Computers & Industrial Engineering 44 (2002) 171–190

www.elsevier.com/locate/dsw

Product planning in quality function deployment using a combined

analytic network process and goal programming approachq
E. Ertugrul Karsak*, Sevin Sozer, S. Emre Alptekin
Industrial Engineering Department, Faculty of Engineering and Technology, Galatasaray University, Ciragan Cad. No: 102,
Ortakoy, Istanbul 80840, Turkey
Accepted 10 June 2002

Abstract
Quality function deployment (QFD) is a customer-oriented design tool with cross-functional team members
reaching a consensus in developing a new or improved product to increase customer satisfaction. QFD starts with
the house of quality (HOQ), which is a planning matrix translating the customer needs into measurable product
technical requirements (PTRs). A robust evaluation method should consider the interrelationships among customer
needs and PTRs while determining the importance levels of PTRs in the HOQ. This paper employs the analytic
network process (ANP) to fulfill this requirement. Furthermore, the proposed analytic procedure should take into
account the multi-objective nature of the problem, and thus, incorporate other goals such as cost, extendibility and
manufacturability of PTRs. This paper presents a zero – one goal programming methodology that includes
importance levels of PTRs derived using the ANP, cost budget, extendibility level and manufacturability level
goals to determine the PTRs to be considered in designing the product. A numerical example is presented to
illustrate the application of the decision approach. q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Quality function deployment; House of quality; Dependence; Analytic network process; Zero –one goal
programming

1. Introduction

Global competitiveness has recently become the biggest concern of both manufacturing and service
companies, which seek for higher levels of quality for their products and services and continuous
improvement to keep up with the rapid pace of development and change throughout the world. Total

q
This manuscript was processed by E.A. Elsayed.
* Corresponding author. Fax: þ90-212-259-5557.
E-mail address: karsak@gsunv.gsu.edu.tr (E.E. Karsak).

0360-8352/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 6 0 - 8 3 5 2 ( 0 2 ) 0 0 1 9 1 - 2
172 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

quality management, which offers a vast selection of techniques to ensure the improvement of quality
and productivity, has been a topic on the research agenda for the last four decades.
Quality function deployment (QFD) is one of these techniques that aim the satisfaction of the
customer at the very beginning, namely the product design phase. It enables the companies to
become proactive to quality problems rather than taking a reactive position by acting on customer
complaints. As an interdisciplinary team process, QFD is used to plan and design new or improved
products or services. QFD employs a cross-functional team to determine customer needs and
translate them into product designs through a structured and well-documented framework.
QFD helps the companies to maintain their competitiveness using three strategies: decreasing
costs, increasing revenues, and reducing the time to produce new products or services (cycle time
reduction). QFD allows for the company to allocate resources and to coordinate skills and functions
based on customer needs, and thus, may result in lower production costs by ignoring aspects
meaning little or nothing to the customer. Its systematic nature also evaluates the necessary
decisions for change and development at the beginning of the design process, reducing and even
avoiding the mid-project changes and corrections. Enabling to develop the right product or service
for the customers with the lowest possible cost, QFD attracts the customers, which results in higher
selling rates, leading to higher revenues. In this way, QFD facilitates the entire development
process, minimizing the corrections and waste during this phase, and as a matter of fact, optimizing
the time required for introducing a new or improved product or service to the market.
After World War II, the concept of product development evolved from copying and imitation to
a product development based on originality. The importance of design quality became apparent.
This dramatic change entailed the development of a totally new concept, the QFD. QFD was first
conceptualized in the late 1960s (Akao, 1997). It was immediately adapted by various companies
but it did not draw much public attention. A few years later, in 1972, QFD was implemented at the
Kobe shipyards of Mitsubishi Heavy Industries Ltd. Even though its application was followed by
successful implementations throughout Japan, e.g. at Toyota, it remained a Japanese tool until the
early 1980s. Following the article by Kogure and Akao (1983) and through Ford Motor Company
and the Cambridge Corporation, QFD has entered the borders of the US and has started to play an
important role at companies such as General Motors, Chrysler, Digital Equipment, Hewlett-Packard,
AT&T, Procter and Gamble, and Baxter Healthcare (Prasad, 1998).
There are two major organizations as sources of QFD, namely the American Supplier Institute
(ASI) and GOAL/QPC that both developed their own models having many similarities to each
other. The ASI employs a basic four-matrix method developed by Macabe, a Japanese reliability
engineer, while GOAL/QPC uses a multiple matrix developed by Akao that incorporates many
disciplines into a less structured format consisting of a matrix of matrices (Shillito, 1994).
The basic concept of QFD is to translate the desires of customers, i.e. the voice of customer, into
product technical requirements (PTRs) or engineering characteristics, and subsequently into parts
characteristics, process plans and production requirements. In order to establish these relationships QFD
usually requires four matrices: product planning, parts planning, process planning, and production
planning matrices, respectively. Product planning matrix translates customer needs into product design
requirements; part planning matrix translates important design requirements into product/part
characteristics; process planning matrix translates important product/part characteristics into
manufacturing operations; production/operation planning matrix translates important manufacturing
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 173

operations into day-to-day operations and controls. In this paper, we focus on the first of the four
matrices, also called the house of quality (HOQ).
There has been some research on quantifying the planning issues in HOQ within the past decade,
mainly focusing on customer needs. Chan, Kao, Ng, and Wu (1999) and Khoo and Ho (1996) employ
fuzzy set theory for rating the customer needs. Other researchers use the analytic hierarchy process
(AHP) to determine the degree of importance of the customer needs (Armacost, Componation, Mullens,
& Swart, 1994; Lu, Madu, Kuei, & Winokur, 1994; Park & Kim, 1998).
In this paper, we propose a novel approach for determining the PTRs that will be considered in
designing the product by integrating two decision making techniques, namely the analytic network
process (ANP) and zero– one goal programming (ZOGP). The ANP is a decision aid to incorporate the
dependence issues into the analysis. Hence it enables us to take into account the degree of the
interdependence between the customer needs and PTRs, and the inner dependence among them.
Although the ANP facilitates the analysis involving dependence of PTRs on customer needs, inner
dependence within customer needs and inner dependence within PTRs, it falls short of taking into
account resource limitations and other metrics required in the determination of the PTRs for product
design. Wasserman (1993) develops a linear programming model for maximizing customer satisfaction
subject to a budget constraint in QFD planning process. Considering the multi-objective nature of the
design problem and having calculated the final relative importance of the PTRs via the ANP, we use the
highest possible consideration of these PTRs in the design phase as a goal to be satisfied along with other
goals such as cost budget, extendibility, and manufacturability of the PTRs. Extendibility incorporates to
what degree improvements in one PTR can be extended to other PTRs, and thus, enables a PTR with a
high cost but also with high extendibility to be considered as worthwhile in product improvement.
Manufacturability highlights the difficulties, which are efforts required to implement the desired
improvement. For instance, one PTR could require a unique technology for improvement, whereas
another PTR could be easily improved with the present technology. The relative importance weights of
these goals are determined by pairwise comparisons. The ZOGP model is solved to determine the PTRs,
which will be taken into account in the design phase, in a way to minimize the deviations from the
prioritized goals.
The paper is organized in the following order. In Section 2, we present a brief description of the HOQ
with a concise treatment of its steps. Section 3 describes the basics of the ANP and presents the
supermatrix approach developed by Saaty (Saaty, 1996; Saaty & Takizawa, 1986). In Section 4, the
decision methodology to determine the PTRs to be considered in product design is presented. In Section
5, the design of a pencil is selected as an example to apply the developed procedure. Section 6 provides
the concluding remarks.

2. The house of quality

In today’s competitive environment, the HOQ is a key strategic tool to aid companies in developing
products that satisfy customer needs. According to Hauser and Clausing (1988), the HOQ is a kind of
conceptual map that provides the means for interfunctional planning and communications. The seven
elements of the HOQ shown in Fig. 1 can be described as follows.
(1) Customer needs (WHATs). They are also known as voice of the customer, customer attributes,
customer requirements or demanded quality. As the initial input for the HOQ, they highlight the product
174 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Fig. 1. House of quality.

characteristics that should be paid attention to. Customer needs, usually collected by focus groups or
individual interviews should be expressed in customers’ own phrases. According to Griffin and Hauser
(1993), individual one-to-one interviews may be more cost effective than focus groups, and at least 20 –
30 customers should be interviewed to obtain 90– 95% of all the possible customer needs. Mail or
telephone surveys should be avoided due to the difficulties in controlling the scope of responses.
Preserving customers’ own words usually causes problems during the phase of translation and
interpretation since they are usually too general and/or detailed to be directly used as customer needs. To
overcome this problem, a number of approaches are used. Initially the words are collected, and then they
are organized to form a tree-like hierarchy usually with three or four levels. Those at the appropriate
level are chosen as the final customer needs. Affinity diagram, which is a method used to gather large
amounts of qualitative data and to organize them into subgroupings based on the similarities between
them, can be used for this purpose (Cohen, 1995). Cluster analysis can also be used to form and structure
customer needs (Griffin & Hauser, 1993).
(2) PTRs (HOWs). PTRs are also known as design requirements, product features, engineering
attributes, engineering characteristics or substitute quality characteristics. They can also be developed
using the affinity diagram and tree diagram. They describe the product in the language of the engineer;
therefore, are sometimes referred to as the voice of the company. The PTRs are used to determine how
well the company satisfies the customer needs. Customer needs tell the company ‘what to do’ while the
PTRs tell ‘how to do it’. They must be stated in measurable and benchmarkable terms, e.g. for a pencil,
lead dust generated in milligrams.
(3) Relative importance of the customer needs. Since the collected and organized data from the
customer usually contain too many needs to deal with simultaneously, they must be rated. The company
should trade off one benefit against another, and work on the most important needs while disregarding
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 175

relatively unimportant ones. In this manner, customers are surveyed for each WHAT using 5-, 7- or 9-
point scales. More detailed 1-to-10 and anchored scale can also be used. In contrast to determination of
customer needs, importance of customer needs are obtained through mail or telephone surveys. Focus
groups and individual interviews are not appropriate since a large number of people should be surveyed
with a higher cost prospect due to statistical significance concerns.
(4) Relationships between WHATs and HOWs. The relationship matrix indicates how much each PTR
affects each customer need. The relations can either be presented in numbers or symbols. In this paper,
we will use numbers to denote the relationship between WHATs and HOWs.
(5) Inner dependence among the customer needs. In general, customer needs have inner dependence
among them. Some of them will support each other whereas others will adversely affect the achievement
of others. These supporting and conflicting needs can be identified by a correlation matrix emphasizing
necessary trade-offs.
(6) Inner dependence among the PTRs. The HOQ’s roof matrix is used to specify the various PTRs
that have to be improved collaterally, providing a basis to calculate to what extent a change in one
feature will affect other features. A desirable change in one feature may result in a negative effect on
another feature. This correlation facilitates the necessary engineering impacts and trade-offs.
(7) Overall priorities of the PTRs and additional goals. Here the results obtained from preceding steps
are used to calculate a final rank order of HOWs, also called PTR ratings. Additional design metrics such
as cost, extendibility, manufacturability, etc. can also be incorporated into the analysis at this step
(Shillito, 1994). These metrics help in determining priorities and directions for improvement, as well as
providing an objective means of assuring that requirements have been met.
A structured communication device, the HOQ, is built using the seven elements mentioned above.
With its design-oriented nature, the HOQ serves not only as a valuable resource for designers but also as
a way to summarize and convert feedback from customers into information for engineers. In addition,
marketing can benefit from it since it is based on the voice of customer, and upper management can use it
to develop strategic opportunities. Hence, the HOQ strengthens vertical and horizontal communications.
Once having identified critical PTRs that demand change, they will be driven to the next matrix as
WHATs to identify the critical parts characteristics. Similarly, the manufacturing operations, and day-
to-day operations and controls are defined. Thus finally, the company has a product fulfilling both
customer needs and producer requirements within a shorter development time in its hands.

3. The analytic network process

The ANP generalizes a widely used multi-criteria decision making tool, the AHP, by replacing
hierarchies with networks. The AHP is a well-known technique that decomposes a problem into several
levels in such a way that they form a hierarchy (Saaty, 1980). Each element in the hierarchy is supposed
to be independent, and a relative ratio scale of measurement is derived from pairwise comparisons of the
elements in a level of the hierarchy with respect to an element of the preceding level. However, in many
cases, there is interdependence among criteria and alternatives. The ANP can be used as an effective
tool in those cases where the interactions among the elements of a system form a network structure
(Saaty, 1996).
While AHP employs a unidirectional hierarchical relationship among decision levels, ANP
enables interrelationships among the decision levels and attributes to be taken into consideration
176 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

in a more general form. ANP uses ratio scale measurements based on pairwise comparisons;
however, it does not impose a strict hierarchical structure as in AHP, and models a decision
problem using a systems-with-feedback approach. Fig. 2a and b shows the structural difference
between the hierarchy and network. Nodes of the network represent components of the system,
and arcs denote interactions between them. The directions of the arcs represent dependence,
whereas loops signify inner dependence of the elements in a cluster. As we can observe, a
hierarchy is a simple and special case of a network.
In ANP, the relative importance values are determined similar to AHP using pairwise comparisons
with a scale of 1 –9, where a score of 1 indicates equal importance between the two elements and 9
represents the extreme importance of one element compared to the other one. The relations aji ¼ 1=aij ;
where aij denotes the importance of the ith element compared to the jth element, and aii ¼ 1 are
preserved in the pairwise comparison matrix to improve the consistency of the judgments.
From a general point of view, the ANP consists of two stages: the first one is the construction of
the network, and the second one is the calculation of the priorities of the elements. In order to construct
the structure of the problem, all of the interactions among the elements should be considered. When the
elements of a component Y depend on another component X, we represent this relation with an arrow
from component X to Y. All of these relations are evaluated by pairwise comparisons and a supermatrix,
which is a matrix of influence among the elements, is obtained by these priority vectors. The supermatrix
is raised to limiting powers to calculate the overall priorities, and thus the cumulative influence of each
element on every other element with which it interacts is obtained (Saaty & Vargas, 1998). The
supermatrix of a hierarchy with three levels is as follows

G C A
0 1
GoalðGÞ 0 0 0
W¼ B C ð1Þ
CriteriaðCÞ B w21 0 0C
@ A
AlternativesðAÞ 0 W32 I

where w21 is a vector that represents the impact of the goal on the criteria, W32 is a matrix that represents
the impact of the criteria on each of the alternatives, and I is the identity matrix. The supermatrix of a

Fig. 2. (a) A hierarchy. (b) A nonlinear network.

 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 177

system of N clusters is denoted as follows

C1 ··· Ck ··· CN
e11 · · ·e1n1 ··· ek1 · · ·eknk ··· eN1 · · ·eNnN
e11
0 1
..
. B C
C1 BW W1k ··· W1N C
B 11 ··· C
B C
e1n1 B C
B C
B C
.. .. B . .. .. .. C
. B . C
. B . . . . C
B C
B C
W¼ ek1 B C ð2Þ
B C
B C
.. B C
Ck . B Wk1 ··· Wkk ··· WkN C
B C
B C
B C
eknk B C
B C
.. B .. .. .. .. C
. .. B . . . . C
. B C
B C
B C
B C
eN1 B C
B ··· C
CN B WN1 WNk ··· WNN C
.. @ A
.
eNnN
where Ck is the kth cluster ðk ¼ 1; 2; …; NÞ; which has nk elements denoted as ek1 ; ek2 ; …; eknk : A ratio
scale priority vector obtained by pairwise comparisons indicates the impact of a cluster’s elements on the
elements of another cluster. These vectors are located in appropriate positions. Since W is a column
stochastic matrix, its limiting priorities depend on the reducibility, primitivity, and cyclicity of that
matrix. For instance, if the matrix is irreducible and primitive, the limiting value is obtained by raising
W to powers (Saaty, 1996; Saaty & Vargas, 1998).
When a network consists of only two clusters apart from the goal, namely criteria and alternatives, the
matrix manipulation approach proposed by Saaty and Takizawa (1986) can be employed to deal with
dependence of the elements of a system. This approach, which will be described in Section 5, is used
herein to incorporate the dependencies inherent in QFD process into the analysis.

4. The decision methodology

The decision algorithm addresses the problem of selecting the PTRs on which to focus in the design
process considering the predetermined goals. The algorithm can be divided into two major phases. In the
first phase, we construct the HOQ using the ANP approach, and in the second phase, we integrate the
ANP results with a ZOGP model to determine the set of PTRs that the design team needs to concentrate
on. The ANP approach, which allows for modeling interrelationships within the HOQ, is employed to
determine the representation of the voice of the customer in the HOQ, considered as one of the goals
178 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

within the decision framework. There are other goals such as cost budget, extendibility, and
manufacturability that need to be taken into account in the analysis. The unit budget costs with respect to
PTRs are determined, and the priorities of the PTRs with respect to goals such as extendibility and
manufacturability are obtained using pairwise comparisons. Following that, the adjusted priorities of the
PTRs with respect to these goals are calculated to account for dependencies encountered in the HOQ.
Then, the relative importance weights of the goals that are taken into consideration are computed
employing pairwise comparisons. Finally, all the previously calculated data are integrated within a
ZOGP formulation to determine the PTRs to be considered in the design process. Fig. 3 denotes the
stepwise representation of this algorithm.
We prefer using weighted ZOGP as a decision tool since it can handle multiple objectives and
seeks to minimize the total deviation from the desired goals. This property of ZOGP enables us
to incorporate multiple goals including cost budget, extendibility, and manufacturability into the
product design process. The weighted goal programming model considers all the goals
simultaneously by forming an achievement function that minimizes the total weighted deviation
from all the goals stated in the model. The weights are not preemptive but reflect the decision
makers’ preferences regarding the relative importance of each goal. The incommensurability issue
faced in weighted goal programming when employing goals that are measured using different
units such as cost budget goal and extendibility goal can be resolved using a normalization
scheme (Schniederjans, 1995).
In formulating goal programming models that include multiple qualitative goals, a method based on
pairwise comparisons such as AHP or ANP appears as an effective means for assessing relative weights.
Numerous studies in diverse areas concerning integrated AHP and ZOGP models can be noted in the
literature, reaching to the conclusion that combined models provide more realistic solutions by avoiding
infeasibility (Badri, 1999; Schniederjans & Garvin, 1997). ANP enables modeling more complex
relationships which include dependence among decision levels, and thus, will be employed in the
integrated decision approach proposed in this paper.
The first step of the algorithm is the identification of the customer needs and PTRs. Then, the
determination of the importance of the customer needs, which corresponds to the first step of the matrix
manipulation concept of the ANP, follows (Lee & Kim, 2000; Saaty & Takizawa, 1986). Next, we fill
the body of the house by the weights obtained through comparing the PTRs with respect to each
customer need, and then, obtaining the interdependent priorities of the PTRs by analyzing dependence
among the customer needs and PTRs, respectively.
The supermatrix representation of the QFD model used in this paper is as follows:

G CNs PTRs
0 1
GoalðGÞ 0 0 0
W¼ B C ð3Þ
Customer NeedsðCNsÞ B w1 W3 0 C
@ A
PTRs 0 W2 W4

where w1 is a vector that represents the impact of the goal, namely manufacturing a product that satisfies
the customers, on the customer needs, W2 is a matrix that denotes the impact of the customer needs on
each of the PTRs, W3 and W4 are the matrices that represent the inner dependence of the customer needs
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 179

Fig. 3. Stepwise representation of the algorithm to determine the set of PTRs to be considered in designing a product.

and the inner dependence of the PTRs, respectively. Here, ‘impact’ denotes the potential of the PTRs to
satisfy the requirements implicit in each of the customer needs, and in a similar way for the customer
needs in terms of the goal.
The network representation of the QFD model is depicted in Fig. 4. This is the case of a hierarchy with
inner dependence within components and no feedback. The customer needs correspond to criteria
whereas the PTRs correspond to alternatives in standard ANP terms, both of which have inner
dependence within themselves; however, there is no feedback, i.e. the customer needs do not depend on
the PTRs.
Using the notation given above, the interdependent priorities of the customer needs (wC) are computed
by multiplying W3 by w1, and similarly the interdependent priorities of the PTRs (WA) are obtained by
multiplying W4 by W2. Then, the overall priorities of the PTRs (w ANP), which reflect the
interrelationships within the HOQ, are calculated by multiplying WA and wC.
In order to determine the set of PTRs that will be considered in product design, we construct a ZOGP
model using the first phase results and goals related to other metrics of the PTRs such as cost,
extendibility, and manufacturability. The set of metrics contains two types of specifications: metrics that
have some sort of resource limitations, e.g. cost, and the second type metrics that result in a rate of
preference for PTRs such as extendibility and manufacturability. In order to incorporate the second type
metrics into the formulation, we determine a preference rating for each PTR using pairwise comparisons,
and we penalize the negative deviation of these metrics from 1 in the objective function. Following
El-Gayar and Leung (2001) to rectify the likely incommensurability problem related to individual goals,
180 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Fig. 4. The network representation of the QFD model.

the general form of the ZOGP model employed in the decision framework is as follows
!
X
s
di2 dþ X
m
min vANP
1 ðd12 Þ þ vi þ i þ vi ðdi2 Þ
i¼2
Ri Ri i¼sþ1

s.t.
X
n
wANP
j xj þ d12 2 d1þ ¼ 1
j¼1

X
n
rij xj þ di2 2 diþ ¼ Ri ; i ¼ 2; …; s
j¼1 ð4Þ
X
n
wij xj þ di2 2 diþ ¼ 1; i ¼ s þ 1; …; m
j¼1

xj [ {0; 1}; j ¼ 1; …; n di2 ; diþ $ 0; i ¼ 1; …; m

where the vi are the weights of the goals ði ¼ 1; 2; …; mÞ; di2 and diþ represent the negative and positive
deviation variables of the ith goal ði ¼ 1; …; mÞ; xj is the binary selection variable representing the jth
PTR ðj ¼ 1; …; nÞ; wANP
j denotes the interdependent priority of the jth PTR ðj ¼ 1; …; nÞ; rij indicates the
amount of the ith resource used by the jth PTR (i ¼ 2; …; s; j ¼ 1; …; n) Ri represents the limitation of
the ith resource, and wij is the preference rating of the jth PTR with respect to the metric i
(i ¼ s þ 1; …; m; j ¼ 1; …; n).

5. Illustrative example

Shillito (1994) applies the QFD process for the design of a hypothetical writing instrument. This
section presents an illustration of the developed decision procedure based on the pencil design example
by Shillito. The algorithm presented in Section 4 is applied to determine the PTRs to be considered in
designing a pencil. The application is demonstrated in stepwise form as given below. Due to space
limitations, only a limited number of pairwise comparison matrices are presented, and the resulting
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 181

weight vectors are generally provided without the corresponding pairwise comparison matrices in the
illustration.
Step 1. As mentioned in Section 2, the QFD product planning process begins with the determination of
the customer needs and then the PTRs. The collected and organized customer phrases are placed in the
upper left part of the HOQ. In our example, the five customer needs in designing a pencil that are
determined as the most important according to a market survey are ‘Easy to Hold’, ‘Does not Smear’,
‘Point Lasts’, ‘Does not Roll’, and ‘Easy to Erase’. Having agreed upon the customer needs, the PTRs
that are likely to affect those needs are identified as ‘Length’, ‘Time Between Sharpenings’, ‘Lead Dust
Generated’, ‘Incline Angle to Roll’, ‘Pages per Pencil’, ‘Pressure Cycles to Erase’, and ‘Eraser Dust
Generated’.
Step 2. In this step, we initially determine the relative importance of the customer needs by asking the
following question: ‘Which customer need should be emphasized more in designing a pencil, and how
much more?’. Assuming that there is no dependence among the customer needs, the following
eigenvector for the customer needs is obtained by performing pairwise comparisons with respect to the
goal of achieving the best design.
0 1 0 1
Easy to Hold 0:073
B C B C
B Does not Smear C B 0:281 C
B C B C
B C B C
B C B C
w1 ¼ B Point Lasts C ¼ B 0:467 C:
B C B C
B C B C
B Does not Roll C B 0:065 C
@ A @ A
Easy to Erase 0:114

Then, assuming that there is no dependence among the PTRs, they are compared with respect to each
customer need yielding the column eigenvectors regarding each customer need. For example, one of the
possible questions for determining the degree of relative importance of the PTRs for Does not Smear can be
as follows: ‘What is the relative importance of Lead Dust Generated when compared to Time Between
Sharpenings with respect to the customer need Does not Smear?’, yielding 3 as represented in Table 1. The
degree of relative importance of the PTRs for the remaining customer needs are calculated in a similar way
and presented in Table 2. The transpose of the data shown in Table 2 will be placed in the body of the HOQ.
Following that, the inner dependence among the customer needs are determined through analyzing
the impact of each customer need on other customer needs by using pairwise comparisons. A possible
question is as follows: ‘What is the relative importance of Does not Smear when compared to Easy to
Hold on controlling Point Lasts?’, resulting in 3 as denoted in Table 3. The schematic representation of
Table 1
Relative importance of the PTRs for Does not Smear

Does not Smear Time Between Lead Dust Pages per Pressure Cycles Relative importance
Sharpenings Generated Pencil to Erase weights

Time Between Sharpenings 1 1/3 1/3 3 0.143

Lead Dust Generated 3 1 2 6 0.462
Pages per Pencil 3 1/2 1 7 0.339
Pressure Cycles to Erase 1/3 1/6 1/7 1 0.056
182 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Table 2
The column eigenvectors with respect to each customer need
W2 Easy to Hold Does not Smear Point Lasts Does not Roll Easy to Erase
(ETH) (DNS) (PL) (DNR) (ETE)

Length 1.000 0.000 0.053 0.000 0.000

Time Between Sharpenings 0.000 0.143 0.472 0.000 0.000
Lead Dust Generated 0.000 0.462 0.141 0.000 0.000
Incline Angle to Roll 0.000 0.000 0.000 1.000 0.000
Pages per Pencil 0.000 0.339 0.334 0.000 0.051
Pressure Cycles to Erase 0.000 0.056 0.000 0.000 0.582
Eraser Dust Generated 0.000 0.000 0.000 0.000 0.367

Table 3
The inner dependence matrix of the customer needs with respect to Point Lasts (other customer needs which do not have an
impact on Point Lasts are not included in comparison matrix)

Point Lasts Easy to Hold Does not Smear Point Lasts Relative importance weights

Easy to Hold 1 1/3 1/5 0.109

Does not Smear 3 1 1/2 0.309
Point Lasts 5 2 1 0.582

the relationship between the customer needs is shown in Fig. 5. The resulting eigenvectors obtained from
pairwise comparisons are presented in Table 4, where zeros are assigned to the eigenvector weights of
the customer needs that are independent.
Next, we deal with the dependence among the PTRs. As previously accomplished for customer needs,
the inner dependencies are determined and required pairwise comparisons are performed. We utilize
questions such as ‘What is the relative importance of Lead Dust Generated when compared to Length on
controlling Time Between Sharpenings?’, yielding 3 as shown in Table 5. The dependencies among the
PTRs are depicted in Fig. 6 and the relative importance weights obtained from the pairwise comparisons
are presented in Table 6.
Afterwards, we obtain the interdependent priorities of the customer needs (wC) as
0 1
0:106
B C
B 0:374 C
B C
B C
B C
wC ¼ W3 £ w1 ¼ B 0:376 C:
B C
B C
B 0:083 C
@ A
0:061

Fig. 5. The inner dependence among the customer needs.

 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 183

Table 4
The inner dependence matrix of the customer needs

W3 Easy to Hold Does not Smear Point Lasts Does not Roll Easy to Erase

Easy to Hold 0.750 0.000 0.109 0.000 0.000

Does not Smear 0.000 0.750 0.309 0.000 0.163
Point Lasts 0.000 0.250 0.582 0.000 0.297
Does not Roll 0.250 0.000 0.000 1.000 0.000
Easy to Erase 0.000 0.000 0.000 0.000 0.540

Then, the interdependent priorities of the PTRs, WA, are calculated as follows:
0 1
0:648 0:089 0:139 0:000 0:012
B C
B 0:122 0:152 0:301 0:000 0:056 C
B C
B C
B C
B 0:000 0:401 0:233 0:000 0:005 C
B C
B C
WA ¼ W4 £ W2 ¼ B 0:000 0:000 0:000 1:000 0:000 C
B
C:
B C
B 0:230 0:280 0:268 0:000 0:152 C
B C
B C
B C
B 0:000 0:066 0:047 0:000 0:452 C
@ A
0:000 0:012 0:012 0:000 0:323
Following that, overall priorities of the PTRs (w ANP), reflecting the interrelationships within the HOQ,
are obtained by multiplying WA and wC.
0 1 0 1
Length 0:155
B C B C
B Time Between Sharpenings C B 0:186 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 0:238 C
B C B C
B C B C
w ANP B C
¼ WA £ wC ¼ B Incline Angle to Roll C ¼ B 0:083 C B
C:
B C B C
B C B 0:239 C
B Pages per Pencil C B C
B C B C
B C B C
B Pressure Cycles to Erase C B 0:070 C
@ A @ A
Eraser Dust Generated 0:029
The ANP analysis results indicate that the most important design feature is the Pages per Pencil with a
relative importance value of 0.239, which is slightly more important than Lead Dust Generated. Time
Between Sharpenings with a relative weight of 0.186, and Length with a relative weight of 0.155 follow
the most important PTRs. Eraser Dust Generated is the least important design feature according to the
ANP analysis.
If the AHP approach were employed, Time Between Sharpenings would appear as the most important
design feature with a relative importance value of 0.261, followed by Pages per Pencil with a relative
importance value of 0.257. However, when interdependencies are incorporated into the analysis by
applying ANP, Time Between Sharpenings declines to the third place in relative importance value
184 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Table 5
The inner dependence matrix of PTRs with respect to Time Between Sharpenings (Incline Angle to Roll which does not have an
impact on Time Between Sharpenings is not included in comparison matrix)

Time Between Sharpenings Length Time Between Lead Dust Pages Per Pressure Cycles Relative importance
Sharpenings Generated Pencil to Erase weights

Length 1 1/9 1/3 1/3 1/2 0.051

Time Between Sharpenings 9 1 3 3 9 0.530
Lead Dust Generated 3 1/3 1 2 3 0.203
Pages per Pencil 3 1/3 1/2 1 2 0.142
Pressure Cycles to Erase 2 1/9 1/3 1/2 1 0.074

Table 6
The inner dependence matrix of the PTRs

W4 Length Time Between Lead Dust Incline Angle Pages per Pressure Cycles Eraser Dust
Sharpenings Generated to Roll Pencil to Erase Generated

Length 0.648 0.051 0.000 0.000 0.241 0.000 0.000

Time Between Sharpenings 0.122 0.530 0.081 0.000 0.099 0.088 0.000
Lead Dust Generated 0.000 0.203 0.731 0.000 0.102 0.000 0.000
Incline Angle to Roll 0.000 0.000 0.000 1.000 0.000 0.000 0.000
Pages per Pencil 0.230 0.142 0.188 0.000 0.486 0.139 0.125
Pressure Cycles to Erase 0.000 0.074 0.000 0.000 0.036 0.773 0.000
Eraser Dust Generated 0.000 0.000 0.000 0.000 0.036 0.000 0.875

whereas Pages per Pencil ranks in the first place. Consequently, the comparative treatment of AHP and
ANP importance values for the PTRs denote that results of the analysis incorporating the
interdependencies differ from the relative importance vector obtained by applying AHP. As for the
pencil design example, Time Between Sharpenings is the most negatively influenced design feature by
the interdependency adjustment.
Step 3. This step enables us to include the goals related to resource limitations in the decision
process. Cost budget is the only resource limitation for the pencil design example. Hence, we
determine the unit cost of PTRs, and set the budget limit per pencil as 7¢. Table 7 presents the unit
cost of PTRs.
Considering the budget limit of 7¢ per pencil, the ANP solution yields Pages per Pencil, Lead

Fig. 6. The inner dependence among the PTRs.

 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 185

Dust Generated, Time Between Sharpenings, Length and Incline Angle to Roll, which are higher
ranking PTRs in terms of relative importance weights, as the PTRs to be focused on while
designing the pencil. On the other hand, if we had ignored interdependency issues in the HOQ and
applied AHP under budget limit of 7¢ per pencil, we would have considered Time Between
Sharpenings, Pages per Pencil, Lead Dust Generated and Length, namely the four PTRs with the
highest relative importance weights, in the design phase. The comparative results for ANP and
AHP approaches under the budget limit are provided in Table 8. Table 8 indicates that the ANP
solution yields a 0.1¢ unused budget resource per pencil while the AHP solution results in a 0.6¢
unused budget resource per pencil.
Step 4. This step allows for taking into account additional goals, which we name here as the second
type metrics, for determining the set of the PTRs to be focused on while designing a pencil. Extendibility
and manufacturability are selected as the second type metrics to be considered in determining the set of
PTRs for pencil design. In the context of the pencil design example, extendibility indicates to what
degree improvements in one PTR can be extended to other pencil types, while a PTR that does not
demand a change in current technology in pencil design represents higher manufacturability, which
decreases with the amount of effort required for an improvement in design. In order to incorporate
extendibility and manufacturability into the goal programming model, the PTRs should be rated with
respect to each metric using pairwise comparisons. The weight vector of extendibility (w E) and the
weight vector of manufacturability (w M) calculated by pairwise comparisons are given below.

0 1 0 1
Length 0:024
B C B C
B Time Between Sharpenings C B 0:175 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 0:128 C
B C B C
B C B C
wE ¼ B
B Incline Angle to Roll C ¼ B 0:035 C
C B C
B C B C
B C B C
B Pages per Pencil C B 0:207 C
B C B C
B C B C
B Pressure Cycles to Erase C B 0:098 C
@ A @ A
Eraser Dust Generated 0:333

Table 7
The unit cost of the PTRs

PTRs Cost in ¢ (c)

Length 0.6
Time Between Sharpenings 1.4
Lead Dust Generated 2.0
Incline Angle to Roll 0.5
Pages per Pencil 2.4
Pressure Cycles to Erase 1.7
Eraser Dust Generated 0.6
186 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Table 8
Comparison of the ANP solution and the AHP solution under budget limit

PTRs ANP AHP

Relative importance Selected design Relative importance Selected design

weights feature? weights feature?

Length 0.155 Yes 0.097 Yes

Time Between Sharpenings 0.186 Yes 0.261 Yes
Lead Dust Generated 0.238 Yes 0.196 Yes
Incline Angle to Roll 0.083 Yes 0.065 No
Pages per Pencil 0.239 Yes 0.257 Yes
Pressure Cycles to Erase 0.070 No 0.082 No
Eraser Dust Generated 0.029 No 0.042 No

0 1 0 1
Length 0:045
B C B C
B Time Between Sharpenings C B 0:104 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 0:218 C
B C B C
B C B C
wM ¼ B
B Incline Angle to Roll C ¼ B 0:022 C:
C B C
B C B C
B C B C
B Pages per Pencil C B 0:277 C
B C B C
B C B C
B Pressure Cycles to Erase C B 0:124 C
@ A @ A
Eraser Dust Generated 0:210

Step 5. The weight vectors of PTRs for cost budget, extendibility, and manufacturability metrics need
to reflect dependencies among the PTRs that are denoted in the roof of the HOQ. In order to incorporate
the adjustment for dependencies, the inner dependence matrix of the PTRs (W4) is multiplied with the
unit cost vector of the PTRs (c), resulting in the adjusted unit cost vector (c0 ) given below.
0 1 0 1
Length 1:039
B C B C
B Time Between Sharpenings C B 1:364 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 1:991 C
B C B C
B C B C
0 B C B
c ¼ W4 £ c ¼ B Incline Angle to Roll C ¼ B 0:500 CC:
B C B C
B C B 2:191 C
B Pages per Pencil C B C
B C B C
B C B C
B Pressure Cycles to Erase C B 1:504 C
@ A @ A
Eraser Dust Generated 0:611

Similarly, in order to account for dependencies among the PTRs, the inner dependence matrix of the
PTRs (W4) is multiplied, respectively, with the weight vector of extendibility (w E) and the weight vector
of manufacturability (w M) to obtain the adjusted weight vectors for extendibility and manufacturability
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 187

as follows:
0 1 0 1
Length 0:074
B C B C
B Time Between Sharpenings C B 0:135 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 0:150 C
B C B C
0 B C B C
wE ¼ W 4 £ wE ¼ B
B Incline Angle to Roll C ¼ B 0:035 C
C B C
B C B C
B C B C
B Pages per Pencil C B 0:210 C
B C B C
B C B C
B Pressure Cycles to Erase C B 0:096 C
@ A @ A
Eraser Dust Generated 0:298
0 1 0 1
Length 0:101
B C B C
B Time Between Sharpenings C B 0:116 C
B C B C
B C B C
B C B C
B Lead Dust Generated C B 0:208 C
B C B C
0 B C B C
wM ¼ W4 £ wM ¼ B
B Incline Angle to Roll C ¼ B 0:022 C:
C B C
B C B C
B C B C
B Pages per Pencil C B 0:244 C
B C B C
B C B C
B Pressure Cycles to Erase C B 0:114 C
@ A @ A
Eraser Dust Generated 0:194

The HOQ obtained using all the data from the previous steps is depicted in Fig. 7.
Step 6. In this step, we determine the weights of the goals considered in the design of a pencil by

Fig. 7. House of quality for the design of a pencil.

188 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Table 9
Relative importance weights of the goals considered in pencil design

ANP Cost budget Extendibility Manufacturability Relative importance weights (v )

ANP 1 1/3 2 3 0.239

Cost budget 3 1 3 5 0.522
Extendibility 1/2 1/3 1 2 0.153
Manufacturability 1/3 1/5 1/2 1 0.086

employing pairwise comparisons. The pairwise comparison matrix and the resulting weight vector (v )
are given in Table 9.
Step 7. We construct the ZOGP model employing Eq. (4) and using the data obtained from the
previous steps as follows:
min 0:239d12 þ ð0:522=7Þd2þ þ 0:153d32 þ 0:086d42
s.t.

0:155x1 þ 0:186x2 þ 0:238x3 þ 0:083x4 þ 0:239x5 þ 0:070x6 þ 0:029x7 þ d12 2 d1þ ¼ 1

1:039x1 þ 1:364x2 þ 1:991x3 þ 0:500x4 þ 2:191x5 þ 1:504x6 þ 0:611x7 þ d22 2 d2þ ¼ 7

0:074x1 þ 0:135x2 þ 0:150x3 þ 0:035x4 þ 0:210x5 þ 0:096x6 þ 0:298x7 þ d32 2 d3þ ¼ 1 ð5Þ

0:101x1 þ 0:116x2 þ 0:208x3 þ 0:022x4 þ 0:244x5 þ 0:114x6 þ 0:194x7 þ d42 2 d4þ ¼ 1

xj [ {0; 1}; j ¼ 1; 2; …; 7 di2 ; diþ $ 0; i ¼ 1; 2; 3; 4:

The model given above is solved using LINDO software yielding the results shown in Table 10.
As Table 10 shows, the ZOGP solution yields Length, Time Between Sharpenings, Lead Dust
Generated, Pages per Pencil, and Eraser Dust Generated as the PTRs to be considered in designing a
pencil. The set of PTRs determined using the combined ANP and ZOGP approach is different from the
solutions that are obtained by applying either AHP or ANP by itself. One should note that consideration
of interdependencies in the HOQ and analysis of the design problem from a multi-objective perspective
result in different design attributes to be focused on. The combined ANP and ZOGP approach, which
Table 10
Set of PTRs selected by employing combined ANP and ZOGP approach

PTRs ZOGP solution

Length x1 ¼1
Time Between Sharpenings x2 ¼1
Lead Dust Generated x3 ¼1
Incline Angle to Roll x4 ¼0
Pages per Pencil x5 ¼1
Pressure Cycles to Erase x6 ¼0
Eraser Dust Generated x7 ¼1
 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190 189

aims to quantify the interdependencies and multiple objectives inherent in the design problem in a
systematic way, appears as an effective solution aid.

6. Conclusions

The QFD approach, which enables companies to translate customer needs to relevant product design
requirements, is a design tool of vital importance. In this paper, we present a systematic decision
procedure to be used in QFD product planning, which has been traditionally based on expert opinions.
The decision approach aims to consider the interdependence between the customer needs and PTRs, and
the inner dependence within themselves, along with resource limitations, and design metrics such as
extendibility and manufacturability.
In a period of intensifying competition, the interaction of different approaches should be embraced
and incorporated within the QFD process in order to realize its full potential. This paper employs a
combined ANP and ZOGP approach to incorporate the customer needs and the PTRs systematically into
the product design phase in QFD. The dependencies inherent in the QFD process are taken into account
using the ANP approach. Considering resource limitations and multi-objective nature of the problem, a
ZOGP model is constructed to determine the set of PTRs that will be taken into account in the product
design phase. The use of ANP weights, resource limitations, and other design metrics such as
extendibility and manufacturability in the ZOGP model provides feasible and more consistent solutions.
The application of the decision procedure is demonstrated via an illustrative example. The proposed
framework adds quantitative precision to an otherwise judgmental decision process. The decision
approach presented in this paper can be easily extended for real-world applications of QFD by
considering additional resource limitations and design metrics.

References

Akao, Y (1997). QFD: Past, present, and future. Proceedings of the International Symposium on QFD’97, Linköping.
Armacost, R. L., Componation, P. J., Mullens, M. A., & Swart, W. W. (1994). An AHP framework for prioritizing customer
requirements in QFD: An industrialized housing application. IIE Transactions, 26(4), 72 – 79.
Badri, M. A. (1999). Combining the analytic hierarchy process and goal programming for global facility location – allocation
problem. International Journal of Production Economics, 62, 237–248.
Chan, L. K., Kao, H. P., Ng, A., & Wu, M. L. (1999). Rating the importance of customer needs in quality function deployment
by fuzzy and entropy methods. International Journal of Production Research, 37(11), 2499– 2518.
Cohen, L. (1995). Quality function deployment: How to make QFD work for you. Reading, MA: Addison-Wesley.
El-Gayar, O. F., & Leung, P. S. (2001). A multiple criteria decision making framework for regional aquaculture development.
European Journal of Operational Research, 133, 462– 482.
Griffin, A., & Hauser, J. R. (1993). The voice of customer. Marketing Science, 12(1), 1 – 27.
Hauser, J. R., & Clausing, D. (1988). The house of quality. Harvard Business Review, 66, 63 – 73.
Khoo, L. P., & Ho, N. C. (1996). Framework of a fuzzy quality function deployment system. International Journal of
Production Research, 34(2), 299– 311.
Kogure, M., & Akao, Y. (1983). Quality function deployment and company wide quality control in Japan: A strategy for
assuring that quality is built into products. Quality Progress, 25– 29.
Lee, J. W., & Kim, S. H. (2000). Using analytic network process and goal programming for interdependent information system
project selection. Computers and Operations Research, 27, 367– 382.
190 E.E. Karsak et al. / Computers & Industrial Engineering 44 (2002) 171–190

Lu, M., Madu, C. N., Kuei, C., & Winokur, D. (1994). Integrating QFD, AHP, and benchmarking in strategic marketing.
Journal of Business and Industrial Marketing, 9(1), 41 – 50.
Park, T., & Kim, K. (1998). Determination of an optimal set of design requirements using house of quality. Journal of
Operations Management, 16, 569– 581.
Prasad, B. (1998). Review of QFD and related deployment techniques. Journal of Manufacturing Systems, 17(3), 221– 234.
Saaty, T. L. (1980). The analytic hierarchy process. New York, New York: McGraw-Hill.
Saaty, T. L. (1996). Decision making with dependence and feedback: The analytic network process. Pittsburgh: RWS
Publications.
Saaty, T. L., & Takizawa, M. (1986). Dependence and independence: From linear hierarchies to nonlinear networks. European
Journal of Operational Research, 26, 229– 237.
Saaty, T. L., & Vargas, L. G. (1998). Diagnosis with dependent symptoms: Bayes theorem and the analytic hierarchy process.
Operations Research, 46(4), 491– 502.
Schniederjans, M. J. (1995). Goal programming: Methodology and applications. Norwell: Kluwer.
Schniederjans, M. J., & Garvin, T. (1997). Using the analytic hierarchy process and multi-objective programming for the
selection of cost drivers in activity-based costing. European Journal of Operational Research, 100, 72 – 80.
Shillito, M. L. (1994). Advanced QFD—Linking technology to market and company needs. New York: Wiley.
Wasserman, G. S. (1993). On how to prioritize design requirements during the QFD planning process. IIE Transactions, 25(3),
59 – 65.