMM, et al. (2000) Steps and strategies in
process improvement, Quality and Reliability Engineering International, 16 (4):
pp. 301-311, Jul-Aug.
Deming, W.E., (1986) Out of the Crisis, Cambridge: MIT, Center for
Advanced Engineering Study.
147
Deshpande P.B., (1998) Emergine technologies and Six Sigma, Hydrocarbon
Processing, 77(4): pp. 55, Apr.
Dimock P.V., Technical Editor, (1977) Engineering and Operations in the Bell
System, Bell Telephone Laboratories, Inc.
Does R., van den Heuvel E., de Mast J. and Bisgaard S. (2002) Quality
quandaries: comparing nonmanufacturing with traditional applications of Six
Sigma, Quality Engineering, 15(1): pp. 177-182
Doganaksoy N., Hahn G.J., Keeker W.Q. (2000) Product life data analysis: A
case study, Quality Progress, 33(6): pp. 115, June
Dornheim M.A. (2001) Implement Six Sigma, Aviation Week & Space
Technology, 155(1): pp. 25, July
Douglas P.C., Erwin J. (2000) Six sigma focus on total customer satisfaction,
Journal of Quality & Participation, 23(2): pp. 45-49
148
Feng CXJ and Kusiak A. (1997) Robust tolerance design with the integer
programming approach, Journal of Manufacturing Science and Engineering
Transactions of the ASME, 119 (4A): pp. 603-610, Nov.
Ferrin D.M., Muthler D. and Miller M.J. (2002) Six Sigma and simulation, so
what’s the correlation? Proceedings of the 2002 Winter Simulation Conference,
pp. 1439-1443.
Folaron J. (2003) The Evolution of Six Sigma, Six Sigma Forum Magazine,
pp. 38-44, Aug.
Fuller H.T. (2000) Observations about the success and evaluation of Six
Sigma at Seagate, Quality Engineering, 12: pp. 311-315
Gano D.L. (2001) Effective problem solving a new way of thinking, Annual
Quality Congress Transactions, pp.110-122.
Garvin D.A. (1988) Managing Quality: the strategic and competitive edge,
New York: Free Press
Gautreau N., Yacout S., and Hall R, (1997) Simulation of Partially Observed
Markov Decision Process and Dynamic Quality Improvement, Computers &
Industrial Engineering, 32 (4): pp. 691-700
Gill M.S. (1990) Stalking Six Sigma, Business Month, pp. 42-46, Jan.
Gnibus R.J. (2000) Six Sigma’s Missing Link – Understanding the quality
tool needed to calculate sigma ratings, Quality Progress, 33 (11): pp. 77+, Nov.
149
Goh T.N. (2001) A pragmatic approach to experimental design in industry,
Journal of Applied Statistics, 28 (3-4): pp. 391-398, Mar-May.
Goh T.N., Xie M. (2003) Statistical control of a Six Sigma process, Quality
Engineering, 15 (5): pp. 587-592
Goh T.N., Low P.C., Tsui K.L. and Xie M. (2003) Impact of Six Sigma
implementation on stock price performance, Total Quality Management &
Business Excellence, 14 (7): pp. 753-763
Greek D. (2000) Inefficiency won’t wash, Professional Engineering, pp. 45, Jun
Gross J.M. (2001) A road map to Six Sigma quality, Quality Progress, 34
(11): 24-29, Nov.
Hahn G. J. and Hoerl R.W. (1998) Key challenges for statisticians in business
and industry, Quality Progress, 31 (8): pp. 195-200, Aug.
Hahn G. J., Hill W.J., Hoerl R.W., and Zinkgraf S.A. (1999) The Impact of
Six Sigma Improvement – A Glimpse Into the Future of Statistics, The American
Statistician, 53 (3): pp. 208-215, Aug.
Hahn G.J., Doganaksoy N., and Hoerl R. (2000) The Evolution of Six Sigma,
Quality Engineering, 12 (3): pp. 317-326
Hahn G.J., Doganaksoy N., and Stanard C. (2001) Statistical tools for Six
Sigma – What to emphasize and de-emphasize in training, Quality Progress, 34
(9): pp. 78-82, Sep.
Hahn G.J. (2002) Deming and the proactive statistician, The American
Statistician, 56 (4): 290-298, Nov.
150
Hammer M. (2002) Process management and the future of Six Sigma, IEEE
Engineering Management Review, 30 (4): pp. 56-63
Harrold D., Bartos F.J. (1999) Optimize existing processes to achieve Six
Sigma capability, Control Engineering, 46(3): pp. 87-103, Mar.
Harry, M.J. (1994) The Vision of Six Sigma: A Roadmap for Breakthrough,
Phoenix: Sigma Publishing Company
Harry, M.J. (1988) The Nature of Six Sigma Quality, Illinois: Motorola
University Press.
Harry M.J. (1998b) Six sigma article inaccurate – Author’s reply, Quality
Progress, 31 (8): pp. 10, Aug.
Harry M.J. (2000a) A new definition aims to connect quality with financial
performance, Quality Progress, 33 (1) pp. 64-66
Harry M.J. (2000b) Six Sigma leads enterprises to coordinate efforts, Quality
Progress, 33 (3): pp. 70-72, Mar.
Harry M.J. (2000d) Abatement of business risk is key to Six Sigma, Quality
Progress, 33 (7), 72+, Jul.
Harry M.J. (2000e) The quality twilight zone, Quality Progress, 33(2): pp. 68,
Feb.
Harry M.J. (2000f) Quality leads, answers follow, Quality Progress, 33(5): pp.
82, May
Harry MJ. and Crawford D. (2005) Six Sigma – The next generation, Machine
Design, pp. 126-132, Feb. 17
151
Harry, M.J. and Schroeder, R. (2000) Six Sigma: The Breakthrough
Management Strategy Revolutionizing the World’s To Corporations, Currency
Henretta K., Walker J., and Bellafiore L. (2003) Applying “Six Sigma” to
chromatography – Tutorial: Cutting costs through process improvements, Genetic
Engineering News 23 (1): 54-56 Jan.
Hill W.J. (2001) Discussion - Six Sigma Black Belts: What do they need to
know? Journal of Quality Technology, 33 (4): pp. 421-423, Oct.
Hoerl R.W. (1998) Six Sigma and the future of the quality profession, Quality
Progress, 31 (6): pp. 35-42, Jun.
Hoerl R. W. (2001a) Six Sigma Black Belts: What Do They Need to Know?
Journal of Quality Technology, 33 (4): PP. 391-406, Oct.
Hoerl R.W. (2001b) Response - Six Sigma Black Belts: What do they need to
know? Journal of Quality Technology, 33 (4): pp. 432-435, Oct,
Horst R.L. (1999) Safety and Six Sigma, Manufacturing Engineering, 122 (2):
pp. 14, Feb.
Howell D. (2000) The power of six, Professional Engineering, 13 (14), pp. 34-
35, Jul. 19
Howell D. (2001) At sixes and sevens, Professional Engineering, pp. 27, May
Hunter D. (1999) Six Sigma steps, Chemical Week, 161 (33): pp. 3, Sep.
152
Hunter J.S. (1989) A one Point Plot Equivalent to the Shewhart Chart with
Western Electric Rules, Quality Engineering, Vol. 2
Hutchins G. (2000) The branding of Six Sigma, Quality Progress, 33 (9): pp.
120-121, Sep.
Ingle S. and Roe W. (2001) Six sigma black belt implementation, The TQM
Magazine, 13(4): pp. 273-280
Johnson A. and Swisher. B. (2003) Now Six Sigma improves R&D, Research-
Technology Management 46 (2): pp.12-15 Mar-Apr.
Johnstone P.A. and Dernbach A.H. (2002) Six sigma quality and delivery of
radiation therapy, Cancer Journal, 8 (6): pp. 44, Nov-Dec.
Juran, J.M. and Gryna F. (1980) Quality Planning and Analysis, New York:
McGraw-Hill
Juran, J.M. (1989) Juran on Leadership for Quality, New York: The Free
Press
Kandebo S.W. (1999) Lean, Six Sigma yield dividends for C-130J, Aviation
Week and Space Technology, 151 (2): pp. 59-61, Jul. 21
Kane L.A. (1998) The quest for Six Sigma, Hydrocarbon Processing, 77 (2):
pp. 15, Feb.
153
Kazmierczak S.C. (2003) Laboratory quality control: Using patient data to
assess analytical performance, Clin Chem Lab Med, 41(5): pp. 617-627
Kendall J. and Fulenwider D.O. (2000) Six sigma, e-commerce pose new
challenges, Quality Progress, 33 (7): 72+, Jul.
Kenett R.S., Coleman S., and Stewardson D. (2003) Statistical efficiency: The
practical perspective, Quality and Reliability Engineering International, 19: pp.
265-272
Koch P.N. (2002) Probabilistic design: optimizing for Six Sigma quality,
AIAA-2002-1471
Koonce D., Judd R., Sormaz D, et al. (2003) A hierarchical cost estimation
tool, Computers In Industry, 50 (3): pp. 293-302 Apr.
Kunes R. (2002) Six Sigma article is misleading, Quality Progress, 35 (3): pp.
8 Mar.
Lucier G.T. and Seshadri S. (2001) GE Takes Six Sigma Beyond The Bottom
Line, Strategic Finance, May, pp. 40-46
Leffew K.W., Yerrapragada S.S., and Deshpande P.B. (2001) 6 sigma and
solid-state polymerization, Chemical Engineering Communications, 188: 109-114.
Linderman K., Schroeder R.G., Zaheer S. and Choo A.S. (2003) Six Sigma: A
goal-theoretic perspective, Journal of Operations Management, 21, (2), pp. 193-
203.
Lucas J. (2002a) The essential Six Sigma – How successful Six Sigma
implementation can improve the bottom line, Quality Progress, 35 (1), pp. 27-31,
Jan.
154
Lucas J. (2002b) Six Sigma article is misleading – Response, Quality
Progress, 35 (3): pp. 8-8 Mar.
Mader D.P. (2002) Design for Six Sigma, Quality Progress, 35 (7): pp. 82+,
Jul.
Maguire M. (1999a) Six sigma saga, Quality Progress, 32 (10): pp. 6, Oct.
Magure M. (1999b) Cowboy Quality: Mikel Harry’s riding tall in the saddle
as Six Sigma makes its mark, Quality Progress, 32 (10): pp. 27-34, Oct.
Main, J., (1994) Quality Wars: The Triumphs and Defeats of American
Business, New York: The Free Press
Mandal P., Howell A. and Sohal A.S. (1998) A systemic approach to quality
improvements: The interactions between the technical, human and quality sytems,
Total Quality Management, 9 (1): pp. 79-100
Mason R.C. and Young J.C. (2000) Interpretive features of a T(2) chart in
multivariate SPC, Quality Progress, 33(4): pp. 84-89, Apr.
McFadden F.R. (1993) Six sigma quality programs, Quality Progress, pp. 37-
42, Jun.
McKeachie, W.J. (1993), Teaching Tips, D.C. Heath and Co., Lexington, MA.
155
Montgomery D. (2002) Changing roles for the Industrial Statistician, Quality
and Reliability Engineering International, 18(5):
Montgomery D.C., Lawson C., Molnau W.E, et al. (2001) Discussion - Six
Sigma Black Belts: What do they need to know? Journal of Quality Technology.
33 (4): pp. 407-409 Oct.
Motyka, M. (2000) Six Sigma, QS-9000 article has one minor flaw, Quality
Progress, vol. 33, no. 8, Aug., p.8
Munro R. (2000) Linking Six Sigma with QS-9000. Quality Progress, pp. 47-
53, May
Munro R. (2000) Six sigma, QS-9000 article has one minor flaw – Response,
Quality Progress, 33 (8): pp. 8, Aug.
Murugappan M. and Keeni G. (2003) Blending CMM and Six Sigma to meet
business goals, IEEE Software, (2): 42+ Mar-Apr.
Nave D. (2002) How to compare Six Sigma, lean and the theory of constraints
– A framework for choosing what’s best for your organization, Quality Progress,
35 (3): pp. 73-78, Mar.
Neuscheler F.D. and Norris R. (2001) Capturing financial benefits from Six
Sigma – five lessons learned will resonate with top management, Quality Progress,
34 (5): pp. 39-44, May.
Nevalainen D.E., Berte L., Kraft C., et. al. (2000a) Evaluating laboratory
performance on quality indicators with the Six Sigma scale, Archives of Pathology
& Laboratory Medicine, 124 (4), pp. 516-519, Apr.
Noble T. (2001) Six sigma boosts the bottom line, Chemical Engineering
Progress, 97 (4): pp. 9-11, Apr.
156
Nolan, D. and T. P. Speed (1999), “Teaching Statistics Theory Through
Applications,” The American Statistician, Vol. 53, No. 4, pp. 370-375.
Pande P.S., Neuman R.P., and Cavanagu R.R., (2000) The Six Sigma Way,
McGraw-Hill
Pearson T.A. (2001) Measure for Six Sigma success, Quality Progress, 34 (2):
pp. 35-40, Feb.
Plotkin C.W., Carlson C.S., Gregory F.D., et. al. (1999) Panel: Advisory
board – What are the successful companies doing? Annual Reliability and
Maintainability Symposium 1999 Proceedings, pp. 219-223
Pyzdek, T., (2000) Six Sigma Handbook: A Complete Guide for Greenbelts,
Blackbelts, & Managers at All Levels, New York: McGraw-Hill
Pyzdek T. (2001a) Why Six Sigma is not TQM, Quality Digest, pp. 26, Feb.
Pyzdek T. (2001b) Discussion - Six Sigma Black Belts: What do they need to
know? Journal of Quality Technology, 33 (4): pp. 418-420, Oct.
157
Rasis D., Gitlow H.S., and Popovich E. (2003b) “Paper Organizers
International: A Fictitious Six Sigma Green Belt Case Study, II”, Quality
Engineering, 15 (2): pp. 259-274
Riley J.B., Justison G.A., Povrzenic D., et al. (2002) Designing an integrated
extracorporeal therapy service quality system, Therap Apher, 6 (4): 282-287, Aug.
Sanders D. and Hild C. (2001) Common myths about Six Sigma, Quality
Engineering, 13 (2): pp. 269-276
Sarewitz S.J. (2000) Evaluating laboratory performance with the Six Sigma
scale, Archives of Pathology & Laboratory Medicine, 124 (12): pp. 1748, Dec.
Scalise D. (2001) Six sigma: the west for quality, Hospitals & Health
Networks, 75(12): pp. 41, Dec.
Scalise D. (2003) Six Sigma in action – Case studies in quality put theory into
practice, Hospitals & Health Networks 77 (5): pp. 57+ May
Schmitt B. (2001) Expanding Six Sigma, Chemical Week, Feb. 13: pp. 34
Schmitt B. (2002) A slow spread for Six Sigma, Chemical Week, Feb. 21: pp. 21
158
Shewhart W.A. (1931) Economic Control of Manufactured Product, New
York: D. Van Nostrand, Inc.
Shina, S. G., Six Sigma for Electronics Design and Manufacturing, McGraw-
Hill 2002
Smith B. (2003) Lean and Six Sigma – A one-two punch, Quality Progress, 36
(4): pp. 37-41 Apr.
Snee R.D. (1999) Why should statisticians pay attention to Six Sigma?
Quality Progress, 32 (9): pp. 100-103, Sept.
Snee R.D. (2000b) Six sigma improves both statistical training and processes,
Quality Progress, 33 (10): pp. 68-72, Oct.
Snee R.D. (2001a) Dealing with the Achilles’ heel of Six Sigma initiatives –
Project selection is key to success, Quality Progress, 34 (3): pp. 66 Mar.
Snee R.D. (2001b) Discussion - Six Sigma Black Belts: What do they need to
know? Journal of Quality Technology. 33 (4): pp. 414-417, Oct.
Snee R.D. (2003) The Six Sigma Sweep, Quality Progress, 36 (9): pp. 76+
Sousa R., and Voss C.A. (2002) Quality management re-visited: a reflective
review and agenda for future research, Journal of Operations Management, 20: pp.
91-109
Stamatis (2000) Who needs Six Sigma, anyway? Quality Digest, pp. 33-38,
May
Stein P. (2001) Measurements for business, Quality Progress, 34(2): pp. 29,
Feb.
159
Studt T. (2002) Implementing Six Sigma in R&D, R&D Magazine, 44 (8): pp.
21-23, Aug.
Tang L.C., Than S.E., and Ang B.W. (1997) A graphical approach to
obtaining confidence limits of C-pk, Quality and Reliability Engineering
International, 13 (6): pp. 337-346, Nov-Dec.
Treichler D., Carmichael R., Kusamanoff A., et al. (2002) Design for Six
Sigma: 15 lessons learned – Leading corporations find out how to avoid pitfalls,
Quality Progress, 35 (1): 33-42, Jan.
Tylutki T.P., and Fox D.G. (2002) Mooooving toward Six Sigma, Quality
Progress, 35 (2): 34-41, Feb.
Vandenbrande (1998) How to use FMEA to reduce the size of your quality
toolbox. Quality Progress, pp. 97-100, Nov.
Vaugham T.S. (1998) Defect rate estimation for “Six Sigma” processes,
Production & Inventory Management Journal, 39(4): pp. 5-9, Oct.
Velocci A.L. (1998) Pursit of Six Sigma emerges as industry trend, Aviation
Week and Space Technology, 149 (20): 52+, Nov. 16
Velocci A.L. (1998) High hopes riding on Six Sigma at Raytheon, Aviation
Week and Space Technology, 149 (20): 59, Nov. 16
Velocci A.L. (2000) Raytheon Six Sigma meets initial target, Aviation Week
and Space Technology, 152 (13): pp. 59, Mar. 27
Velocci A.L. (2002) Full potential of Six Sigma eludes most companies,
Aviation Week and Space Technology, 157 (14): 56-60 Sep. 30
160
Voelkel J.G. (2002) Something’s missing – An education in statistical
methods will make employees more valuable to Six Sigma corporations, Quality
Progress, 35 (5): 98-101, May
Walsh K., Fuller J., Wood A., et al. (2000) Six Sigma – Marshaling an attack
on costs, Chemical Week, 162 (9): pp. 25-27, Mar. 1
Watson G. H., (2002a) Selling Six Sigma to Upper Management, Six Sigma
Forum Magazine, 1 (4): pp. 26-37, Aug.
Wiklund H. and Wiklund P.S. (2002) Widening the Six Sigma concept: An
approach to improve organizational learning, Total Quality Management, 13 (2):
pp. 233-239, Mar.
161
Wyper B., and Harrison A. (2000) Deployment of Six Sigma methodology in
human resource function: A case study, Total Quality Management, 11 (4-6):
S720-S727 Sp. Iss. SI Jul.
Yeung ACL, Chan L.Y., and Ledd T.S. (2003) An empirical taxonomy for
quality management systems: a study of the Hong Kong electronics industry,
Journal of Operations Management, 21 (1): pp. 45-62
Zain Z.M., Dale B.G., and Kehoe D.F. (2001) Total quality management: an
examination of the writings form a UK perspective, The TQM Magazine, 13 (2):
pp. 129-137
162