See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4916730

A combined AHP-GP model for quality control systems

Article  in  International Journal of Production Economics · June 2001

DOI: 10.1016/S0925-5273(00)00077-3 · Source: RePEc

CITATIONS READS
246 2,216

1 author:

Masood Abdulla Badri

United Arab Emirates University and Department of Community Development
105 PUBLICATIONS   3,205 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Professional development of teachers in Abu Dhabi View project

Schools Evaluation Rating View project

All content following this page was uploaded by Masood Abdulla Badri on 28 August 2019.

The user has requested enhancement of the downloaded file.

 Int. J. Production Economics 72 (2001) 27}40

A combined AHP}GP model for quality control systems

Masood A. Badri*
Department of Management, College of Business and Economics, United Arab Emirates University, Post Box 17555,
Al-Ain, United Arab Emirates
Accepted 20 June 2000

Abstract

Using the results of previous studies of service quality attributes, "ve sets of quality measures are identi"ed. These
indicators or measures, through the analytic hierarchy process (AHP), are then accurately and consistently weighted. The
priority weights are, in turn, incorporated in a goal-programming model to help select the `besta set of quality control
instruments for customer data collection purposes. The paper proposes a decision aid that will allow weighting
(prioritizing) of a "rm's unique service quality measures, consider the real world resource limitations (i.e., budget, hour,
labor, etc.), and select the optimal set of service quality control instruments. The paper addresses two important issues:
how to incorporate and decide upon quality control measures in a service industry, and how to incorporate the AHP into
the model. A real world case study illustrates the application of this combined analytic hierarchy process and
goal-programming (AHP}GP) model.  2001 Elsevier Science B.V. All rights reserved.

Keywords: Analytic hierarchy process; Goal programming; Service quality

1. Introduction vide the `besta solution in terms of coming as close

as possible to reaching all goals. Through a review
A problem encountered in designing quality con- of the literature they developed a list of indicators
trol systems for service organizations is the de"ning the quality construct. These indicators
measurement of the quality construct. Schnieder- were, in turn, incorporated into a goal-program-
jans and Karuppan [1] developed a goal-program- ming model for the design of a quality control
ming model to aid in selecting the `besta set of system in service organizations. A zero}one goal-
quality control instruments in designing a quality programming model was developed to help select
control system. Goal programming is a procedure the best set of quality control instruments. They
for handling multiple-objective situations within presented a small business application to imple-
the general framework of linear programming. ment the model. The model employed a scoring
Each objective is viewed as a goal. Then, given the method to rate the instruments (on a scale from 1 to
usual resource limitations or constraints, the deci- 10). A simple scoring method was used to establish
sion-maker attempts to develop decisions that pro- the priorities for the quality measures with regard
to each instrument.
* Corresponding author. Tel.: 971-3-7665644; fax: 971-3- One of the problems in any multi-objective
7515538. method is the bias introduced by the initial solution
E-mail address: masood@uaeu.ac.ae (M.A. Badri). provided by the selection process. The simple

0925-5273/01/$ - see front matter  2001 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 5 - 5 2 7 3 ( 0 0 ) 0 0 0 7 7 - 3
28 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

scoring method employed by Schniederjans and manner in which customers receive the service).
Karuppan is highly susceptible to such bias where Lehtinen and Lehtinen [8] discussed three kinds of
there is a tendency to adjust biases and anchor quality: physical, corporate, and interactive quality.
upon initial points. Moreover, Tversky and Kahne- Although there has been an avalanche of publi-
man [2], and Steuer [3] raise the issue that when cations on service quality attributes, only few
simple scoring methods are used, the decision- provided attributes that are developed and tested
maker's consistency is not veri"ed. This method scienti"cally. Garvin [9,10] proposed eight dimen-
does not provide consistency feedback to the sions to measure quality. He did not discriminate
decision-maker. Saaty [4] points out that while between goods producing or service providing
decision-makers theoretically are paragons of con- "rms. The eight dimensions included performance,
sistency, in practice humans have been known to features, reliability, conformance, durability, ser-
change their mind, either through reconsideration, viceability, aesthetics, and perceived quality.
or through the process of learning. The analytic DeSouza [11] reported service quality attributes
hierarchy process (AHP) has been proposed from the Pro"t Impact of Market Strategy (PIMS)
as a means of reconciling initial decision-maker's database. There were 12 attributes: delivery,
expression of preference, as well as means of identi- warranty, repair and maintenance, sales services,
fying the consistency of that expression. It provides corporate viability, advertising and promotional
an estimate of additive utility weight that best material, customization, technical support, loca-
matches the initial information provided by the tion, complaint handling, ordering and billing
decision-maker. Moreover, when the AHP is used simplicity, and communications.
to obtain an initial estimate of the priorities, the Parasuraman et al. [12] proposed 10 determi-
initial points are selected on the bases of pairwise nants of service quality that included reliability,
comparison of alternatives. The e!ect of introduc- responsiveness, competence, access, courtesy of
ing bias is lessened [5]. personnel, communications, credibility or trust-
Even though Schniederjans and Karuppan did worthiness of the organization, security or protec-
not use the AHP, they recommended its use to more tion from risk, understanding of customers' needs,
accurately weight the importance of the quality and tangibles or physical elements attesting to the
measures. The purpose of this paper is to extend the service. In a later study, these determinants were
model presented by Schniederjans and Karuppan factor analyzed and generated "ve principal quality
[1] by using the AHP method to aid in accurately dimensions: tangibles, reliability, responsiveness,
representing the goal-programming (GP) model's assurance, and empathy [13].
objective function and quality measure goals. The A consistent theme emerging from these dimen-
paper will demonstrate how current limitations in sions is that customers might use more than just
decision-making involving selecting quality control service outcome or `corea in assessing service qual-
instruments can be overcome by combining the ity. Customer assessment may also be in#uenced by
AHP and GP. The use of the proposed model is the service process and the `peripheralsa associated
illustrated in a real world case study. with the service. The research conducted by Para-
suraman et al. [12] con"rmed that both outcome
1.1. Service quality attributes in literature and process dimensions in#uence customers' evalu-
ation of service quality regardless of service sector.
Early conceptualizations suggested several gen- The 10 determinants of their study, and identi"ed
eral service attributes that might be used to assess earlier, constitute a more comprehensive set of ser-
service quality. Sasser et al. [6] proposed three vice quality dimensions. However, the researchers
di!erent dimensions of service performance: levels acknowledged the possibility of overlapping
of material, facilities, and personnel. Gronroos [7] dimensions. Through extensive empirical research,
proposed two types of service quality: technical using statistical and psychometric tests, they
quality (what customers actually receive from developed and re"ned the SERVQUAL instrument
the service provider), and functional quality (the to focus on "ve principal quality dimensions, which
 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 29

Table 1
Summary of quality dimensions and attributes for service organizations

Quality dimensions De"nition of the dimension

Tangibles Appearance of physical facilities, equipment, personnel, and communication personnel (modern-looking
equipment, visually appealing physical facilities, neat-appearing employees, and visually appealing mater-
ials associated with the service)
Reliability Ability to perform the promised service dependably and accurately (doing things by the promised time,
showing sincere interest in solving problems, performing the service right the "rst time, and insisting on
error-free records)
Responsiveness Willingness to help customers and provide prompt service (employees telling exactly when services will be
performed, employees giving prompt service, employees always willing to help, and employees never being
too busy to respond to requests)
Assurance Knowledge and courtesy of employees and their ability to inspire trust and con"dence (behavior of
employees instilling con"dence in customers, customers feeling safe in their transaction with the "rm,
employees constantly being courteous, and employees having the knowledge to answer questions)
Empathy Caring, individualized attention the "rm provides its customers (giving individual attention, having
convenient operating hours, giving personal attention, having the best interest at heart, and understanding
the speci"c needs of customers)

has become the most widely used methodology for tals [25]; banking, pest control, dry cleaning, and
measuring service quality (see Table 1). fast food [28]; and higher education [29].
This paper will utilize the dimensions of quality For all these reasons, the basic dimensions of
as derived by Parasuraman et al. [14]. The SER- SERVQUAL will serve as attributes of service
VQUAL instrument has generated considerable quality and will be used in this paper.
interest in service quality measurement among
academic researchers. It has gone through several
enhancement stages as a direct reaction to some 2. Selecting the service quality control instruments
concerns raised by other researchers with regard to
its expectations component [15], the interpretation 2.1. Weighting the quality measures: The AHP
and operationalization of expectations [16], its approach
reliability and validity [17], and its dimensionality
[18]. In response to these questions, the three The AHP, introduced by Saaty [4], addresses
colleagues have presented counter arguments, clari- how to determine the relative importance of a set of
"cations, and additional evidence to rea$rm the activities in a multi-criteria decision problem. The
instrument's psychometric soundness and practical process makes it possible to incorporate judgments
value [13,19,20]. on intangible qualitative criteria alongside tangible
The instrument has served as the basis for quantitative criteria. The method utilizes pairwise
measurement approaches used in published studies comparisons of alternatives (quality control instru-
examining service quality in a variety of contexts ments) as well as pairwise comparisons of the mul-
} e.g., real estate brokers [21]; physicians in private tiple criteria (the "ve attributes of service quality).
practice [22]; public recreation programs [23]; The use of such pairwise comparisons to collect
motor carrier companies [24]; a business school data from the decision-maker o!ers signi"cant
placement center, and a tire shop [25]; an ac- advantages [30]. It allows the decision-maker to
counting "rm [26]; discount and department stores focus on the comparison of just two objects, which
[16]; a gas and electric utility company [27]; hospi- makes the observation as free as possible from
30 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

extraneous in#uences. Additionally, pairwise com- a measure of consistency of our judgment. Gener-
parisons generate meaningful information about ally the consistency ratio should be less than 0.10
the decision problem, improving consistency in the (10%) [31]. As a result, we derive priorities and an
decision-making process, especially if the process inconsistency ratio with respect to each of the "ve
involves group decision-making. quality-of-service criteria. The next step is to make
The `quality control systema model is given in judgments about the "ve criteria. A comparison is
the form of a hierarchy (Fig. 1). The decision prob- made with respect to each pair [the number of
lem is structured in a hierarchy of three basic levels: comparisons will be
u(u!1)/2, where u is the
the goal of the decision is at the top, followed by number of criteria in the model].
a second level of criteria (or objectives) and a third The next step will be to synthesize the derived
level of alternatives. This arrangement makes it priorities that were based on the decision-makers'
possible for decision-makers to focus on each and judgments. Synthesis means adding up the global
every part of such a complex problem, and to derive weights of the common nodes at the bottom level of
priorities from simple pairwise comparisons. In this the hierarchy so as to generate a composite priority
study, the goal of the AHP model will be to decide for an alternative across all criteria. Synthesizing
on prioritizing the set of quality control instru- shows the results of the entire work, the overall
ments given some quality of service criteria. priorities of the alternatives. These alternative pri-
The criteria or objectives of the quality system orities will be used later in the combined AHP}GP
decision consist of "ve elements: reliability, assur- model as weights in the objective function. The
ance, responsiveness, empathy, and tangibles. The derived priorities with respect to each of the "ve
decision alternatives, the options from which quality-of-service criteria will be used in the com-
a choice is made, are the quality instruments for bined model to serve as the contribution that each
customer data collection purposes. criterion makes to each alternative.
Once the model is built, the decision-makers
evaluate the elements by making pairwise compari- 2.2. The goal-programming model
sons. A pairwise comparison is the process of
comparing the relative importance, preference, or In the goal-programming model, the decision
likelihood of two elements with respect to an ele- variable is x (0 or 1). The objective function, given
G
ment in the level above. When all the comparisons by Eq. (1), seeks to minimize deviation from desired
are completed, we calculate the priorities and targets for limited resources (costs, available

Fig. 1. Structure of the AHP model.

 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 31

management hours, and available employee given in Eq. (7):

hours)

 
)
Min Z" P (w d\, w d>) #P (d\, d>)
Min Z"P (d\, d>)#P (d\, d>)#P (d\, d>). (1) I I I I I ? ? ?
A A A @ @ @ C C C I
The goal constraints in Eqs. (2), (3) and (4) #P (d\, d>)#P (d\, d>)#P (d\, d>),
A A A @ @ @ C C C
represent the availability of limited resources. The
(5)
right-hand side of each equation re#ects the tar-
geted or desired level of the resource utilization, K
a x #d\!d>"Q (for k"1,2, 2, K), (6)
where C denotes cost, B available management IG G I I I
hours, and E available employee hours. We could G
also express these limitations of available resources K
w x #d\!d>"1. (7)
as `system constraintsa by removing the deviation G G ? ?
G
variables from the constraints and the objective
Eq. (6) shows that there will be one equation
function and by changing the equality signs of the
for each of the quality measures. The ordering
constraints to less than or equal signs. For conveni-
of goal constraints in Eq. (6) depends on the
ence, this will be done through the combined
nature of the problem situation. The classic
model:
lexicographic ordering of preemptive priorities
K may be appropriate if the highest ranked quality
a x #d\!d>"C, (2)
AG G A A measures are preemptively important. To better
G utilize the results of the AHP method, the a
IG
K will denote the contribution that each criterion
a x #d\!d>"B, (3)
@G G @ @ makes to each alternative. Compared to previous
G research, and as shown in Eq. (1), more
K appropriate weighting of quality measures is
a x #d\!d>"E. (4)
CG G C C desired and used. By utilizing the obtained
G AHP weightings we have established the ranks
The ordering of these goal constraints depends of the individual quality control instrument,
on the nature of the problem situation. Either by establishing a relative weighting w for each
preemptive or non-preemptive goals could be used I
of them. The term P is some k priority
depending on the order of importance, if any, of the I
lexicographic rankings where P 'P '2'P .
goals.   I
In this model, Eq. (1) is the objective function,
which seeks to minimize deviation from the
2.3. The combined model desired goals consistent with the AHP ranking
of the quality control instruments. The lexi-
There are several studies that used the AHP cographic nature of goal programming treats
methodology in combination with goal program- the AHP weights given by w as a sub-ranking
I
ming [30,32}35]. In the combined model, the ob- of the quality control instruments within their
jective function also includes deviation variables speci"c P . The greater the w , the more desirable
I I
associated with the quality measure goals. It will the selection of quality instrument in the decision
seek to minimize such deviations from desired process.
levels. The revised objective function is given in Eq. As mentioned before, the constraints for
(5). Moreover, a set of constraints, as shown in Eq. limited resources of total cost, available
(6), will be added to re#ect the quality target of management hour, and available employee hour
Q in each of the `goal constraintsa. An equation could also be expressed as `system constrainta
I
associated with the AHP weights for the quality by removing the deviation variables and replacing
control instruments will be added to re#ect the the equality signs with inequality signs in Eqs.
preferences for the di!erent instruments. This is (2)}(4).
32 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

3. Application of the combined model 3.1. Deriving the AHP weights

Based on a real life problem involving a large To facilitate the use of AHP, the problem is
department store, the proposed AHP}GP meth- decomposed into a multilevel hierarchy showing
odology is illustrated to show how it is applied. The the overall goal of the decision process, each deci-
department store is evaluating seven potential sion criterion to be used, and the decision alterna-
alternatives to better design a quality control sys- tives to be considered. The "ve decision criteria are:
tem for customer data collection purposes. Unfor- reliability, empathy, responsiveness, tangibles, and
tunately, each instrument captures the principal assurance. The decision alternatives to be
quality measures to a di!erent degree and requires considered are: large-scale mail survey of known
varying levels of resources. The problem for a large customers, large-scale telephone survey of known
department store with limited resources is therefore customers, open-ended comment cards, personal
to select the best set of instruments to obtain cus- interviews of customers while shopping, point of
tomer feedback and perception on key quality purchase survey to measure employee service deliv-
measures. ery, point of purchase survey to measure product
Our model requires a group process rather than line service delivery, and small-scale telephone sur-
aggregation of individual inputs. In this process, vey of known customers. The multilevel hierarchy
the group establishes a single set of weights for the is shown in Fig. 2.
decision criteria, and then rates the decision alter- A relative preference matrix is formulated for
natives. The group process has several advantages each subordinate level of hierarchy. The decision-
over aggregation of individual ratings [36]. It facil- making team was given the chance to express its
itates a common understanding of the meaning and preferential expert judgment (the pairwise compari-
signi"cance of each criterion. This commonality of sons) in the selection process. Each decision alter-
understanding is not achieved through aggregating native is compared against another decision
the inputs of individual evaluations. The group is alternative with respect to one particular decision
often able to clarify misunderstanding and di!er- criterion at a time. These within-pairwise compari-
ences in interpretation of the data so that there is sons result in priority values of each decision
a more uniform understanding of the facts. In addi- alternative by each decision criterion. The result of
tion, a group process utilizes the dynamics of elicitation at this level of the hierarchy is a matrix
powerful in#uence within the decision-making showing the values expressed by the decision-mak-
group. ing team.
To provide expert judgment and consultations The `right eigenvector methoda is applied to
in a group process, the department store manage- each of the matrices in turn using Expert Choice
ment formed a team, which consisted of "ve Pro software (version 9.0) [37]. From this applica-
members to work side by side with the research tion, the relative priority for each alternative deci-
team. The "ve members consisted of the general sion by criterion is computed. The results are
managers of marketing, "nance and personnel, shown in Table 2. During the elicitation process, an
and the two heads of advertising and public appropriate level of consistency is necessary to
relations. The research was given full support achieve meaningful results. The measure of consist-
by the top management. To derive the AHP ency provided by the AHP is given by Saaty [38]:
weights, a preliminary data collection stage "( !n)/(n!1), where  is the maximum
 
attempted to get some reaction and feedback with eigenvalue of the positive reciprocal matrix, and n is
regard to the "ve quality attributes. A survey of the number of objectives. Maximum consistency
known customers was carried out to identify the occurs as  approaches zero. Researchers use a con-
relative importance of each criterion. The results sistency ratio of 0.10 or less as guidelines in evaluat-
obtained, along with the management-team's input ing consistencies [31]. As can be seen in Table 2, all
were used to later structure and synthesize the of the ratios are below the maximum 0.10 level. Via
AHP model. `synthesizinga [31, p. 159] the sets of priority
 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 33

Fig. 2. The AHP model for a department store.

Table 2
Resulting priority values for each decision alternative by each criterion

Decision alternatives Service quality criterion decisions

Responsiveness Assurance Reliability Empathy Tangibles

x 0.031 0.219 0.028 0.026 0.415


x 0.335 0.160 0.397 0.376 0.055

x 0.036 0.145 0.041 0.040 0.097

x 0.099 0.068 0.067 0.165 0.154

x 0.099 0.070 0.071 0.078 0.034

x 0.099 0.130 0.105 0.075 0.036

x 0.302 0.208 0.291 0.240 0.208

Total 1.000 1.000 1.000 1.000 1.000
Inconsistency ratio 0.050 0.010 0.020 0.050 0.030

values (from Table 2) are combined into a matrix overall rankings (in terms of weights) of the seven
and multiplied by the overall criteria relative priori- quality instrument candidates. Before we discuss
ties as presented in Table 3. The result is an overall the combined model application, and for simpli"-
prioritization of the decision alternatives. The re- cation, Fig. 3 provides an illustration of the
sults of this step, presented in Table 4, are the combined model development steps.
34 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

Table 3
Resulting priority values for each criterion

Criterion decisions Priority

Responsiveness 0.479
Assurance 0.139
Reliability 0.281
Empathy 0.025
Tangibles 0.076
Total 1.000

Table 4
Overall AHP weights of the decision alternatives

Decision AHP weighting Decision preference

alternatives

X 0.086 Fifth preference


x 0.308 First preference

x 0.057 Seventh preference

x 0.091 Fourth preference

x 0.081 Sixth preference

x 0.099 Third preference

x 0.277 Second preference

Total 1.000

Fig. 3. Illustrative diagram of the combined model

development.
3.2. The combined AHP}GP application

Goal programming permits resource limitations

such as budgetary limitations or limited hours of version of the large-scale telephone survey. As a
labor, and other selection limitations (such as result, system constraints would be needed to pre-
system constraints) that must be observed in estab- vent such duplications.
lishing and maintaining a service quality control The decision team also wanted to ensure that the
system. Trying to decide which of the seven service selection of quality control instruments would
control instruments to use in designing the depart- measure those quality attributes that were most
ment store's service quality control system requires important to their customers. To determine the
the recognition of several resource limitations rela- ability of each of the instruments to measure each
tive to the nature of each of the instruments. of the "ve quality measures presented in Table 1,
Table 5 provides a summary of yearly cost and the AHP scores (priority values) for each decision
labor information on the seven service quality con- alternative by each criterion will be used. The prior-
trol instruments. ity scores are presented in Table 2. We recall that
The decision team felt that a minimum of three of the derived priority scores are the relative contribu-
the seven instruments would be needed to justify tion that each of the "ve criteria makes to each of
the expense of the budget. Also, the nature of some the seven alternatives.
of the quality instruments created duplication that Since it is our objective to select quality instru-
would be wasteful and unnecessary. For example, ments with the highest weights, a goal constraint
the minor telephone survey was just a scaled down could be added to maximize the overall weights of
 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 35

Table 5
Yearly cost and labor information on service quality control instruments

Resource item Service quality control instruments (decision alternatives) Total budgeted
yearly resources
x x x x x x x
      
Estimated cost $9760 $9760 $1207 $640 $852 $2500 $4738 $15,000
Management 186 186 54 50 98 98 42 530
hours
Employee hours 47 47 57 93 278 493 600 1070

the instruments selected. We recall that the AHP selected. The equation shows that point of purchase
solution provided us with overall weights for each surveys to measure employee service delivery and
of the seven quality instruments (alternatives). product line service delivery are most expensive
These weights are presented in Table 4. with respect to annual cost.
Moreover, the management of the department
3.3. Formulation of the department store's problem store limited the yearly number of hours that
management could devote to the service of quality
System constraints are needed to reduce resource control instruments to 530 hours, and the number
wastage by not selecting similar quality control of employee labor hours for data collection, inter-
instruments. In our case, the minor telephone sur- views and clerical activities to a maximum of
vey was just a scaled down version of the major 1070 hours:
telephone survey. The management team also re-
186x #186x #54x #50x #98x #98x
quested that only one form of large-scale survey of      
known customers be performed, a telephone or #42x #d\!d>"530, (11)
 @ @
mail survey. We should recall also, that at least
47x #47x #57x #93x #278x #493x
three instruments are to be selected. Mathemat-      
ically, these `system constraintsa are treated as #600x #d\!d>"1070. (12)
simple linear programming constraints. Typically,  C C
they create resource boundaries where an interval The three equations associated with resource
of possible solutions is usually narrowed down to limitations could also be formulated as system con-
the optimal choice by subsequently considering the straints if desired by decision-makers. However, the
priority of goal constraints: management team asked to keep them as goal
constraints to better understand their contribution
x #x )1 and x #x )1, (8) to the overall structure of the model.
   
x #x #x #x #x #x #x *3, (9) The management of the department store
       wanted to also ensure that the selection of the
9760x #9760x #1207x #640x #852x quality control instruments would really measure
    
those quality measures that were important to their
#2500x #4738x #d\!d>"15,000. (10)
  A A customers. One goal constraint is needed to ensure
The three resource limitation constraints will that the instruments with the highest weights
have deviation variables associated with them, and obtained from the AHP analysis will be selected.
will attempt to minimize the positive deviations by Such goal constraint will attempt to maximize the
adding deviation variables to the overall objective weights by selecting the quality control instruments
function. The department store allocated a max- with the highest priorities:
imum yearly operating budget limitation of $15,000
0.086x #0.308x #0.057x #0.091x #0.081x
for the cost of materials, postage, phone calls, print-     
ing, etc., for all of the quality control instruments #0.099x #0.277x #d\!d>"1. (13)
  N N
36 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

To determine the ability of the instruments to The ordering of deviation variables in the objec-
measure each of the "ve measures presented in tive function associated with goal constraints
Table 1, the AHP priority values (or weights) for (d }d ) will rely on the data obtained via the
I I
each quality control instrument by each quality AHP process and reported in Table 3. We recall
measure will be used. The individual weights rep- that the order of the "ve measures of service quality
resent the a parameters in the combined model. were responsiveness, reliability, assurance, tan-
IG
The development of the `goal constraintsa in the gibles, and empathy, respectively. The three
situation originates chie#y from the a data in measures of responsiveness, reliability, and assur-
IG
Table 2. The right-hand-side values for each goal ance accounted for 89.9% of the overall weights,
constraint seek to pressure a selection of service while the remaining two measures of empathy and
quality control instruments or x with the highest tangibles accounted for only 10.1% of the total
G
scores (i.e., the most useful collection of instruments weight. The results also re#ect the importance of
to measure quality). In other words, the best set of the top three measures as the quality control instru-
three quality instruments was chosen for each of ments that are selected portray these facets of
the "ve quality measures. Of course, the right-hand quality. For example, the selection of x (point of

side could be set at any level; however for our case, purchase survey to measure employee service deliv-
each of the right-hand-side Q was simply derived, ery) re#ects the importance of responsiveness
I
by summing the best three a , for each of the "ve (0.335), reliability (0.397), and assurance (0.160). We
IG
quality measures. The model seeks to select the best also notice that the selection of x which assigns its

or perfect set of quality instruments, for each of the largest weights to empathy (0.165) and tangibles
quality measures: (0.154) attempts to help in the creation of an overall
quality control system that does not neglect the
0.031x #0.335x #0.036x #0.099x #0.099x
     contribution of the two measures of empathy and
#0.099x #0.302x #d\ !d> "0.736, (14) tangibles. Management was satis"ed with the solu-
  I I tion since the selected instruments, in general, tried
0.219x #0.160x #0.145x #0.068x #0.070x to re#ect the department store's image to its
    
customers through the "ve measures of service
#0.130x #0.208x #d\ #d> "0.587, (15)
  I I quality.
0.028x #0.397x #0.041x #0.067x #0.071x We could simply use the order of importance as
     reported in the table. Meanwhile, we could make
#0.105x #0.291x #d\ !d> "0.793, (16) better use of the AHP process results by utilizing
  I I
the speci"c weights provided in the table. In this
0.026x #0.376x #0.040x #0.165x #0.078x
     case, as mentioned earlier, the AHP weights pro-
#0.075x #0.240x #d\ !d> "0.784, (17) vide internal relative rankings (or desirability
  I I scores) for each of the quality measures. The
0.415x #0.055x #0.097x #0.154x #0.034x objective function will seek to minimize the overall
    
deviations.
#0.036x #0.208x #d\ #d> "0.777. (18)
  I I
The objective function will attempt to minimize 3.4. Solution of the combined model
the overall deviations in each of the goal
constraints. The goal constraints include resource The solution to this problem was generated using
limitation constraints as well as desired quality LINDO, Version 6.0 [39]. Table 6 presents the
measure goals: optimal selection of service quality control instru-
ments. The optimal selection turned out in this case
Min Z"P (d>#d>#d>)#P d\
 A @ C  N study to be where the department store will use
#P (0.479d\ #0.139d\ #0.281d\ point of purchase survey to measure employee ser-
 I I I vice delivery (x ), open-ended comment cards (x ),
 
#0.025d\ #0.076d\ ). (19) small-scale telephone survey of known customers
I I
 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 37

Table 6
The combined AHP}GP model solution

(a) Decision variables

X "1 Select point of purchase survey to measure employee service delivery

X "1 Select open-ended comment cards

X "1 Select small-scale telephone survey of known customers

X "1 Select large-scale mail survey of known customers

X "0 Do not select point of purchase to measure product line service delivery

X "0 Do not select large-scale telephone survey of known customers

X "0 Do not select personal interviews of customers while shopping

(b) Slack in resource system constraints

Constraints Usage Total available Slack

Budget $14,107 $15,000 $893

Management hours 388 530 142
Employee hours 690 1070 380

(c) Deviation in quality measure constraints

Constraint Target Obtained Underachievement

Responsiveness 0.736 0.569 0.167

Assurance 0.587 0.503 0.084
Reliability 0.793 0.610 0.183
Empathy 0.784 0.656 0.128
Tangibles 0.777 0.342 0.435

(x ), and large-scale mail survey of known cus- tem than low-priority attributes such as empathy

tomers (x ). The selected instruments will act as and tangibles.

a collective system designed to monitor perfor- The management of the department store was
mance on the "ve quality measures de"ned in delighted to see that the resulting quality control
Table 1. The other three instruments were not system would be achievable under-budget (a saving
chosen because of the limitations posed by the of $893). With regard to the "ve quality measures,
system constraints, goal constraints, as well as the we note that the resulting solution adequately
assigned priorities. addresses the attributes with the highest priorities.
The slack in the goal constraints associated with For example, with regard to the measures of
resource limitations that were used to model cost responsiveness, reliability, and assurance, the three
and hours usage also provides some summary attributes with the highest priorities, they are
information on those resources. The amount of adequately represented by the selected instruments.
deviation actually provides a crude measure of de- Being happy with cost "gures of the model solu-
viation of the resulting quality control system in tion, management requested a scenario to be
measuring the "ve quality attributes. The smaller checked where the priorities for the "ve quality
the resulting deviation, the better the system measures are placed ahead of the resource limita-
measures the respective quality attributes. The tion goals in the objective function. The scenario
high-quality attributes like responsiveness, reliabil- called for a solution that overshoots the budget by
ity, and assurance should be more comprehensively $11,105, which was totally unacceptable. Other
measured under the proposed quality control sys- scenarios were also looked at but most of them
38 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

a!ected the budget cost and therefore were not tomers and management. The derived weights were
considered further. used to establish preemptive priorities in the objec-
Due to the importance of the three quality tive function, and utility weights for the "ve quality
measures of responsiveness, reliability, and assur- measures in the goal constraints.
ance as the top three highest priorities, a right- The proposed combined model has saved
hand-side sensitivity analysis was undertaken on resources by recognizing necessary constrained
these three goals (mainly Eqs. (14)}(16)). Results resources. It also presented a more thorough deci-
show that substantial changes to the right-hand- sion process. For example, the completeness and
side goals must occur before altering the existing non-bias nature of the pairwise comparisons in the
solution. In other words, the management team AHP method was a more thorough means of
that assisted in this research would have had erred weighting quality control instruments in the deci-
in the assessment of the weights assigned to the "ve sion process relative to ranking methods employed
quality measures. in previous studies dealing with quality control
The department store management implemented instrument selection. The most obvious advantage
the suggested results obtained from the combined of using the AHP model for deriving weights for the
AHP}GP model in the summer of 1998. After quality control instruments as well as measures is
almost one year of its implementation, the top that it provides for consistent decision-making. All
management of the department store has conveyed instruments are evaluated on a single set of
its complete satisfaction with the results. The imple- weighted criteria (quality control measures). This
mentation of the study was reported several times should help to reduce the subjectivity of the pro-
in the daily-newspapers of the city. Management cess. In addition, using a mathematical program-
feels that many customers have noticed increased ming method like goal programming allowed for
awareness in service quality, re#ecting to some additional trade-o! information that permitted
degree the comprehensiveness of the quality con- decision-makers to see these tradeo!s in terms of
trol system to reach the customers. The top cost and hours.
management has also promised more money to be The combined AHP}GP method o!ers a system-
available for future implementations of similar atic, easy-to-use approach to the service quality
projects. control instruments selection decision problem. It
extends previous research in the area by incorpor-
ating a comprehensive prioritization system within
4. Concluding remarks an optimization resource allocation process. The
method accurately represents the model's objective
The quality constructs suggested in this paper for function and quality measure goals, while avoiding
designing quality control systems in service organ- some pitfalls of goal-programming modeling
izations were measured via "ve criteria (or service encountered in previous studies.
attributes): responsiveness, reliability, assurance, The management team formally presented the
empathy, and tangibles. The combined AHP}GP model to top management of the department store.
model presented has helped a large department The top management viewed the model's structure
store select from a set of seven survey instruments and mechanism favorably, believing that it is a tool
those four that will collectively measure the ranked that can be used in the development of a quality
quality measures de"ned by the management team. control system and ordered the immediate imple-
The AHP is "rst used to prioritize the set of quality mentation of its results. Top management has sug-
control instruments along the "ve measures in gested that the research team should continuously
a consistent manner. The selected set of four survey re"ne the hierarchy, modify the judgments, and
instruments collectively represents the department generate new set of weights, since customer taste is
store's service quality control systems. The AHP dynamically sensitive. In addition, they have sug-
process was used to derive weights for each of these gested that the model should also expand on the
quality measures using solicitations from both cus- resource limitation part to include others such as
 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40 39

advertisement costs, employee training costs for [14] A. Parasuraman, V. Zeithaml, L. Berry, SEVQUAL:
better implementation of some of the instruments, a multiple-item scale for measuring consumer perceptions
and other costs associated with providing cus- of service quality, Journal of Retailing 64 (1) (1988) 12}40.
[15] S. Taylor, J. Nicholson, J. Milan, R. Martinez, Assessing
tomers with promotion gifts to better solicit their the roles of service quality and customer satisfaction in the
participation in data collection. They also recom- formation of the purchase intentions of Mexican con-
mended that future e!orts should also examine the sumers, Journal of Marketing Theory and Practice 5 (1)
viability of other forms of quality control instru- (1997) 78}90.
ments such as home visits of some loyal customers, [16] R. Teas, Expectations, performance evaluation and con-
sumer's perceptions of quality, Journal of Marketing 57 (4)
personal interviews with suppliers and personal (1993) 18}34.
interviews with service providing sta! members. [17] T. Brown, G. Churchill, P. Peter, Improving the measure
A revised model should be ready for implementa- of service quality, Journal of Retailing 69 (1) (1993)
tion in the fall of 2000. 127}139.
[18] D. Finn, C. Lamb, An evaluation of the SERVQUAL
scales in retail setting, in: R.H. Holman, M.R. Solomon
(Eds.), Advances in Consumer Research, Vol. 18, Associ-
References ation for Consumer Research, Provo, UT, 1991.
[19] A. Parasuraman, L. Berry, V. Zeithaml, Re"nement and
[1] M. Schniederjans, C. Karuppan, Designing a quality con- reassessment of the SERVQUAL scale, Journal of Retail-
trol system in a service organization: A goal programming ing 67 (1991) 420}450.
case study, European Journal of Operational Research 81 [20] A. Parasuraman, L. Berry, V. Zeithaml, More on improv-
(1995) 249}258. ing service quality measurement, Journal of Retailing 69
[2] A. Tversky, D. Kahneman, Judgment under uncertainty: (1993) 40}147.
Heuristics and biases, Science 185 (1974) 1124}1131. [21] L. Johnson, M. Dotson, B. Dunlop, Service quality deter-
[3] R. Steuer, An interactive multiple objective linear pro- minants and e!ectiveness in the real estate brokerage
gramming procedure, TIMS Studies on Management industry, The Journal of Real Estate Research 3 (1988)
Science 6 (1977) 225}239. 21}36.
[4] T. Saaty, Ascaling method for priorities in hierarchical [22] S. Brown, T. Swartz, Agap analysis of professional service
structures, Journal of Mathematical Psychology 15 (1977) quality, Journal of Marketing 53 (1989) 92}98.
234}281. [23] J. Crompton, K. Mackay, User's perceptions of the relative
[5] D. Olson, M. Venkataraman, J. Mote, Atechnique using importance of service quality dimensions in selected public
the analytic hierarchy process in multiple objective plann- recreation programs, Leisure Sciences 11 (1989) 367}375.
ing models, Socio-Economic Planning Science 20 (6) (1986) [24] R. Brensinger, D. Lambert, Can the SERVQUAL scale be
361}368. generalized to business-to-business services? Knowledge
[6] W. Sasser, P. Olsen, D. Wycko!, Management of Service Development in Marketing, 1990 AMA's Summer Educa-
Operations: Text and Cases, Allyn and Bacon, Boston, tors' Conference Proceedings, 1990.
MA, 1978. [25] J. Carman, Consumer perceptions of service quality: An
[7] C. Gronroos, Strategic management and marketing in the assessment of the SERVQUAL dimensions, Journal of
service sector, Swedish School of Economics and Business Marketing Research 66 (1) (1990) 33}55.
Administration, Helsinki, Finland, 1982. [26] D. Bojanic, Quality measurement in professional service
[8] U. Lehtinen, J. Lehtinen, Service quality: A study of qual- "rms, Journal of Professional Services Marketing 7 (2)
ity dimensions, Unpublished Research Report, Service (1991) 27}36.
Management Group OY, Finland, 1982. [27] E. Babakus, G. Boller, An empirical assessment of the
[9] D. Garvin, What does `product qualitya really mean? SERVQUAL scale, Journal of Business Research 24 (1992)
Sloan Management Review 26 (1) (1984) 25}43. 253}268.
[10] D. Garvin, Competing on the eight dimensions of quality, [28] J. Cronin, S. Taylor, Measuring service quality:
Harvard Business Review 65 (5) (1987) 101}109. A reexamination and extension, Journal of Marketing 56
[11] G. DeSouza, Now service business must manage quality, (1992) 55}68.
Journal of Business Strategy 10 (3) (1989) 21}25. [29] W. Boulding, A. Kalra, R. Staelin, V. Zeithaml, A dynamic
[12] A. Parasuraman, V. Zeithaml, L. Berry, A conceptual process model of service quality: From expectations to
model of service quality and its implications for future behavioral intentions, Journal of Marketing Research 30
research, Journal of Marketing 49 (4) (1985) 41}50. (1993) 7}27.
[13] A. Parasuraman, V. Zeithaml, L. Berry, Reassessment of [30] M. Schniederjans, R. Wilson, Using the analytic hierarchy
expectations as a comparison standard in measuring ser- process and goal programming for the information system
vice quality: Implications for further research, Journal of project selection, Information and Management 20 (1991)
Marketing 58 (1) (1994) 111}124. 333}342.
40 M.A. Badri / Int. J. Production Economics 72 (2001) 27}40

[31] T. Saaty, Decision Making for Leaders, RWS Publica- of cost drivers in activity-based costing, European Journal
tions, Pittsburgh, PA, 1995. of Operational Research 100 (1997) 72}80.
[32] N. Bryson, A goal-programming method for generating [36] E. Erkut, S. Moran, Locating obnoxious facilities in
priority vectors, Journal of the Operational Research the public sector: An application of the analytic
Society 46 (5) (1995) 641}648. hierarchy process to the municipal land"ll sitting deci-
[33] D. Despotis, Fractional minimax goal-programming: sions, Socio-Economic Planning Science 25 (2) (1991)
A uni"ed approach to priority estimation and preference 89}102.
analysis in MCDM, Journal of the Operational Research [37] Expert Choice, Decision Support Software, Pittsburgh,
Society 47 (8) (1996) 989}999. PA, 1995.
[34] S. Myint, M. Tabucanon, Amultiple criteria approach [38] T. Saaty, The Analytic Hierarchy Process, McGraw-Hill,
to machine selection for #exible manufacturing systems, New York, 1980.
International Journal of Production Economics 33 [39] L. Schrage, Optimization Modeling with LINDO, 5th
(1}3) (1994) 121}131. Edition, Brooks/Cole Publishing Co., Paci"c Grove, CA,
[35] M. Schniederjans, T. Garvin, Using the analytic hierarchy 1997.
process and multi-objective programming for the selection

View publication stats