ARTICLE IN PRESS

Int. J. Production Economics 111 (2008) 456467

www.elsevier.com/locate/ijpe

On the optimal selection of process alternatives

in a Six Sigma implementation
U. Dinesh Kumara, David Nowickib,
Jose Emmanuel Ramrez-Marquezb,, Dinesh Vermab
a
Indian Institute of Management Bangalore, Bannerghatta Road, Bangalore 560076, India
b
Department of Systems Engineering and Engineering Management, Stevens Institute of Technology, Castle Point on Hudson,
Hoboken, NJ 07030, USA
Received 1 January 2006; accepted 1 February 2007
Available online 27 February 2007

Abstract

Six Sigma is at the top of the agenda for many companies that try to reduce cost and improve productivity. Many of the
top manufacturing companies implement thousands of Six Sigma projects every year and this implementation demands a
signicant investment of capital that requires a careful analysis to make sure that the benets obtained are much higher
than the actual investment. This cost benet analysis is crucial, especially for companies whose products have a small prot
margin. In this paper, two optimization models that will assist management to choose process improvement opportunities
are presented. These models consider a multi-stage, asynchronous manufacturing process with the opportunity to improve
quality (scrap and rework rates) at each of the stages. The rst model is to maximizing the sigma quality level of a process
under cost constraint while the selection of Six Sigma alternatives to maximize process returns is considered by the second
model. Process quality improvement usually results in costs associated with the purchase of new technology, modication
of existing equipment, training employees, hiring new employees and investment in information technology infrastructure.
The proposed models recognize that a company competes for funds and that benets can result in either improved revenue
or reduction in costs. An example illustrates the application of the optimization models developed and results show that in
some scenarios implementing Six Sigma may not be nancially benecial.
r 2007 Elsevier B.V. All rights reserved.

Keywords: Mathematical programming; Optimal allocation; Process improvement; Six Sigma

1. Introduction to reduce manufacturing defects. The concept

Six Sigma, a trademark of Motorola, was
introduced more than 20 years back as a method
Corresponding author. Tel.: +98 201 216 8003;
fax: +98 201 216 5080.
E-mail addresses: dineshk@iimb.ernet.in (U.D. Kumar),
dnowicki@stevens.edu (D. Nowicki), jmarquez@stevens.edu
(J.E. Ramrez-Marquez), dverma@stevens.edu (D. Verma).
behind this method was developed by William
Smith to deal with the high failure rate experienced
by the systems produced. Smith proposed Six Sigma
as a tool to improve the reliability and the quality of
products and thus, focused it at reducing defects by
improving manufacturing processes. Initially devel-
oped as an operational strategy, Six Sigma has
evolved into a competitive corporate strategy used
extensively throughout the corporate world. Even

0925-5273/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.ijpe.2007.02.002
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467 457

Nomenclature cm
i Marginal production cost of an item at
the ith stage
i ith stage of process, i 1, y, n Qik Implementation cost for the kth process
Y Process yield alternative for the ith stage
j jth defect type for each stage, j 1,2. S Raw material available at the rst stage
(1 scrap, 2 rework) D Process demand
k kth Six Sigma implementation alterna- B Budget available for Six Sigma alterna-
tive, k 1, y, qi tives
fij Rate of defect type j at state i Ki Production capacity for the ith produc-
rijk Rate of defect type j for alternative k in tion stage
state i Ri Amount of rework resulting from the ith
dik Decision variable related to the kth stage
process alternative for the ith stage cR
i Rework cost at the ith stage
Ii Number of units that are processed at the ci X Scrap cost at the ith stage
ith stage Z+ Set of positive integers
U Unit selling price

traditional companies that adhere to conventional and Weiss (2005) also point that 60% of the
management frameworks have started embracing companies in the survey selected opportunities for
Six Sigma as a method of substance with the improvement on an ad-hoc basis and only 31%
potential to increase market share and protability relied on portfolio approach. It is interesting to note
(Harry, 1998). The benets of implementing Six that companies that used a portfolio approach
Sigma programs have been extensively reported in gained better results.
the literature (Hendricks and Kelbaugh, 1998; Hahn As illustrated in Table 1, Six Sigma is the process
et al., 1999; Lanyon, 2003; Robinson, 2005) and management tool that has yielded the greatest
range from the simple reduction in the number of results (Dusharme, 2006). Moreover, the fact that
manufacturing defects to the improvement of the in this table Six Sigma is ranked much higher than
market share and the competitive advantage of a other process improvement techniques, illustrates
company. In this sense, Anon (2003) conducted a the effect of concurrently implementing various
study of 13 high prole corporate houses in the US process improvement techniques given that most of
from a wide variety of industries and reported that these techniques constitute the Six Sigma toolbox.
Six Sigma programs returned more than double the This fact is important because none of the remain-
investment. ing quality improvement initiatives have much
However, as discussed by Deleryd (1999), with application outside manufacturing industry.
the increasing number of organizations using Although Six Sigma was originally conceived to
process capability studies, warnings have been reduce waste due to process deciencies in
launched that imprudent use of numerical measures
of capability, might lead the user to make erroneous
decisions. In this respect, recently there have been Table 1
many cases reported in the literature where Six Rating of process improvement techniques (Dusharme, 2006)
Sigma has failed to deliver the desired results. Process improvement tool Impact (%)
Zimmerman and Weiss (2005), quote a survey
conducted by Aviation Week magazine among Six Sigma 53.60
Process mapping 35.30
major aerospace companies, which reported that
Root cause analysis 33.50
less than 50% of the companies expressed satisfac- Cause and effect analysis 31.30
tion with results from Six Sigma projects, nearly ISO 9001 21.00
20% were somewhat satised and around 30% were Statistical process control 20.10
dissatised. Even at these levels of satisfaction, Six Total quality management 10.30
Malcolm Baldridge criteria 9.80
Sigma has been accounted to do better than many
Knowledge management 5.80
other process improvement techniques. Zimmerman
 ARTICLE IN PRESS
458 U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467

manufacturing, it is now used by almost all
industries including service industries such as health
care management (Krupar, 2003; Antony, 2004;
Antony and Fergusson, 2004; Moorman, 2005;
Frings and Grant, 2005). That is, Six Sigma has
the exibility to be used as an operational strategy
to reduce the number of defects or as a business
strategy to improve business processes and evolve
new business models. Many proponents of Six
Sigma stress that the power of Six Sigma lies in
the fact that it can be used as a business strategy to
improve market share and protability. Table 2
provides some of the different areas where Six
Sigma has been applied to improve manufacturing
and business processes. Table 3 provides an updated
list of the benets of Six Sigma as reported in Kwak
and Anbari (2006).
In general, Six Sigma can be considered as an
extension of other quality improvement initiatives
such as Demings statistical quality control and
Total Quality Management (TQM). Furthermore,
the main objective of Six Sigma, like most of other
management strategies on quality initiatives, is
focused around meeting the customer requirements.
Anbari (2002) and Kwak and Anbari (2006)
summarize Six Sigma as a strategy, which includes
TQM, strong customer focus, additional data
analysis tools, nancial results and project manage-
ment, to satisfy costumer needs.
Six Sigma uses the following ve major phases to
achieve process improvement: Dene, Measure,
Analyze, Improve and Control (DMAIC). The
DMAIC cycle has a lot of similarities with Demings
Plan-Do-Check-Act cycle (Bertels, 2003). How-
ever, Six Sigma provides a well-dened target for
quality that the process defect rate should not
exceed 3.4 defects per million opportunities. This
focused target, along with a well-dened DMAIC
procedure, has probably resulted in higher success
rate for Six Sigma as compared to TQM. Another
important difference between Six Sigma and TQM
is that Six Sigma is mostly a business results
oriented model compared to a return on investment
orientation of TQM (Bertels, 2003).
For manufacturing companies the direct benet
of Six Sigma, results from the reduction in the
number of defects due to improved manufacturing
processes. For these companies, Six Sigma or Sigma
quality level is a measure of the process defect rate
and thus, can be used to measure the quality of the
manufacturing process (i.e. a high Sigma level
indicates that the process results in a lower defect
rate whereas, a low Sigma level illustrates a higher
defect rate). Moreover, Sigma quality level also
helps to set a realistic target for improvement of
process quality during the DMAIC cycle (i.e. it can
be used as a benchmarking tool).
Reducing process variations is the core objective
of Six Sigma projects, since process variations result
in higher quality loss. In this respect, Taguchi and
Clausing (1990) reported a classic example on the
impact of process variations using the case of Ford
versus Mazda. Ford, which owned 25% of Mazda,
had asked the Japanese company to build transmis-
sions for the cars that it sold in the United States.
The transmissions were built to identical specica-
tions and Ford was adopting zero defects as its
standard. However, after the cars had been in the
market it was observed that Fords transmission
system was generating far higher warranty costs as
compared to the transmission systems built by the
Mazda. The reasons was traced to the fact the

Table 2
Six Sigma industrial applications

References Industrial application

Hendricks and Kelbaugh (1998)
Lanyon (2003)
Motwani et al. (2004)
Knowles et al. (2004)
Banuelas et al. (2005)
Snee (2005)
Edgeman et al. (2005)
Ehie and Sheu (2005)
Liu (2006)
Mukhopadhyay and Ray (2006)
Successful implementation of several Six Sigma projects that improved the net prot
Improvement of HR process using Six Sigma
Dow chemicals which implemented Six Sigma on a corporate wide basis in 2000, achieved its
target of $1.5 billion in cumulative earning before interests and taxes
Successful application of Six Sigma within a UK confectionery plant of a major food producer
Use of Six Sigma methodology to reduce waste in a coating process
Claims that Six Sigma beneted Motorola, Allied Signal, General electric etc
Claims saving between $2 and $3 million at ofce of the chief of technology ofcer (OCTO),
Washington DC using Six Sigma strategies
Demonstrates the value of Six Sigma and Theory of Constraint
Presented an application of Six Sigma to reduce cycle time and defects in clinical report entry
Used Six Sigma to reduce the yarn packing defects
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467 459

Table 3
Benets of Six Sigma in manufacturing sector

Company/project Metric/measures Benets/savings

Motorola (1992) In-process defect levels 150 times reduction

Raytheon/aircraft integration Depot maintenance inspection time (measured in Reduced 88%
system days)
GE/Railcar leasing business Turnaround time at repair shops 62% reduction
Allied Signal (Honeywell)/ Capacity/Cycle time/Inventory/On-time delivery Up 50%/Down 50%/Down 50%/Increased
laminates plant in South Carolina to near 100%
Allied Signal (Honeywell)/Bendix Concept to shipment cycle time Reduced from 18 months to 8 months)
IQ brake pads
Hughes aircrafts missile systems Quality/productivity Improved 1000%/improved 500%
group/wave soldering operations
Continental Teves/Brake and axle Failure rate More than 50% reduction in failure rate
assemblies
Borg Warner Turbo Systems Financial $ 1.5 million annually since 2002
General electric Financial $2 Billion in 1999
Motorola (1999) Financial $15 Billion over 11 years
Dow chemaical/rail delivery project Financial Savings of $2.45 million in capital
expenditures
DuPont/Yerkes plant in New York Financial Savings of more than $25 million
(2000)
Telefonica de espana (2001) Financial Savings and increase in revenue 30 million
Euro in rst 10 months
Texas instruments Financial $600 million
Johnson and Johnson Financial $ 500 million
Honeywell Financial $ 1.2 billion
Ford motor compant/exterior Financial $500,000
Surface defects

(Source: Kwak and Anbari, 2006; Weiner, 2004; De Feo and Bar-El, 2002; Antony and Banuelas, 2002; Buss and Ivey, 2001; McClusky,
2000; Anon, 2007a,b).

Fords transmissions had much higher process 1

0.9
variability compared to the transmissions built by 0.8
Mazda. Sony Corporation reported a similar case 0.7
0.6
Yield

for their televisions manufactured at Tokyo and San 0.5

Diego. The televisions manufactured at Tokyo had 0.4
0.3
less variability on the color density compared to the 0.2
0.1
televisions manufactured at San Diego. As a result 0
the customer satisfaction levels for televisions 0 1 2 3 4 5 6
manufactures at Tokyo was much higher than the Sigma quality level
televisions manufactures at San Diego.
Fig. 1. Decreasing return to scale in Yield as Sigma quality
It is well understood that the marginal benet of
increases.
any Six Sigma project decreases as the sigma quality
level increases. It is also well understood that the
process yield increases (at a decreasing rate) as requires high investment. For example, consider
the sigma quality level increases. Fig. 1 shows the improving a process from a 3 sigma level to a 4
relationship between sigma quality level and de- sigma level. This improvement reduces the number
crease in number of defects as a function of process of defects per million opportunities (DPMO) from
yield. In economical terms, this gure illustrates that 66,811 to 6210. From a Taguchi quality loss
because of the decreasing returns to scale in process perspective (Taguchi, 1986), at 3 sigma level the
yield as sigma quality increases, at some point it Taguchi quality loss is higher compared to the
may not be economically benecial to increase Taguchi quality loss at 4 sigma level. Similarly
sigma quality level, especially if the process change improving sigma level from a 5 sigma level to a 6
 ARTICLE IN PRESS
460 U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467
sigma level reduces the DPMO from 233 to 3.4.
Note however, that in most cases more effort will be
required to improve a process from 5 to 6 sigma
level as compared to improving a process from 3 to
4 sigma level. The improvement of sigma level
obviously requires reduction in the process varia-
bility, i.e., the process sigma itself.
Any improvement in Sigma level is likely to
reduce the Cost of Poor Quality (CoPQ). CoPQ is a
result of manufacturing defects and, a function of
rework cost, excessive use of material, warranty
related costs and unnecessary use of resources.
Under the assumption that CoPQ is linearly related
to the number of defects produced by the process,
the marginal benet of increasing sigma levels (i.e.
process improvement) decreases. Models such as
Taguchis quality loss function (Taguchi, 1986;
Chou and Chen, 2001; Ganeshan et al., 2001) are
used to estimate the loss due to poor quality.
Taguchi claims that the loss due to variability
follows a quadratic function (Taguchi, 1986;
Taguchi and Clausing, 1990). However, in our
model, we assume that the loss is a linear function
which depends on the number of rework and scrap
generated by a process. The most appropriate model
for evaluating loss due to poor quality would
require analysis of relevant data on case-by-case
basis.
Process improvement may involve replacing the
existing process with a new improvement alterna-
tive. In this respect, Gowen and Tallon (2005) point
out that Six Sigma programs must take into
consideration the level of technological intensity of
the organization to determine the impact of
implementing Six Sigma. The replacement or
upgrade of an existing process with a new process
alternative can be costly. Moreover, if the resulting
process does not improve the sigma level signi-
cantly, the investment made may not translate into
signicant returns. Like any other process improve-
ment technique, Six Sigma would also fail to deliver
if the management fails to understand the cost of
implementing Six Sigma and its effectiveness.
This paper presents two mathematical models
that will assist management in choosing among
different process improvement alternatives. These
models acknowledge that the implementation cost
for a successful Six Sigma initiative can be
considerable and failure to select an optimal
alternative may further translate into loss of
expected returns. The solution to the proposed
models represent the rst attempt to optimally select
Sigma process alternatives that maximize the
process sigma levels while taking into account the
cost associated with such alternatives. Any change
in a process requires investment in the form of
procurement cost of new processes and, in costs
related to operation and maintenance.
The selection of alternatives for process improve-
ment has been traditionally modeled as a multi-
criteria decision-making problem. Parkan and Wu
(2000) have demonstrated use of operational com-
petitiveness rating analysis, analytic hierarchy pro-
cess, and data envelopment analysis for process
selection using a case study by treating production
quality as one of the criteria. A mixed integer
programming approach was used by Stuart et al.
(1999) to select process under environmental con-
straints such as material consumption, energy
consumption and waste rate. Bowman and Schmee
(2004) used Monte Carlo simulation to demonstrate
sensitivity of process capability to each parameter of
the input variables and proposed a method to
improve process performance. Although the basis of
Six Sigma is process improvement, few research
studies have been carried out on the actual
alternative selection itself. Thus, this research is an
attempt to ll such a gap by providing a quantita-
tive model for selection of process in a Six Sigma
implementation project that would result in max-
imum benet
The rest of the paper is organized as follows:
Section 2 introduces a mathematical relationship
between a process Sigma level and its yield. This
relationship is subsequently used as a surrogate
metric to develop a mathematical programming
model to optimize the process sigma level, under a
cost constraint. Section 3 presents a mathematical
programming model to maximize prot based on
the optimal selection of improvement alternatives.
In Section 4, an example is used to illustrate and
compare the models presented. Finally, conclusions
and future research are presented in Section 5.
2. Sigma level optimization subject to a budget
constraint
As discussed, Six Sigma initiatives are often
viewed as a guide to improve business prots
through reduced costs and/or increased revenue.
In practice, this translates into a foregone conclu-
sion that implementing Six Sigma will provide a de
facto, justiable return on investment. However, the
implementation cost for successful Six Sigma
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467 461

initiatives can be considerably high for many the shift is likely to be as much as Ls (where s is the
companies, especially those companies with small process standard deviation), the following mathe-
prot margins and limited resources. To improve matical relationship (Kumar et al., 2006) can be
the quality of a process often requires signicant used to approximate the value of Z:
capital investment in new, emerging process alter-  
C 0 C 1 Q C 2 Q2
natives. In addition to the capital expenditure Z L Q , (1)
needed to force process change, Six Sigma initiatives 1 d 1 Q d 2 Q2 d 3 Q3
require investment in training and, quite often, where C0 2.515517, C1 0.802853, C2
cultural and behavioral change. 0.010328, d1 1.432788, d2 0.189269, d3
To account for these issues, this section presents a 0.001308 and,
mathematical programming model that can be used v
u
u   !
as a management tool to solve the problem of t Y 2
optimizing the sigma level of a process subject to a Q ln 1 . (2)
100
budget constraint. The mathematical model is
constructed by assuming that units are processed The process yield, dened by Y in (2), can be
through an n stage sequential process. For each obtained with the following formula:
stage, there are qi process alternatives that can be Number of opportunities  Number of defects
implemented to improve the process Sigma level. Y  100%.
Number of opportunites
Moreover, each of these stage improvement alter- (3)
natives has a known improvement rate (sigma level
improvement) and an associated implementation It is important to notice that Y is a measurement
cost. Then, the problem is to nd the alternatives of process performance and thus, the mathematical
that will maximize process quality improvement model presented in Model 1 uses Y as a surrogate
without exceeding the allocated budget. The sche- measure to maximize the sigma quality level.
matic of the process under consideration is pre- Model 1:
!
sented in Fig. 2. Yn Xqi
Max Y dik 1  f i1 ri1k
i1 k1
2.1. Process yield as a measure to optimize sigma
level Subject to:
qi
X
For the model under consideration, quality I. dik 18i
improvement is dened as the reduction in both k1
rework and scrap. Thus, a relationship between the
Sigma level and the stage quality improvement X qi
n X
needs to be dened. In general, the Sigma level or II. dik Qik pB
Sigma quality level can be understood as the Z i1 k1

value of the standard normal distribution under the

assumption that there is no shift in the process III. I 1 pS
mean. Several polynomial approximations (Abra-
mowitz and Stegun, 1972) are available to calculate IV. I i pK i 8i
the value of Z. Similarly, when the assumption of no
shift in the process mean does not hold and when V. I n1 D where; I n1 On

Scrap 1 Scrap i Scrap n

O1 = I2 Oi = Ii+1 On-1=In On
Stage 1 Stage i Stage n

Rework 1 Rework i Rework n

Fig. 2. n-Stage Processes Under Consideration.

 ARTICLE IN PRESS
462 U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467
!
qi
X dik dene the stage improvement alternatives that
VI. I i1 I i dik 1  f i1 ri1k 8i
k1
VII. dik Bin0; 1 and I i 2 Z
The improvement of the process is driven by the
binary decision variables, dik in the objective
function of Model 1. That is, for each stage i, these
variables dictate by how much the current scrap rate
(dened by fi1 ) can be reduced by selecting the kth
alternative (dened by ri1k).
The initial constraint in Model 1 species that
among the qi possible improvement alternatives for
stage i, only one can be selected. It is important to
mention that among the alternatives for each stage,
the current process should be included with the
appropriate associated parameters. That is, the
model should consider the possibility that it may
not be economically viable to implement a Sigma
stage improvement alternative.
Given that most companies do not have unlimited
funds to implement process improvement alterna-
tives, there is generally a predetermined budget for
these decisions. Thus, the second constraint guar-
antees that this upper limit is not exceeded.
The next three constraints of the model are
related to the units processed at each stage. Under
the model rationale, the third constraint presumes
that there is a limit on the availability of raw
material coming into the rst production stage.
Equivalently, each stage is capacitated as described
by the fourth constraint. The fth constraint
guarantees not to build up any unnecessary
inventory at the nal stage of the process.
The sixth constraint acknowledges that the input
of each stage is a function of the input of the previous
stage and the reduction in the scrap rate obtained
from the implementation of an improvement alter-
native. Even though for each stage some potential
rework may result, under Model 1 formulation it has
been assumed that the rework is successful and the
reworked units move downstream into the next
production stage. Notice also, that this constraint
ensures balance from stage to stage (i.e. the output of
a production stage is forced to be the input of the
subsequent stage). Finally, the last constraint denes
the decision variables dik as binary and the stage
inputs to be restricted to non-negative integers.
Thus, Model 1 denes a non-linear mixed integer-
programming model in which the decision variables
may be chosen to replace the existing process and,
maximize its sigma quality level based on the yield
as a surrogate measure. Finally, it must be noted
that: (1) the reduction in the occurrence of either
rework or scrap cannot exceed its corresponding
existing value. That is, for each stage, the alter-
natives cannot degrade its current Sigma level.
Although this is not represented in Model 1, it is
used as a necessary condition prior to running the
model and, (2) The corresponding sigma level based
on the solution of Model 1 can be calculated using
(1) and (2).
Although the objective function of Model 1 is
non-linear, an equivalent linear objective function
can be used to transform Model 1 into a Mixed
Linear Integer Program (MLIP). As noted by
Ramirez-Marquez et al. (2004): An equivalent
objective function is one that has the same optimal
solution, although the objective function value is
not equal. These types of functions allow the
creation of a new problem i.e., easier to solve and
yields the optimal solution to the original problem.
The original objective function in Model 1 can be
replaced by the following equivalent functions:
!
qi
i. X X
n
ln dik 1  f i1 ri1k ,
i1 k1
qi
!
ii. X X
n
dik gi1k ; gi1k  ln1  f i1 ri1k .
i1 k1
The second equivalent function can be used
because the rst constraint of Model 1 dictates that
only one of the qi alternatives for stage i can be
chosen. Thus, based on this last equation, Model 1
can be transformed into an MILP.
3. Optimal selection of alternatives to maximize
prot in a Six Sigma project
For cases where management is only considering
the selection of alternatives to maximize the process
Sigma quality level, the optimization model stated
in Model 1 provides the best tool for guiding the
decision-making process. However, in some in-
stances the decision may not be directly related to
maximizing Sigma level, but to choose the stage
improvement alternatives that maximize the prot
generated by the process.
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467 463

Under this new scenario, this section presents a rework and scrap, respectively. Mathematically:
mathematical programming model that can be used X
i
as a management tool to solve the problem of Di I i cm
l , (4)
maximizing prot in a Six Sigma project subject l1
to a budget constraint. The model is presented in
qi
X
Model 2.
O i R i cR
i dik f i2  ri2k , (5)
Model 2: k1
b
z}|{
" a ! # qi
X
z}|{ X X qi
n
Ci I i c X
i dik f i1  ri1k . (6)
Max Profit uI n1  dik Qik Di Oi Ci
k1
i1 k1
Similar to Model 1, stage improvement alterna-
Subject to: tives are dened in terms of their costs, rework
improvement rates and scrap improvement rates
qi
X that are assumed to be known in advance. Model 2
I. dik 18i
is a MILP model for which a solution provides the
k1
values of the decision variables dik (i.e. the
qi
n X
X alternatives chosen to maximize prot). The for-
II. dik Qik pB mulations in Model 1 and 2 can be solved using
i1 k1 special purpose software such as LINDO, SOL-
VER, etc.
III. I 1 pS
4. Illustrative example
IV. I i pK i 8i
The purpose of this example is to illustrate that
V. I n1 D where; I n1 On the selection of process improvement alternatives
! play a crucial role in the successful implementation
qi
X and improvement in the Sigma level of the process.
VI. I i1 I i dik 1  f i1 ri1k 8i This example illustrate that the Six Sigma imple-
k1
mentation team should rst carryout an analysis
regarding the benets of the different alternatives
VII. dik Bin0; 1 and I i 2 Z before engaging in a particular Six Sigma alter-
native.
The objective function for Model 2 represents The process depicted in Fig. 3 is a four-stage
prot at its highest level. That is, the difference process. For this process, the current yield can be
between revenue generated at the end of the process improved by selecting among three different stage
(i.e. revenue is generated when units are sold to meet improvement alternatives for each stage. The
market demand, represented by term a) and the cost current scrap and rework rates at each stage are
consumed at each process stage (represented by shown in Table 4. Additionally, Table 4 displays the
term b). current rework, marginal production costs and
The costs considered in term b of the objective capacity for each stage. Table 5 illustrates the
function are categorized into four major cost potential reduction in scrap and rework rates of
components: quality improvement, production, re- implementing the stage improvement alternatives
work and scrap. The rst cost accounted in this along with its associated implementation cost. The
term is the quality improvement cost associated to total budget available for implementation of Six
each of the alternatives. Then, term b, accounts for Sigma is $1,500,000. Finally, the market price for
the production cost at stage i, represented by Di, as each unit is $50 and raw material supply is limited
the product of the marginal cost accrued during to 50,000 units to meet the existing process demand
each stage and the amount of items going through of 20,000 units.
it. Finally, the variables Oi, and Ci, represent the For this example, the two models presented in this
incremental savings associated with the implemen- paper have been solved using LINDO (Linear
tation of a specic stage improvement alternative in Interactive Discrete Optimizer). The optimal solution
 ARTICLE IN PRESS
464 U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467

Scrap 1 Scrap 2 Scrap 3 Scrap 4

O1 = I2 O2 = I3 O3 = I4 O4 = I5
Stage 1 Stage 2 Stage 3 Stage 4

Rework 1 Rework 2 Rework 3 Rework 4

Fig. 3. Four-Stage Process.

Table 4 Table 6
Current process rates, costs and capacities Optimal values for dik under Model 1 solution

Stage fi1 fi,2 cR

i
cm
i Ki Stage Alternativea

0a 1.0 0 1 2 3
1 0.30 0.05 1.00 2.0 100,000
2 0.08 0.10 1.50 3.0 90,000 1 0 1 0 0
3 0.20 0.05 2.00 4.0 80,000 2 0 0 1 0
4 0.10 0.10 2.50 5.0 70,000 3 0 1 0 0
4 1 0 0 0
a
Stage 0 accounts for the raw material cost needed to start the
a
rst process. Value of dik at optimal solution.

Table 5 Table 7
Reduction in scrap and rework rates and associated cost for each Optimal solution for Model 2
stage improvement alternative
Stage Alternativea Input
a
Stage Alternative
0 1 2 3
0 1 2 3
1 1 0 0 0 43,136
ri10 ri20 Qi0 ri11 ri21 Qi1 ri12 ri22 Qi2 ri13 ri23 Qi3 2 1 0 0 0 30,195
1 0 0 0 0.030 0.030 40 0.050 0.040 70 0.007 0.007 25 3 1 0 0 0 27,779
2 0 0 0 0.003 0.003 30 0.060 0.030 85 0.050 0.060 110 4 1 0 0 0 22,222
3 0 0 0 0.008 0.008 25 0.007 0.007 30 0.007 0.002 15
a
4 0 0 0 0.060 0.060 110 0.009 0.009 40 0.007 0.007 35 Value of dik at optimal solution.

a
Qik104.
for Model 1, maximizing the process yield subject to
a budget constraint provides a value of Y 52.0%
(assuming only scrap rate). The Sigma stage im-
provement alternatives resulting in the optimal defect
reduction are presented in Table 6. This table,
illustrates that alternatives one, two, one and no
action are chosen to improve stages one, two, three
and four, respectively.
The solution to Model 2 for this example is
presented in Table 7. It is interesting to note that
this solution recommends maintaining the status
quo of the current process which provides a yield of
46.4%. The yield certainly can be improved but the
prot of $400.5K for the existing processes cannot
be improved based on the proposed Six Sigma
process improvement alternatives and the allowable
budget. In other words, it is not economically
desirable to implement any of the dened Six Sigma
alternatives to reduce defects.
One of the premises of the paper is that in the
eagerness of a company to implement a Six Sigma
program little, if any, quantitative analysis is
performed to assist in the decision process on how
the implementation is to be performed (i.e. choosing
among competing alternatives). Model 1 introduces
a mathematical model consistent with a common
management approach of trying to maximize the
overall yield when making Six Sigma decisions. In
formulating Model 2, the authors intent is to show
that economics should be central in making
decisions among competing quality alternatives
and can be quantitatively measured by prot. The
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467
examples illustrate that it is possible to drive the
most revenue through a company by choosing to
implement no quality program. In this example,
yield certainly can be improved but the economic
benet does not exist. The authors recognize this is
an illustrative example constructed to prove this
point, but we believe it is an important onethe use
of economic models are an essential component
when deciding if and how quality improvement
efforts should be implemented.
5. Conclusions and future research
This paper underscores the need to perform both
economic and non-economic analyses to evaluate
the value of selecting competing quality improve-
ment alternatives. Far too often corporations, based
upon other companies or industries experience, rush
into implementing quality improvements to their
existing processes without quantifying the economic
consequences. Experience has shown that the initial,
perceived euphoria of implementing a Six Sigma
initiative, e.g., can bear little resemblance to its
resulting economic consequences. Six Sigma con-
tinues to be a predominate target to try and obtain a
competitive advantage. However, not all companies
are successful in implementing many of these
quality improvement strategies. Although many
companies attribute their success to following a
quality improvement program such as TQM and Six
Sigma, there are a signicant number of companies
that fail to gain any measurable benet after
implementing these quality strategies.
Given that quality improvements are far reaching
and potentially costly, this paper introduced and
developed two economic-based mathematical mod-
els to assist managers in making the decision on
what, if any, quality improvement alternatives to
implement. The objective of this paper is to develop
mathematical models that can be used to select the
process improvement techniques in an optimal way.
Two mathematical programming models have been
developed in the paper, one nds an optimal sigma
quality level using yield as a surrogate revenue
measure and another model maximizes the prot
from a Six Sigma project by selecting the best
processes for Six Sigma implementation.
An illustrative example is used to show how the
mathematical models developed in the paper can be
used in practice. Additional research is required to
develop models that fully capture the economic
advantages of Six Sigma, most notably increased
465
market share, apart from reduction in wastage cost.
Once the relationship between benet due to
improved quality of a manufacturing process and
a higher sigma level is established, one can modify
the optimization models presented in this paper.
References
Abramowitz, M., Stegun, I.A., 1972. Handbook of Mathematical
Functions. Dover, New York.
Anbari, F.T., 2002. Six Sigma method and its applications in
project management. In: Proceedings of the Project Manage-
ment Institute Annual Seminar and Symposium. San Anto-
nio, Texas.
Anon, 2003. Revealing study of six sigma: gains but missed
opportunities. Strategic Direction 19 (8), 3436.
Anon, 2007a. Six Sigma project saves continental Teves more
than $800,000 annually. Internet article accessed on 31st
January 2007 at http://www.ies.ncsu.edu/_library/docs/
ContinentalTeves_SixSigma.cfm?RID=172&WebPLID=1.
Anon, 2007b. Borg Warner turbo systems saves almost $1.5
million annually with Six Sigma. Internet article accessed on
31st January 2007 at /http://www.ies.ncsu.edu/_library/docs/
BorgWarner.cfm?RID=110&WebPLID=1S.
Antony, J., 2004. Six Sigma in the UK service organizations:
results from a pilot survey. Managerial Auditing Journal 19,
10061013.
Antony, J., Banuelas, R., 2002. Key ingredients for the effective
implementation of Six Sigma program. Measuring Business
Excellence 6 (2), 2027.
Antony, J., Fergusson, C., 2004. Six sigma in a software industry:
results from a pilot study. Managerial Auditing Journal 19,
10251032.
Banuelas, R., Antony, J., Brace, M., 2005. An application of Six
Sigma to reduce waste. Quality and Reliability Engineering
International 21, 553570.
Bertels, T. (Ed.), 2003. Rath and Strongs Six Sigma Leadership
Handbook. Wiley, New Jersey.
Bowman, A., Schmee, J., 2004. Estimating sensitivity of process
capability modeled by a transfer function. Journal of Quality
Technology 36 (2), 223239.
Buss, P., Ivey, N., 2001. Dow chemical design for Six Sigma rail
delivery project. In: Proceedings of the 2001 Winter Simula-
tion Conference 2001, pp. 12481251.
Chou, C.Y., Chen, C.H., 2001. On the present worth of
multivariate quality loss. International Journal of Production
Economics 70, 279288.
De Feo, J., Bar-El, Z., 2002. Creating strategic change more
efciently with a new design for Six Sigma process. Journal of
Change Management 3 (1), 6080.
Deleryd, M., 1999. A pragmatic view on process capability
studies. International Journal of Production Economics 58
(3), 319330.
Dusharme, D., 2006. Six Sigma Survey: Big Successy What
About Other 98 Percent? Quality Digest, accessed on 10th
February 2006 at /http://www.qualitydigest.com/feb03/articles/
01_article.shtmlS.
Edgeman, R.L., Bigio, D., Ferleman, T., 2005. Six sigma and
business excellence: strategic and tactical examination of IT
Service Level Management at the Ofce of the Chief of
 ARTICLE IN PRESS
466 U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467
Technology Ofcer of Washington, DC. Quality and Relia-
bility Engineering International 21, 257273.
Ehie, I., Sheu, C., 2005. Integrating Six Sigma and theory of
constraints for continuous improvement: a case study.
Journal of Manufacturing Technology Management 16 (5),
542553.
Frings, G.W., Grant, L., 2005. Who moved my Sigmay effective
implementation of the Six Sigma methodology to hospitals.
Quality and Reliability Engineering International 21, 311328.
Ganeshan, R., Kulkarni, S., Boone, T., 2001. Production
economics and process quality: a Taguchi perspective.
International Journal of Production Economics 71, 343350.
Gowen, R.C., Tallon, W.J., 2005. Effect of technological
intensity on the relationship among Six Sigma design,
electronic business, and competitive advantage: a dynamic
capability model. Journal of High Technology Management
Research 16, 5987.
Hahn, G.J., Hill, W.J., Hoerl, R.W., Zinkgraf, S.A., 1999. The
impact of Six Sigma improvementa glimpse into the future
of statistics. The American Statistician 53 (3), 208215.
Harry, M.J., 1998. Six Sigma: a breakthrough strategy for
protability. Quality Progress 31 (5), 6064.
Hendricks, C.A., Kelbaugh, R.L., 1998. Implementing Six Sigma
at GE. The Journal of Quality and Participation 21 (4), 4348.
Knowles, G., Johnson, M., Warwood, S., 2004. Medicated sweet
variabilitya six sigma application at a UK. Food Manu-
facturer 16 (4), 284292.
Krupar, J., 2003. Yes, Six Sigma can work for nancial
institutions. ABA Banking Journal 95 (9), 9394.
Kumar, U.D., Crocker, J., Chitra, T., Saranga, H., 2006.
Reliability and Six Sigma. Springer, Berlin.
Kwak, Y.H., Anbari, F.T., 2006. Benets, obstacles and future of
Six Sigma. Technovation: The International Journal of
Technological Innovation, Entrepreneurship and Technology
Management 26 (5-6), 708715.
Lanyon, S., 2003. At Raytheon Six Sigma works, too, to Improve
HR management processes. Journal of Organizational Ex-
cellence 22 (4), 2942.
Liu, E.W., 2006. Clinical Research the Six Sigma Way. Journal of
the Association for Laboratory Automation 11 (1), 4249.
McClusky, R., 2000. The rise, fall and revival of Six Sigma.
Measuring Business Excellence 4 (2), 617.
Moorman, D.W., 2005. On the quest for Six Sigma. The
American Journal of Surgery 189, 253258.
Motwani, J., Kumar, A., Antony, J., 2004. A business process
change framework for examining the implementation of Six
Sigma: a case study of Dow chemicals. TQM Magazine 16 (4),
273283.
Mukhopadhyay, A.R., Ray, S., 2006. Reduction of yarn packing
defects using Six Sigma methods: a case study. Quality
Engineering 18 (2), 189206.
Parkan, C., Wu, M.L., 2000. Comparison of three modern multi-
criteria decision making tools. International Journal of
System Science 31 (4), 497517.
Ramirez-Marquez, J.E., Coit, D., Konak, A., 2004. Reliability
optimization of series-parallel systems using a MaxMin
approach. IIE Transactions 36 (9), 891898.
Robinson, B., 2005. Build a management system based on Six
Sigma. ASQ Six Sigma Forum Magazine 5 (1), 2834.
Stuart, J.A., Ammons, J.C., Turbini, L.J., 1999. A product and
process selection model with multidisciplinary environmental
considerations. Operations Research 47 (2), 221234.
Taguchi, G., 1986. Introduction to Quality Engineering. Asian
Productivity Organization, Tokyo.
Taguchi, G., Clausing, D., 1990. Robust quality. Harvard
Business Review, 6575 (JanuaryFebruary).
Weiner, M., 2004. Six Sigma. Communication World 21 (1),
2629.
Zimmerman, J.P., Weiss, J., 2005. Six Sigmas seven deadly sins.
Quality 44 (1), 6266.
Dinesh Kumar is a Professor at Indian Institute of Management
Bangalore (IIM Bangalore). This work was carried out during his
Sabbatical at the Systems Engineering and Engineering Manage-
ment department at Stevens Institute of Technology. Dr. Dinesh
Kumar holds a Ph.D. in mathematics from Indian Institute of
Technology, Bombay and M.Sc. in Applied Mathematics from P
S G College of Technology, Coimbatore, India. Before joining
IIM Calcutta, Dr Kumar worked as a faculty member at the
Indian Institute of Management Calcutta, India; University of
Exeter, UK; and as a post-doctoral fellow at University of
Toronto, Canada; and Swiss Federal Institute of Technology,
Zurich, Switzerland. Dr. Kumar is a visiting professor at the
national graduate school of management, Australian National
University. He also served as a visiting fellow at the Queensland
University of Technology in 2004. Dr. Kumar has written two
books and more than 50 articles in leading academic journals in
the areas of systems engineering, reliability, fault tolerant
software, maintainability, logistics and supply chain manage-
ment, optimization and six sigma. Email: dineshk@iimb.ernet.in
David Nowicki is an associate research professor at Stevens
Institute of Technology. Over the last 15 years Dr. Nowicki has
held executive positions at i2 Technologies where he built i2s
European development organization and later ran their Con-
sumer Packaged Goods consulting and solution practice, at
Servigistics where he was responsible for product management
in the eld of Service Parts Management and, at Tools for
Decision where he was engaged on modeling LCC, Spares
Optimization and LORA solutions to drive new business
ventures. Dr. Nowicki is currently focused on applying advanced
analytical techniques to solve Supply Chain Management
problems from a Systems Engineering context with special
interest on Performance Based Logistics modeling and multi-
asset optimization. Additional areas of interest are in the elds of
reliability theory and spares optimization. Dr. Nowicki received
his Masters Degree in Industrial and Systems Engineering from
Virginia Tech and both his bachelors and doctorate degrees in
Industrial Engineering from the University of Wisconsin
Madison. Email: dnowicki@stevens.edu
Jose E. Ramirez-Marquez is an Assistant Professor at Stevens
Institute of Technology in the Department of Systems Engineer-
ing and Engineering Management. His research interests include
system reliability and quality assurance, uncertainty modeling,
advanced heuristics for system reliability analysis, applied
probability and statistical models and, applied operations
research. He has authored more than a dozen articles in leading
technical journals on these topics. He obtained his Ph.D. at
Rutgers University in Industrial and Systems Engineering. He
 ARTICLE IN PRESS
U.D. Kumar et al. / Int. J. Production Economics 111 (2008) 456467 467

received his B.S. degree in Actuarial Science from the UNAM in
Mexico City in 1998 and M.S. degrees in Industrial Engineering
and Statistics from Rutgers University. He is a member of IIE,
IFORS and INFORMS. Email: jmarquez@stevens.edu
Dinesh Verma is Professor in Systems Engineering and concur-
rently serves as the Associate Dean at Stevens Institute of
Technology and as the Scientic Advisory to the Director of the
Embedded Systems Institute in Eindhoven, Holland. Dr Verma
served as Technical Director at Lockheed Martin Undersea
Systems, in the area of adapted systems and supportability
engineering processes. His current research activities emphasize
systems engineering and design with a focus on conceptual design
evaluation, preliminary design and system architecture, design
decision-making, life cycle costing, and supportability engineer-
ing. Dr. Verma has authored over 85 technical papers and co-
authored two textbooks. He is a Fellow of the International
Council on Systems Engineering (INCOSE), a senior member of
SOLE, and was elected to Sigma Xi, the honorary research
society of America. He serves an a member of the External
Advisory Board on Systems Engineering at SAIC, on the Systems
Engineering Advisory Council (SEAC) of the Systems and
Software Consortium, and the Advisory Board of the
Center for Systems Engineering at the Air Force Institute of
Technology. Dr. Verma received the Ph.D. and the M.S. in
Industrial and Systems Engineering from Virginia Tech. Email:
dverma@stevens.edu