The DMAIC (Define-Measure-Analyze-Improve-Control) method in Six Sigma is often described as an approach for problem solving. This paper compares critically the DMAIC method with insights from scientific theories in the field of problem solving. As a single authoritative account of the DMAIC method does not exist, the study uses a large number of sources, consisting of prescriptive accounts of the method in the practitioner literature. Five themes are selected from the problem solving literature for the analysis of DMAIC—generality versus domain specificity of methods; problem structure; generic problem solving tasks; diagnostic problem solving; and remedial problem solving. The study provides a characterization of the types of problems for which DMAIC is a suitable method, but also identifies problems for which it may be ineffective. An important limitation of the method is its generality, which limits the methodological support it provides, and which fails to exploit task-domain specific knowledge. Domain-specific adaptations of the method partly overcome these weaknesses. Among the method's strengths are the powerful statistical techniques for fact finding and empirical verification of ideas, and the DMAIC stage model, which acts as a problem structuring device. The most prominent limitation identified in this study is Six Sigma's inferior methodology for efficient problem diagnosis. Methodological support for the identification of potential problem causes is offered as an incoherent and poorly structured collection of techniques, without strategic guidance to ensure efficiency of the diagnostic search. Adopters of the method should be aware of its potential limitations.
Six Sigma is defined by Linderman et al. (2003) as “(…) an organized and systematic method for strategic process improvement and new product and service development that relies on statistical methods and the scientific method to make dramatic reductions in customer defined defect rates.”
Academic research, such as Zu et al. (2008) and Schroeder et al. (2008), has tried to determine which elements in Six Sigma make it effective. Besides its role structure and focus on metrics, Six Sigma's structured improvement procedure is seen as a novel and effective contribution to quality management. This improvement procedure is generally known under the acronym DMAIC, standing for Define, Measure, Analyze, Improve and Control.
DMAIC is similar in function as its predecessors in manufacturing problem solving, such as Plan-Do-Check-Act and the Seven Step method of Juran and Gryna (Balakrishnan et al., 1995). In the theory of organizational routines, DMAIC is a metaroutine: a routine for changing established routines or for designing new routines (Schroeder et al., 2008). Originally described as a method for variation reduction, DMAIC is applied in practice as a generic problem solving and improvement approach (McAdam and Lafferty, 2004). It is instrumental in the implementation of Six Sigma as a process improvement methodology (Chakravorty, 2009).
Six Sigma and its DMAIC method emerged and developed in practice. It built on insights from the quality engineering field, incorporating ideas from statistical quality control, total quality management and Taguchi's off-line quality control. Their wide adoption in practice warrants a critical scientific analysis. One aspect of a scientific evaluation of Six Sigma is to critically compare its principles with insights from established scientific theories.
This work aims to study the Six Sigma DMAIC method from the perspective of scientific theories in the field of problem solving as published in the operations research and management science (OR/MS) and industrial engineering (IE) literatures. Six Sigma is often described as a problem solving methodology, and for that reason, theoretical insights from the problem solving literature should provide insights on DMAIC. The purpose of the analysis is to identify limitations of the method. These identified limitations may be an inducement for attempts at improving the method. But some limitations may be inherent to DMAIC, as it is not plausible that a strong method can be applicable without restrictions in all circumstances. In those cases, the practical value of identified limitations is that they provide a basis for advising users when the DMAIC method is suited.
Since an authoritative or uniform account of the DMAIC method does not exist, we have worked with a large number of sources, varying in degrees of quality, clarity and coverage. In the next section, we describe the sources we have used to obtain a carefully balanced understanding and rendering of the DMAIC procedure. We also outline our approach for studying DMAIC from the perspective of a number of themes in the problem solving literature. The subsequent sections treat these themes, and each formulates a number of conclusions. In Section 8, we seek to integrate the individual conclusions into a comprehensive characterization of DMAIC as a problem solving method.
We start this concluding section by pointing out the most
important limitation of this research. In this study, we conceive of
DMAIC as a problem solving method, and analyze it from that
perspective. We claim that this perspective gives useful insightsand results in useful conclusions about the method, but we
emphasize that this perspective is by no means exclusive, and
other perspectives (for instance, the goal-theoretic perspective
chosen in
Linderman et al., 2003
) may result in equally interesting
conclusions.
The seven major conclusions of our studies have been pre-
sented in the preceding sections. Here we seek to integrate this
set of conclusions into a comprehensive view on DMAIC as a
problem solving method.
Our study has brought to light some characteristics of problem
tasks for which DMAIC may be a suitable method. DMAIC is
applicable to empirical problems ranging from well-structured to
semi-structured, but not to ill-structured problems or pluralistic
messes of subjective problems (
people problem solving
, in the
framework used in the paper). DMAIC is suitable for rather
extensive problem solving tasks, requiring all of the components
of problem definition, diagnosis, and the design of remedies. It is
less suited for problem tasks of a smaller scope.
Six Sigma is a
generic
method. The advantage of such methods
is that they are versatile. The disadvantage is that task-domain
specific methods can be more powerful because they can be more
specific and operational in the guidance they can provide. Task-
domain specific methods can also benefit from advanced task-
specific domain knowledge, which we found to be absent in
generic accounts of Six Sigma. Domain specific elaborations of
DMAIC partly overcome these weaknesses inherent to general
methods. The limitations of generic versions of DMAIC are not
generally recognized in the practitioners’ literature. An exception
is
Goh (2010)
, who mentions as the first out of a number of ‘Six
Sigma tragedies’: ‘‘The belief that Six Sigma (as typical Black Belts
know it) is universally applicable’’, and thus, that mastery of
DMAIC obviates domain-specific expertise.
Six Sigma has its origin in quality engineering, which has
traditionally had a strong emphasis on statistical methods. For
example,
ASQ (2007a)
suggests that around 45% of the questions
in the ASQ’s Black Belt exam are about statistical concepts and
techniques, as are, for example, 55% of the pages in a book on Six
Sigma techniques such as
Pyzdek (2003)
. Even in books on
Lean
Six Sigma, statistical techniques are dominant. For example, 45%
of the pages in
George et al. (2004)
are devoted to statistics. The
strong basis in statistical methodology provides strength to the
method, which offers powerful techniques for fact finding and
empirical testing of ideas before they are accepted. It is also
responsible for some of the limitations of the method, in that
methods originating in fields other than statistics are under-
represented. For example, SPC (statistical process control) tech-
niques are emphasized for process control, but PID (proportional-
integral-derivative) controllers (
Box et al., 2009
), which originate
in the field of control engineering, are virtually absent in the
method’s toolbox. Statistical techniques for empirical model
building, such as the theory of the design and analysis of
experiments, are emphasized, while other methods for model
building, such as the finite element method (
Reddy, 2005
)or
techniques from operations research, that may be more appro-
priate in many situations, are not offered as alternatives.
The strong and somewhat one-sided origins in statistics may
also be responsible for the unsatisfactory methodological support
that DMAIC offers for efficient problem diagnosis. We propose
that in this respect, it should be considered inferior to competing
problem solving methodologies in industrial engineering, such as
Shainin’s (
Steiner et al., 2008
) and Kepner and Tregoe’s (
Kepner
and Tregoe, 1997
) problem solving approaches.
One of the acclaimed strengths of Six Sigma is its structured
method (
Zu et al., 2008
;
Schroeder et al., 2008
). Reasoning from
the perspective of problem solving, we note that the DMAIC
model functions as a problem structuring device. It breaks down aproblem solving task into a sequence of generic subtasks, repre-
sented and defined by the Define-Measure-Analyze-Improve-
Control stages. More detailed accounts break down these subtasks
into more specific deliverables (see the generic steps in
Table 1
).
Deliverables and subtasks, finally, are associated to problem
solving techniques, such as gage R&R studies, process capability
analysis and design and analysis of experiments (see
De Koning
and De Mast, 2006
). In this fashion, the DMAIC procedure helps a
user to find a strategy for analyzing and solving a problem, and
thus structure the problem at hand.
We believe that adopters of Six Sigma methodology and the
DMAIC problem solving approach should be aware of their
characteristics and potential limitations. This paper has high-
lighted the characteristics of the DMAIC approach and its limita-
tions, specifically from a problem solving perspective. In addition,
the paper has pinpointed directions where the approach may be
improved