Sm

AE

a

ARRAA

KSMQHCD

1

iccmviccaostiEatoms

((

1d

Applied Soft Computing 10 (2010) 1013–1027

Contents lists available at ScienceDirect

Applied Soft Computing

journa l homepage: www.e lsev ier .com/ locate /asoc

upplier selection paradigm: An integrated hierarchical QFD methodology underultiple-criteria environment

rijit Bhattacharya ∗, John Geraghty, Paul Youngnterprise Process Research Centre, School of Mechanical & Manufacturing Engineering, Dublin City University, Glasnevin, Dublin, Ireland

r t i c l e i n f o

rticle history:eceived 22 September 2008eceived in revised form 5 May 2010ccepted 30 May 2010vailable online 8 June 2010

a b s t r a c t

A concurrent engineering approach integrating analytic hierarchy process (AHP) with quality functiondeployment (QFD) in combination with cost factor measure (CFM) has been delineated to rank and sub-sequently select candidate-suppliers under multiple, conflicting-in-nature criteria environment within avalue-chain framework. Engineering requirements and customer requirements governing the selectiondecision have been identified. The hierarchical QFD methodology allows the decision maker (DM) to rank

eywords:upplier selectionulti-criteria decision-making (MCDM)uality function deployment (QFD)ierarchical MCDM

the candidate-suppliers considering both CFM and the subjective factors. The sensitivity of the proposedmethodology is elucidated considering a parameter called objective factor decision weight. The devisedmethodology has been tested with the dataset adopted from Yahya and Kingsman [89]. Liu and Hai [51]tested their model with the same dataset. A comparative analysis using design of experiment has beenelucidated so as to demonstrate the efficacy of the devised hierarchical concurrent engineering approach.

FMecision support

. Introduction

Supply chain management is a process of planning, implement-ng, and controlling the operations of the supply-chain networkatering to the requirements of customers (purchasers) as effi-iently as possible. One of the primary activities of a value chainodel [62] is to provide service to the customers thereby adding

alue to the value-chain network. Further, the goal of any organ-sation is to maximise the value creation while minimising theosts. Thus, selection of a supplier plays a crucial role in a valuehain, or present days’ supply-chain network of any organisations it demands trading off among cardinal and ordinal preferencesf the decision makers (DM) in an optimal way. The supplierelection process is the most significant variable in the effec-ive management of modern supply-chain networks as it helpsn achieving high quality products and customer satisfaction [33].ffective supplier selection calls for robust analytical methodsnd decision support tools [57] that are able to trade off mul-

iple subjective and objective criteria. In an exhaustive reviewf 76 articles Weber et al. [84] found that 47 address involve-ent of more than one criterion [84]. Dickson [28] identifies a

et of 23 criteria considered by purchasing managers under dif-

∗ Corresponding author. Tel.: +353 1 7006496.E-mail addresses: arijit.bhattacharya@dcu.ie, arijit.bhattacharya2005@gmail.com

A. Bhattacharya), john.geraghty@dcu.ie (J. Geraghty), paul.young@dcu.ieP. Young).

568-4946/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.asoc.2010.05.025

© 2010 Elsevier B.V. All rights reserved.

ferent supplier selection scenario. A supplier selection decisionis inherently a multi-criteria problem and a decision of strategicimportance to companies [41]. Thus, the selection decision withina supply-chain framework is a complex process involving multiplecriteria. Supplier selection decisions within a supply-chain networkare complicated as potential options for such selection decisionsare evaluated on more than one criterion [85]. Criteria may varydepending on the type of product being considered and includemany qualitative factors in addition to the quantitative criteria [81].Therefore, supplier selection is a multi-criteria decision-makingproblem which includes both qualitative and quantitative factors[88,92].

This paper addresses the relationship among the criteria forsupplier selection decision-making. Both cardinal and ordinalpreferences have been considered for evaluation of candidate-suppliers. In a supply-chain framework such decision-makinginvolves cost factors. Thus, cost factor components have beenincluded and a trade off among all the criteria has been estab-lished integrating the quality function deployment (QFD) technique[1] suitably with analytic hierarchy process (AHP) [69,70]. Sup-plier selection is viewed as a combination of both customerrequirements and engineering requirements. Customers are thecompanies that purchase the technical expertise of the suppliers.

Therefore, such a company–supplier relation can be viewed as a‘house of quality’ model. The outcome of the integrated method-ology presented in this paper is determined with indices tradingoff all the types of information available within the supply-chainframework.

1014 A. Bhattacharya et al. / Applied Soft Co

Nomenclature

SymbolsSI selection index of the ith candidate-supplierSFM subjective factor measureOFM objective factor measureOFC objective factor cost�max principal eigenvalueQj sum of column vectordij element of the decision matrixn number of rows/colums in a decision matrixUV utility valuesCI consistency indexRI random consistency indexCR consistency ratiowj importance weight for the jth EReq

wj normalized importance weight for the jth EReq

Rij quantified relationship between the ith CReq and thejth EReq

ci importance weight of the ith CReq

eij utility values of the jth alternative on the i th tech-

2pcwoha3ieiKcdc

2

twcrmtro

2

mifiaiag

nical criteriaSj overall score for the jth candidate-alternative

The remainder of the paper is organised as follows: Sectionpresents a survey of existing literature in the field of sup-

lier/vendor selection. An attempt has been made in this regard tolassify the tools/methodologies used in evaluating the suppliersithin the scope of the research. This leads framing of the research

bjectives in the later part of Section 2. QFD and AHP techniquesave been briefed and subsequently the proposed integrated hier-rchical methodology for supplier selection is delineated in Section. Section 4 is directed towards the development of the case study,

nteraction of criteria, sub-criteria and cost factor components rel-vant to the selection decision. The devised methodology has beenmplemented in a real-world problem adapted from Yahya andingsman [89] and Liu and Hai [51] in Section 5. Section 6 providesomparative analyses of the results obtained which are furtherirected to ascertain the scope for future work. Finally, Section 7oncludes with the criticism of the supplier selection process.

. Survey of existing literature and research objectives

Literature reveal a vast number of published works in regardo the selection of suppliers within different supply-chain frame-orks. Supplier selection as well as evaluation is one of the most

ritical activities of any firm [8]. The supplier selection literature isich in terms of conceptual/empirical works and decision supportethods [8]. A critical review of the decision methods supporting

he supplier selection process is found in [14,75,78]. It has beeneported that any real-life supplier selection process is of a multi-bjective nature [30].

.1. AHP, ANP and integrated models for supplier selection

Among the decision support methods, application of the AHPethod [6,12,53,61,69,71,77] to the supplier selection problem

s not new in the art. It has been reported that AHP provides aramework to cope with multiple-criteria situations involving intu-

tive, rational, qualitative and quantitative aspects [3]. Due to thesedvantages of AHP, researchers widely use the AHP framework tontegrate with linear programming (LP) [34], data envelopmentnalysis (DAE) [51,63], goal programming (GP) [26,44], lexico-raphic goal programming (LGP) [17], multi-objective pre-emptive

mputing 10 (2010) 1013–1027

goal programming (PGP) [60], grey relational analysis [90], roughsets theory and multi-objective mixed integer programming [88].

In the literature the receptivity of decision makers to the use offormal decision tools in terms of formulation of decision criteria,the qualification of suitable candidate-suppliers and recognition ofthe need for a new supplier have been argued widely [13]. Amongsuch ‘formal decision tools’, Saaty’s analytic network process (ANP)model [72] is found suitable for the supplier evaluation process[32]. On the other hand, claims of some researchers [5,18,41] tointegrate the cardinal and ordinal preferences using ANP/AHP forvendor selection decisions are not valid. It is argued that integra-tion of conflicting-in-nature quantitative and qualitative factorsis required for an effective supplier selection procedure [55]. Thefocus of these works leads one to systemise the steps like determi-nation of buyer–supplier relationships and formation of selectioncriteria, i.e., data collection, but does not consider the voice of thepurchaser.

2.2. Fuzzy techniques in supplier selection decisions

In the literature, supplier selection and evaluation studies havebeen conducted with fuzzy techniques [7,10,36,80] applied to themulti-attribute selection models [64,76]. One such method utilisesa fuzzy supplier selection algorithm (FSSA) based on predeterminedperformance criteria and product-related performance criteria [7].Adequate argumentations are not present in the literature so as toadopt the FSSA as a ‘realistic approach for supplier selection’ [7].Though it is much discussed that the decisions based on vague orimprecise data are tackled with fuzzified techniques [10], support-ive evidences are lacking in this regard. Moreover, in such cases thevoices of the customers are not heard in a well-structured manner.

There are several fuzzy-AHP techniques adopted for evaluationof candidate-suppliers in a supply-chain network [19–22,29,41,48].Among fuzzy techniques for supplier selection, an integrated AHP-fuzzy LP model has been reported in the art. Sevkli et al. [74]propose a method of supplier selection combining AHP and fuzzyLP. The weights of the supplier selection criteria are calculated usingthe AHP method and simultaneously those weights are consideredas the weights of the fuzzy LP model. In regard to fuzzification, themodel looks good for supplier selection. But the statement “moreuseful than traditional singular multi-criterion methods” has notbeen justified with evidences. Additionally, the model [74] is notable to hear the voice of customers by integrating qualitative andquantitative criteria.

2.3. Other methods on supplier selection decision

Beside AHP/ANP and fuzzy techniques, studies have beenconducted in the arena of multi-attribute utility theory [37], multi-objective programming [26,27,43] and expert systems [4,86]. Inthese works [4,86] the voices of the purchasers are not the primeissues and, therefore, they are not heard by the DMs. There areexamples [23,25] wherein intelligent system, case based reason-ing and artificial neural network (ANN) tools have been used toevaluate the supplier selection process. Evolutionary fuzzy systems[58], data-mining-based hybrid approach [39,79], expert systems[46,81,91,93], hybrid intelligent algorithm [56] are much appliedin evaluating potential suppliers’ performances for specific tasks.Wang et al. [82] and Jadidi et al. [38] propose a fuzzy hierarchi-cal ‘Technique for Order Preference by Similarity to Ideal Solution’(TOPSIS) methodology for the supplier selection problem. While

applying such soft computing techniques it should be kept inmind that simple decisions, involving merely a few hundred simplecriteria/candidate-alternatives, are not always required to be intel-ligent, unless the decision variables are related in a very complexmanner, as intelligent decisions involve huge costs.

oft Computing 10 (2010) 1013–1027 1015

2

ssmcc

i[omAntir[toeCtpjvhaittw

2

cmbsArbacmtssttmop

stpdnoosnf

Table 1Criteria, sub-criteria and classification of information.

Classification Criteria Sub-criteria

CReq Delivery (DE) –Quality (Q) Factory audit (QFA)

Customer rejection (QCR)Responsiveness (R) On urgent delivery (RUD)

On quality problems (RQP)Management (M) Business skills (MBS)

Attitude (MA)Discipline (D) Honesty (DH)

Procedural compliance(DPC)

Financial position (FP) –

EReq Facility (F) Machinery (FM)Layout (FL)Infrastructure (FI)

A. Bhattacharya et al. / Applied S

.4. Variants of the QFD technique for supplier selection

Variants of integrated QFD techniques [40] have been used inelecting as well as ranking candidate-suppliers. For example, aupplier selection methodology based on extended-QFD and data-ining technique has been proposed [57]. The technique does not

onsider integration of both subjective and objective factors thoughustomer requirement analysis has been focused.

Researchers have proposed the introduction of fuzzy techniquen the HOQ approach (i.e., in QFD) for supplier selection process9]. One variant of QFD teaches an approach for the managementf customer service based on QFD addressing the issue of deploy-ent of HOQ, using fuzzy logic, to improve logistics processes [15].nother variant of the QFD model is the integrated QFD-AHP tech-ique. There have been constant efforts to integrate AHP with QFDo establish a framework for prioritising CReqs [2,11,24]. In one suchntegrated technique, QFD is applied to develop criteria with cor-esponding evaluating weights wherein AHP is used in two phases24]. In the first phase AHP is used to measure the relative impor-ance weighting for each of the location requirements (i.e., CReqs inur case). In the second phase AHP is used to evaluate the score forach of the candidate-alternatives for each particular criterion [24].huang’s model [24] considers only two types of costs as subjec-ive criteria. Another variant [2] focuses on the identification andrioritisation of CReqs. The model tries to integrate AHP with QFD

ust to establish a framework for prioritising CReqs. Another furtherariant [11] combines AHP with QFD, but does not elucidate theierarchical framework taking into consideration both the cardinalnd ordinal preferences. These variants of QFD do not provide anynsights on how to integrate both the subjective and objective cri-eria. Additionally, these variants are not able to demonstrate howhe decision variables respond to a slight variation in the decisioneights.

.5. Research objectives

A survey of the literature reveals the fact that there have beenontinuous efforts to evaluate suppliers by devising numerousethodologies. Some of the literature suggest techniques to com-

ine both cardinal and ordinal preferences during the selection ofuppliers. Many of the cases, wherein the fuzzified techniques andNN methods are used, need to be tested before applying those ineal-world supplier selection decisions. Sometimes, use of a cum-ersome technique may frustrate the DM as well as managers todopt the methodology wherein the decision does not seek suchomplex techniques. However, the available literature do not revealethodology that explicitly helps in categorising the information

o be used during the supplier selection procedure combining bothubjective as well as objective factors and simultaneously aid inclu-ion of the cost factor components involved with such evaluationechnique when customers’ voices are heard adequately. Thus,he objective of this paper is to devise an integrated hierarchical

ethodology for supplier selection and demonstrate the efficacyf the devised hierarchical methodology using data adopted fromublished literature [51] and [89].

In view of the survey of the literature addressing the supplierelection problem using different tools/methods, it has been foundhat there is a constant need to enhance the effectiveness of theurchasing decision. The DM needs to more precisely model theecision situation taking into account both the cardinal and ordi-al preferences. Further, the decision situations should comprise

f a hierarchical framework. Thus, it is required to illustrate thebjectives of each hierarchy in the selection decision. Hierarchicalupport enables verification of redundancy of criteria and alter-atives, if any. Overall, there is a need to propose a method that

acilitates efficient communication between the suppliers and pur-

Technical capabilities(TC)

Product range (TCPR)Ability to solve technicalproblem (TCTP)

chasers. The proposed model should contain a guiding factor thatsenses the responses of the decision variables to a slight variationin the decision weights.

3. Proposed integrated hierarchical QFD methodology

The methodology adopted in this work is directed to evalu-ate as well as rank the candidate-suppliers among a group of 10candidate-alternatives. As delineated in the earlier section, thereare 8 criteria and 13 sub-criteria (factors) for the selection decisionand the criteria are conflicting-in-nature. The devised integratedhierarchical QFD methodology combines both cardinal as well asordinal preferences of the selection decision trading off amongthe criteria and sub-criteria. There are CFMs included in the sup-plier evaluation process from business perspectives. Integration ofsuch CFMs with subjective factors is a complex process and thus, itdemands the devised model to combine the subjective and objec-tive criteria in the form of indices. Thus, the integrated hierarchicalQFD model first trades-off the governing parameters and the out-comes of such selection decisions are found in the form of indices. Inthe next sections both the QFD and the hierarchical MCDM method-ologies are delineated. The integrated model is devised thereafterand an algorithm is presented.

3.1. Quality function deployment (QFD) technique

QFD is a well-structured, cross-functional planning techniquethat is used to hear the customers’ voice throughout the prod-uct planning, development, engineering and manufacturing stagesof any product. QFD, in fact, is a methodology of continuousproduct improvement, emphasising multifunctional teams to inte-grate planning, development, engineering as well as manufacturingstages. This technique helps to measure the impact of organisa-tional learning through innovation [35]. In general, there are threesteps associated in QFD planning ensuring that the voices of cus-tomers are heard: (i) identification of CReqs, (ii) identification ofEReqs and (iii) determination of the importance weights for theindividual CReq. Bhattacharya et al. [11] have located the customer’sorder decoupling point (CODP) wherein the voices of the customershave been heard during a selection decision. The chief aim of theQFD planning should be maximisation of customer satisfaction [83]

through proper identification of CReqs [2]. Thus, identification of theCReqs plays a pivotal role as depicted in Table 1.

There has been much debate with regards to the appropriateimplementation and use of QFD in organisations [16]. Burke et al.[16] argue on the meaningfulness and validity of the final quantita-

1016 A. Bhattacharya et al. / Applied Soft Co

tdathru

[oogtiQa

3

tqpbwta1pi

((

Fig. 1. Supplier-selection ratings based upon Saaty’s nine-point scale.

ive output as the mathematical structure of the QFD calculationsictates certain constraints on the use of its output. It is stated that ifppropriate methods are not used to establish importance weights,he weights cannot be assumed to be on a valid ratio scale in theouse of quality (HOQ) matrix [16]. Extensive analysis of literatureeveals that the HOQ matrix is an almost universal tool that can besed for prioritising most of the tasks of any industry.

The QFD technique is not designed as a general planning process16]. The technique is aimed at planning one specific product aspposed to quantifying the worth of alternatives. As the resultsf QFD are not intended for use as anything more than a generaluideline for choosing priority items [16], it is required to makehe QFD matrix more acceptable than an ordinary “guideline”. Thus,ntegration with the operational research tool, AHP [69], makes theFD technique more acceptable to DMs rather than using the sames a “general guideline”.

.2. Analytic hierarchy process (AHP) methodology

AHP has been developed by Saaty [69,70]. The methodologyrades-off among various qualitative and quantitative factors anduantifies the qualitative factors with a scale called Saaty’s nine-oint scale [68,69,65]. Fig. 1 elucidates supplier-selection ratingsased upon Saaty’s nine-point scale for generating importanceeights. It may be noted that the intermediate importance between

wo adjacent judgements is illustrated with the weights 2, 4, 6nd 8. Very close activities are represented by ratings such as 1.1,.2, etc. when criteria/sub-criteria/candidate-alternatives are com-ared among themselves within the closest importance level. AHP

s an efficient decision-making tool since it incorporates a tech-

mputing 10 (2010) 1013–1027

nique for measuring consistency of the decision made. Additionally,DMs have the choice to have a more exhaustive conceptual com-parison of the different decision components using AHP. It hasbeen seen in many applications of AHP that consideration of var-ious mutually exclusive multivariate criteria guarantees a higherstandard of the solution through AHP and enhanced consistencythroughout the decision-making process.

AHP is the one of the most systematic analytical techniques ofMCDM within the framework of operational research techniquesthat facilitates a rigorous definition of priorities and preferencesof DMs. It is widely used as an analytical tool in various fields ofstudies. Broadly the technique considers the following steps duringmodeling of any system under consideration:

a) defining a site-specific hierarchic structure;b) calculating weights; and

(c) computing inconsistency ratios.

Precisely, AHP has four axioms: (i) reciprocal judgements, (ii)homogeneous elements, (iii) hierarchical or feedback dependentstructure and (iv) rank order expectations [67]. Numerous stud-ies have been conducted on the afore-stated features of AHP.Researchers have identified AHP as a flexible tool to use undersituations demanding ranking of candidate-alternatives based onsubjective criteria. Thus, the properties of AHP enable this researchto hierarchically judge the information as well as categorise themaccording to the needs of customers and suppliers while using thesame in the HOQ.

There have been considerable research efforts [49,66] on assist-ing a DM to detect inconsistencies and to represent DM’s judgmentsproperly in the AHP model. Based upon the construction of themodel as well as the weights allowed by the DMs, AHP may suf-fer from significant cardinal and/or ordinal inconsistencies in itspair-wise comparison matrix thereby making it difficult to rankrationally the candidate-alternatives [49]. On the other hand, ithas been found during extensive AHP applications that AHP hasa logical way to combine individual judgments [59]. Further, it hasbeen established that AHP is theoretically sound, readily under-stood, easily implemented, and capable of producing results thatagree with expectations [31]. Thus, there is a sound basis of usingthe properties of AHP in classifying hierarchically the informationas well as integrating the same with QFD so as to achieve moreprecise and logical ranking of the candidate-suppliers.

3.3. Detailed description of the hierarchical QFD methodology

Using the concept of AHP, a four-level hierarchical structureof the supplier selection decision has been constructed and indi-cated in Fig. 2. The final scores and importance weights have beenreflected in Fig. 2 using the devised methodology demonstrated inSection 5 later. The concept of hierarchical QFD has evolved fromthe efforts of prescribing methodology that enables the DMs tohear the voice of customers as well as classifies the level of hierar-chies according to the goals. In literature, partial efforts have beenreported proposing a fuzzy AHP with an extent analysis approachto determine the importance weights for the CReqs [45] and priori-tising the EReqs in QFD [83].

In this paper, the integration of the hierarchical MCDM methodwith QFD has been made in such a fashion so that the integrationobeys the set of QFD implementation rules suggested by Burke et al.[16] to act as guidelines for building and scoring QFD matrices. The

rules, as set out by Burke et al. [16], are helpful to the QFD analystwhile fitting any MCDM/scoring model appropriately to the QFDmatrix.

Considering all major aspects of the QFD implementation, stressis given in the devised integrated methodology in such a fashion so

A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027 1017

ction

tbtdodTf

fE

3

dwaoreicbcohps

TS

Fig. 2. The four-level hierarchical structure of the supplier sele

hat the voice of purchaser, i.e., the GFSB company, is heard. It wille noticed from the algorithm of the proposed methodology that inwo different phases of the model customers’ voices are heard, i.e.,uring prioritisation of the CReqs and EReqs and during involvementf the CFMs. It has been shown how the outcome of the selectionecision varies with a slight change in the CReqs as well as EReqs.his effect has been indicated with a parameter called “objectiveactor decision weight” measured using ˛.

The hierarchical QFD matrix offers a symbolic scale in order toacilitate the assignment of importance weights for each CReq andReq. The scale is outlined in Table 2.

.3.1. Algorithm of the integrated modelFig. 3 illustrates the algorithm governing the supplier selection

ecision with the devised QFD technique within the AHP frame-ork involving CFMs. In the algorithm (Fig. 3), QFD is also aimed

t planning tasks considering the customers’ requirements. On thether hand, AHP aids a DM to build the hierarchical structure inegard to the criteria in the selection problem. Thus, the synergisticffect of the hierarchical QFD planning algorithm is aimed at cater-ng to the needs of customers by requiring the DM to decide on theriteria, levels of hierarchies and ranking of the candidate-suppliersased on the selection index of the devised model. The hierarchi-al model considers both subjective as well as objective factors

f the selection problem. The very purpose of the QFD techniqueas remained intact; rather the approach of tackling the decisionroblem has been changed. Hierarchical formation of criteria andub-criteria and to fix the goal accordingly is one objective of the

able 2ymbolic representation of importance weights for the hierarchical QFD model.

Symbols Description Assigned weights

� Strong relationship exists 9Moderate relationship exists 5

� Weak relationship exists 1No symbol No relationship exists 0

decision indicating the final scores of the candidate-suppliers.

proposed model. Another objective is to observe the importanceallowed for each criterion, sub-criterion, and their interactions inthe decision hierarchy so as to reach a consensus decision wheremultiple DMs are associated for supplier selection. The approachis significantly different from that proposed by Yahya and Kings-man [89] and Liu and Hai [51] as delineated in the earlier section.Ranking of the candidate-suppliers are based on the values of selec-tion indices (SI) obtained for each candidate-supplier. SIs are thenegotiated values obtained trading-off between subjective factormeasures (SFMs) and objective factor measures (OFMs). OFM con-siders objective factor costs (OFC), and it is a measure to describethe CFM function in supplier selection. The governing relationshipamong SI, SFM and OFM is as follows [11]:

SIj = [(˛ ∗ SFMj) + (1 − ˛) ∗ OFMj] (1)

where OFMj = [OFCj

∑nj=1OFC−1

j ]−1

and 0 ≤ ˛ ≤ 1.In the next section, the hierarchical QFD model will be imple-

mented within the supplier selection framework as posed by Yahyaand Kingsman [89] earlier. In order to illustrate the efficacy of thedevised methodology the number of candidate-suppliers has beenrestricted to 10 in the present work.

4. Case study

This paper attempts to evaluate 10 candidate-suppliers having8 criteria and 13 sub-criteria as is the case of Liu and Hai [51]which has used the same dataset as that of Yahya and Kingsman[89]. Liu and Hai [51] confine their methodology to evaluate 10candidate-suppliers with a voting AHP method, whereas Yahyaand Kingsman [89] evaluate 68 vendor-alternatives with the 8criteria and 13 sub-criteria using AHP. In this paper, the candidate-

suppliers are identified as S1, S2, S3, . . ., and S10 representing a totalof 10 candidate-suppliers.

Yahya and Kingsman’s [89] model does not encompass thecost-constraints for selecting suppliers. There is no competitionamong the potential suppliers in regard to the price as prices for

1018 A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027

with

tc[iscom

4

psiasa

Fig. 3. Algorithm for the QFD technique

he products, i.e., furniture, have already been set by the pur-haser [89]. The same assumption has been set by Liu and Hai51]. However, literature [50] reveals that costs play a crucial rolen selecting suppliers from the service-receiving industry’s per-pective. This paper takes into consideration the ‘price’ (i.e., cost)omponents. Upon careful observation cost components are tracedut which are further utilised in the proposed hierarchical QFDodel.

.1. The multi-criteria nature of supplier selection

Selection of criteria and sub-criteria plays a pivotal role in therocess of ranking as well as the selection of the best candidate-

upplier. Based upon the criteria, sub-criteria and the type ofnformation available, i.e., cardinal, ordinal and whether theyre conflicting-in-nature type, the methodology for the supplierelection process is selected. There are several MCDM as wells multi-attribute decision-making (MADM) models which have

in the AHP framework involving CFMs.

already been tested in supplier selection framework with varietyof criteria and sub-criteria.

It has been much argued that the interaction of various criteriaand sub-criteria is required for selecting a supplier [42,52]. Factorslike quality management factors, price, delivery time and produc-tion lead-time are a few suggested criteria during the selection andranking of a suppliers in a supply-chain network [47]. The hierarchyin quality management factors includes quality management audit,product testing, engineering work force, capability index, trainingtime, etc., based on a five-interval scale [47]. Among many others,cost is a major factor in supplier selection [50].

The 8 criteria and the 13 sub-criteria as depicted by Yahya andKingsman [89] and Liu and Hai [51] are illustrated in Table 1. Sim-

ple notations have been used in order to represent each criterion aswell as sub-criterion. The criteria as well as sub-criteria are furtherclassified into two categories: (i) customer requirements (CReq) and(ii) engineering requirements (EReq). The word ‘customer’ has beenused here to refer to the company which is engaged for the pro-

A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027 1019

Table 3Cost factor measure for supplier selection.

Cost factor components Ranges (Euro)

Material acquisition cost (C1) 7,000–12,000

crwnbqtFsdot

4

ahta

tttm

acpoctapgto

5

e

Order management cost (C2) 200–350Transportation cost (C3) 150–400Information systems cost (C4) 100–200Finance and planning cost (C5) 150–250

urement of furniture for Government. Readers are encouraged toefer to Yahya and Kingsman [89] for further details. EReqs are thosehich are related to the product’s technical orientation, such asumber of products in a group, etc. Moreover, EReqs are the capa-ilities of the suppliers to perform the specified job mentioned in auote. The categorisation of criteria is made based upon fundamen-al requirements and bonus requirements of the selection decision.undamental requirements are those which directly govern theupplier selection whereas the bonus requirements are implicitlyependent on the selection problem. Thus, the sub-criteria as setut by Yahya and Kingsman [89] have been re-arranged to placehem in the QFD matrix.

.2. Supply chain costs

Supply chain costs are complex [87]. Some of the cost elementsre not easily definable. Cost components become complex due toigh number of suppliers. Supply chain costs are crucial to effec-ively manage inventory. The Supply Chain Council (SCOR) outlinesn operational definition of supply chain cost as follows [73]:

“Total Supply Chain Management Cost is a discrete measure-ment defined as the fixed and operational costs associated withthe Plan, Source, Make, and Deliver supply chain processes”.

After a careful consideration of the operational definition of theotal supply chain cost [73], cost factor components have been iden-ified and tabulated in Table 3. The cost components considered forhe evaluation of the suppliers are referred to as the cost factor

easure (CFM).The costs of procuring the products, i.e., furniture in this case,

re the material acquisition costs. Order management cost is theost of order capture, validation, sourcing and distribution. Trans-ortation cost relates to the freight cost as a percentage of revenuer cost of sales. Information systems cost pertains to the costaptured for supplier database maintenance, exchange of informa-ion for procurement and administration activities. It contributesmaximum of 2–3% of the product acquisition cost. Finance andlanning cost is associated with the costs of demand forecasting,eneral planning within supply-chain network, technology, inven-ory financing. Table 4 depicts the breakdown of the CFM for eachf the candidate-suppliers.

. Application to supplier selection problem and results

Yahya and Kingsman [89] have illustrated comparisons amongight criteria for the supplier selection decision. The proposed hier-

Fig. 4. Aggregate importance weights of the CReqs for decision matrix.

archical QFD methodology categorises the criteria and sub-criteriaas CReqs and EReqs. Thus, the decision matrix considers only 6 cri-teria which have been considered as CReqs. The criteria for thedecision matrix are delivery (DE), quality (Q), responsiveness (R),management (M), discipline (D) and financial position (FP). Thedata for the decision matrix have been adopted from Yahya andKingsman [89]. The decision matrix is illustrated in Matrix 1.

The algorithm computes the principal eigenvalue of the decisionmatrix (�max). In order to calculate the value of �max the sum of eachcolumn vector (Qj) is computed, j being the number of rows of thecolumn vector. Overall summation of the product of utility values(UV) with Qj for a given i and j is the �max. This has been indicatedin Eq. (2):

�max =N∑

i=1

Qj ∗ (UV)i, where Qj =n∑

i=1

dij (2)

Mathematically, utility values (UV) are those which have beenreferred to as priority vectors in AHP. Physically, UV refers to themeasure of the relative satisfaction or desirability of consumptionof the service offered by the suppliers. UV reflects the quality ofsuppliers in regard to their gross performances. UVs are directlydependent on the assessment of the criteria/candidate-alternativesin the pair-wise comparison matrices. The higher the UV is, thebetter is the performance of the supplier.

The hierarchical framework of the QFD model allows a consis-tency ratio less than 10% as an acceptable limit. As delineated inFig. 3, this check is conducted with the following set of equations:

CI = �max − n

n − 1(3)

RI = 1.98(n − 2)n

(4)

CR = CIRI

(5)

It is desirable to assess the relative importance of the CReqs witha comparison among utility values of those CReqs. Thus, a com-parative characterization in regard to UVs has been illustrated inFig. 4.

Using Eqs. (2)–(5) for the decision matrix (Matrix 1) the follow-ing values are obtained: �max = 6.403948; CI = 0.0807896; RI = 1.32;CR = 0.0612 = 6.1204%. It may be noticed that the consistency ratioof the decision matrix is less than 10%, which is within the accept-able limit as suggested by Saaty [54].

The 2nd level of the hierarchy of the QFD planning comprises ofseveral sub-customer requirements (sub-CReqs). The data for thepair-wise comparison matrices (Matrix 2–5) to evaluate the sub-CReqs have also been adopted from Yahya and Kingsman [89]. TheUVs have been illustrated beside each of the pair-wise comparison

1020 A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027

Table 4Cost factors for candidate-suppliers.

Cost components Candidate-suppliers

S1 S2 S3 S4 S5 S6 S7 S8 S9 S10

C1 7,100 7,800 7,650 11,750 8,100 8,900 8,400 11,900 9,600 7,800C2 220 210 230 260 320 305 300 350 270 275C3 190 380 315 210 350 310 180 165 150 260C4 165 110 100 150 135 200 125 190 180 140C5 200 155 180 235 170 250 150 190 165 210

mo

A

M

pcapolmI0(

d

level of the hierarchy have been generated to compare the sub-

Fig. 5. Importance weights of candidate-suppliers for facility (F) criterion.

atrices so as to obtain the information on the level of importancef each of the sub-CReqs represented in the matrices:

1 =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎣

1 4 2 5 8 7 6 9 1 21/4 1 2/4 5/4 8/4 7/4 6/4 9/4 1/4 1/21/2 4/2 1 5/2 4 7/2 6/2 9/2 1/2 1/11/5 4/5 2/5 1 8/5 7/5 6/5 9/5 1/5 2/51/8 4/8 1/4 5/8 1 7/8 6/8 9/8 1/8 1/41/7 4/7 2/7 5/7 8/7 1 6/7 9/7 1/7 2/71/6 4/6 2/6 5/6 8/6 7/6 1 9/6 1/6 2/61/9 4/9 2/9 5/9 8/9 7/9 6/9 1 1/9 2/91/1 4 2/1 5/1 8/1 7/1 6/1 9 1 21/2 2 1 5/2 8/2 7/2 6/2 9/2 1/2 1

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎦

atrix 6: Pair-wise comparison matrix for facility (F) criterionThe criteria considered for EReqs of the QFD matrix allow com-

arison among the 10 candidate-suppliers. In order to perform theomparative analysis with their utility values, comparison matricesre set for each EReq. The data for these preference weights are notresent in Yahya and Kingsman [89] and Liu and Hai [51]. Experts’pinions were sought in order to obtain the weights for the fol-owing two EReq matrices. Based upon the opinions a comparison

atrix (A1) for the facility (F) criterion is constructed (Matrix 6).t has been observed that the consistency index for the matrix is

.0159%, which is well below the maximum acceptable level of CIi.e., 10%).

Fig. 5 depicts how the utility values for the facility criterion areistributed among the 10 candidate-suppliers. It may be observed

Fig. 6. Importance weights of candidate-suppliers for TC criterion.

from Fig. 10a that a good preference has been allowed for suppliersS1 and S9 for the facility criterion.

A2 =

⎡⎢⎢⎢⎢⎢⎢⎢⎢⎣

1 2 6 7 4 3 8 5 9 81/2 1 3 7/2 2 3/2 4 5/2 9/2 41/6 1/3 1 7/6 2/3 1/2 4/3 5/6 3/2 4/31/7 2/7 6/7 1 4/7 3/7 8/7 5/7 9/7 8/71/4 1/2 3/2 7/4 1 3/4 2 5/4 9/4 21/3 2/3 2 7/3 4/3 1 8/3 5/3 3 8/31/8 1/4 3/4 7/8 1/2 3/8 1 5/8 9/8 11/5 2/5 6/5 7/5 4/5 3/5 8/5 1 9/5 8/51/9 2/9 2/3 7/9 4/9 1/3 8/9 5/9 1 8/91/8 1/4 3/4 7/8 1/2 3/8 1 5/8 9/8 1

⎤⎥⎥⎥⎥⎥⎥⎥⎥⎦

Matrix 7: Pair-wise comparison matrix for TC criterionSimilar analysis is conducted for technical capability (TC) crite-

rion. Matrix 7 illustrates the comparison matrix (A2) for the TCcriterion. The CI for this matrix is 2.7545%, which is much lessthan 10%. Thus, the judgemental values allowed in matrix A2 arewithin the acceptable limits of the hierarchical model. A compara-tive measure on the importance weights of candidate-suppliers forTC criterion is illustrated in Fig. 6.

As with sub-CReqs, pair-wise comparison matrices for the 2nd

criteria for EReqs. These matrices are illustrated in Matrices 8 and9. The data for these matrices have been adopted from Yahya andKingsman [89]. It is required to observe how the sub-criteria foreach EReq criterion behave with respect to the importance weights

oft Computing 10 (2010) 1013–1027 1021

ars

hdrthijiE

w

TiiiMr

S

TrsnctS

ct(oi

1istic lines as indicated in Fig. 9, and S1 has constantly obtainedthe highest SI value for every ˛. Thus, S1 is ranked as the bestcandidate-supplier. On the other hand, Fig. 9 indicates that S8 isthe least preferred candidate-supplier. Attention is concentrated on

Fig. 7. Scores for each candidate-supplier (when CFMs are not considered).

A. Bhattacharya et al. / Applied S

llowed in the said matrices. This has been illustrated by UVs rep-esenting the importance levels of the aggregate weights for eachub-EReq.

The central relationship matrix for the hierarchical QFD modelas been developed in Matrix 10. This central relationship matrixepicts the relationships among criteria for CReqs and EReqs. Theelationship utilises a symbolic scale as defined in Table 4. Inter-alia,he utility values for CReqs generated using the hierarchical modelave been considered in the central relationship matrix. The matrix

s generated in order to trade-off among the CReqs and EReqs. At thisuncture, Eqs. (6) and (7) [24] are utilised to compute the degree ofmportance (wj) and normalised degree of importance (wj) for theReqs.

Engineeringrequirements (EReq)

Importance weights forcustomer requirements

F TC

Customer requirements (CReq)DE � � 0.4016Q � � 0.2986R � 0.0461M – 0.0853D – – 0.0559FP � – 0.1124Degree of importance for

selection criteria6.6447 3.9304

Normalised degree ofimportance forselection criteria

62.8334 37.1666

Matrix 10: The QFD matrixfor supplier selection

j =m∑

i=1

(Rij ∗ ci) (6)

wj = wj∑nj=1wj

× 100

= SFMj × 100(7)

able 5 illustrates the calculation of the scores, without consider-ng CFMs, for each candidate-supplier. As illustrated in Table 5, themportance weights of the candidate-suppliers for both the facil-ty and technical capability EReq criteria have been obtained from

atrices 10 and 11, respectively. The computation of the scoresequires application of Eq. (8):

j =n∑

j=1

(wj ∗ eij) (8)

he computed scores are tabulated in Table 5. Fig. 7 is an illustrativeepresentation of the relative comparison among the candidate-uppliers with respect to the computed scores when CFMs areot considered. Subsequently, a blueprint on the selection ofandidate-suppliers is formulated based on the initial ranking ofhe combined AHP-QFD model, when CFMs are not considered:1 > S9 > S2 > S3 > S10 > S6 > S5 > S4 > S8 > S7.

In order to incorporate the CFMs, additional information, viz.,

ost related to product acquisition, order management, transporta-ion, information system, finance and planning, are consideredTables 2 and 3). Fig. 8 consolidates the final ranking basedn the hierarchical QFD approach when the CFMs are takennto account as one of the measures for selection of candidate-

suppliers. Finally, based on the SI values the 10 candidate-suppliersare ranked in the following fashion in a descending order:S1 > S9 > S2 > S3 > S10 > S6 > S5 > S7 > S4 > S8. The differences betweenthis final ranking and the previous ranking are clearly visible whenFig. 8 and Fig. 7 are compared and analysed respectively.

The role of the ‘objective factor decision weight’ (˛) is to guide aDM on how much dominance of CFMs may be allowed in the sup-plier selection decision while evaluating the selection indices. Inorder to elucidate the role of ˛ in the selection decision a charac-teristics plot has been configured in Fig. 9 to analyse the effects ofthe selection of ˛ with respect to the selection indices.

As a measure to analyse the effect of the ˛ value on theselection of the best candidate-supplier from a set of 10 candidate-suppliers, Table 6 has been constructed. Table 6 prioritises acandidate-supplier when comparing the same with the associatedcandidate-alternatives within a range of ˛ value. It may be notedthat the candidate-supplier S does not intersect other character-

Fig. 8. Final ranking based on the hierarchical QFD approach.

1022 A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027

Table 5Scores of the hierarchical QFD model.

EReq Weights Importance weights for candidate-suppliers CR (%)

S1 S2 S3 S4 S5 S6 S7 S8 S9 S10

Facility (F) 62.8334 0.2503 0.0626 0.1251 0.0501 0.0313 0.0358 0.0417 0.0287 0.2503 0.1251 0.01585Technical capability (TC) 37.1666 0.3511 0.1638 0.0546 0.0468 0.0878 0.1092 0.0409 0.0655 0.0364 0.0439 2.7545

Scores 28.7764 10.0213 9.8898 4.8874 5.2299 6.3080 4.1403 4.2377 17.0801 9.4921

ttTtildct�gspe

5e

c

TD

Fig. 9. Characteristics of candidate-suppliers with respect to ‘˛’.

he intersecting lines of Fig. 9. Based upon the intersection pointshe limits of ˛ values are framed and accordingly compared inable 6. As an illustrative example, this paper considers ˛ = 0.64o determine the final ranking as depicted in Fig. 8. From Table 6t is found that the supplier-selection decision stabilises when ˛ies between 0.281 and 0.792. If the value of ˛, which is solelyependent on the DM as well as the management of the firm, ishanged according to the needs of the industry the ranking as illus-rated in Fig. 8 is affected. Ranking is dependent on the choice of. It is prudent from both Fig. 9 and Table 6 that these are helpfuluides for the DMs in deciding the value of the objective factor deci-ion weight (˛) for selection decision, which is directed towardsrioritising the candidate-suppliers while comparing with the oth-rs.

.1. Validating the proposed methodology with design ofxperiment

The validation of the proposed hierarchical QFD methodology isonducted with the help of design of experiment. As a representa-

able 6ependence on ˛ in prioritising the candidate-suppliers.

Range of ˛ values Order of preference

0.000 ≤ ˛ < 0.033 S1 – S3 – S2 – S10 – S5 – S7 – S6 – S9 – S4 – S8

0.033 ≤ ˛ < 0.086 S1 – S3 – S2 – S10 – S5 – S7 – S9 – S6 – S4 – S8

0.086 ≤ ˛ < 0.099 S1 – S3 – S2 – S10 – S5 – S9 – S7 – S6 – S4 – S8

0.099 ≤ ˛ < 0.189 S1 – S3 – S2 – S10 – S9 – S5 – S7 – S6 – S4 – S8

0.189 ≤ ˛ < 0.205 S1 – S3 – S2 – S9 – S10 – S5 – S7 – S6 – S4 – S8

0.205 ≤ ˛ < 0.222 S1 – S3 – S9 – S2 – S10 – S5 – S7 – S6 – S4 – S8

0.222 ≤ ˛ < 0.281 S1 – S9 – S3 – S2 – S10 – S5 – S7 – S6 – S4 – S8

0.281 ≤ ˛ < 0.465 S1 – S9 – S3 – S2 – S10 – S5 – S6 – S7 – S4 – S8

0.465 ≤ ˛ < 0.640 S1 – S9 – S3 – S2 – S10 – S6 – S5 – S7 – S4 – S8

0.640 ≤ ˛ < 0.792 S1 – S9 – S2 – S3 – S10 – S6 – S5 – S7 – S4 – S8

0.792 ≤ ˛ < 0.968 S1 – S9 – S2 – S3 – S10 – S6 – S5 – S4 – S7 – S8

0.968 ≤ ˛ < 1.000 S1 – S9 – S2 – S3 – S10 – S6 – S5 – S4 – S8 – S7

Fig. 10. Illustration on the effect of three-factor interaction on the SI values.

tive case an experiment with the following four sample factors andtwo-levels (Table 7) is performed.

The number of run for the above experiment is sixteen. Theexperiment is conducted with Design-Expert® software, version7.1.5. The considered responses to the system of the 2-level factorialdesign are “Hierarchical QFD” [H-QFD] (response 1) and traditionalQFD (response 2). This is an illustrative example with four fac-tors only. One can extend this analysis for the entire eight factorsand thirteen sub-factors. The full design would give the favourableresults indicating the strength of H-QFD over the traditionalQFD.

It is observed from the experiment that the F-value of the H-QFD model is 16545.39 (Table 8). This indicates that the proposedmodel is significant. There is only a 0.01% chance that the F-valueof the model could have a large difference from the current onedue to noise. Further, the F-value of the traditional QFD model is6.72 (Table 9). This implies that the traditional QFD model is notsignificant relative to the noise. However, there is a 13.68% chancethat the F-value could have this large value due to noise.

Interaction among the three criteria, viz., technical capability(TC), facility and cost, are illustrated in Fig. 10. This plot indicateshow the factor TC interacts in combination with the other two fac-tors to affect the response, i.e., SI values. SI value is highest at theA-, B-, C-settings (lower front left corner of the cube). This indicatesthat TC and facility would give poor results at low cost. However,as seen from the plot, the best setting is A+, C−, B+ which wouldgive high TC and facility at low cost. The interactions among othercriteria can be interpreted in the same manner so as to monitor theeffect of response on the governing criteria.

Sample representations comparing the two methodologies (i.e.,responses R1 and R2) are illustrated in Figs. 11 and 12. The illustra-tions guide a decision maker to select the appropriate methodologysuiting the criteria. It has been observed from such comparativeplots that the hierarchical methodology (i.e., response R1) is best

suited to trade off among the eight criteria and thirteen sub-criteria.Therefore, the H-QFD model is superior over the traditional QFDmethod.

A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027 1023

Table 72-Level factorial design.

Factors Units Lower level Higher level

A: Technical capabilities (TC) Non-dimensional unit 0.03639549 0.35109638B: Facility (F) Non-dimensional unit 0.02780695 0.25027713C: Cost Euro 7875 12,795D: Responsiveness (R) Non-dimensional unit 0.41323616 0.58676384

Table 8Analysis of variance table for response 1 (R1) [partial sum of squares—type III; transform: square root; constant: 0].

Source Sum of Squares df Mean Sqaure F-Value p-Value prob > F

Model 0.076 13 5.857E−003 16545.39 <0.0001a

A: Technical capability 2.723E−003 1 2.723E−003 7692.65 0.0001B: Facility (F) 6.307E−003 1 6.307E−003 17818.46 <0.0001C: Cost 4.202E−003 1 4.202E−003 11870.02 <0.0001D: Responsiveness (R) 8.076E−004 1 8.076E−004 2281.62 0.0004AC 0.024 1 0.024 66601.95 <0.0001AD 4.863E−003 1 4.863E−003 13738.76 <0.0001BC 0.014 1 0.014 39314.40 <0.0001BD 3.896E−003 1 3.896E−003 11005.39 <0.0001CD 9.939E−003 1 9.939E−003 28077.23 <0.0001ABD 4.253E−004 1 4.253E−004 1201.55 0.0008ACD 2.268E−003 1 2.268E−003 6406.81 0.0002BCD 9.453E−004 1 9.453E−004 2670.41 0.0004ABCD 2.269E−003 1 2.269E−003 6410.85 0.0002

6

6tascohttoinlDh

TA

Residual 7.080E−007 2

Cor total 0.076 15

a Significant.

. Discussion

Yahya and Kingsman [89] have used the AHP method to assess8 suppliers (vendors). They have established vendor rating cri-eria with participation of 16 respondents who were managersnd supervisors of GFSB company. Much stress has been put onelection of the criteria relevant for the supplier selection pro-ess. The procedure adopted by Yahya and Kingsman [89] is onef a kind of group decision-making methods in which suppliersave been rated using individual’s voting method and the AHPechnique. Their evaluation process does not consider the cus-omer’s voice. Yahya and Kingsman’s [89] model is not capablef incorporating objective factors like cost components. Surpris-ngly their model does not consider objective cost factors. The

umber of DMs involved with the selection process is 16. The

evel of vagueness increases with the increase in the number ofMs. Therefore, the vagueness level in the decision is considerablyigh.

able 9nalysis of variance table for response 2 (R2) [partial sum of squares—type III; transform

Source Sum of squares df

Model 0.12 13A: Technical capability 4.444E−003 1B: Facility (F) 0.029 1C: Cost 0.015 1D: Responsiveness (R) 5.026E−003 1AB 0.015 1AC 0.011 1AD 6.014E−003 1BC 5.006E−003 1BD 5.333E−004 1CD 0.016 1ABD 4.733E−003 1ACD 4.239E−003 1BCD 2.779E−003 1Residual 2.720E−003 2

Cor total 0.12 15

a Significant.

3.540E−007

On the other hand, Liu and Hai [51] evaluated 10 candidate-suppliers using a voting AHP method combining data envelopmentanalysis with AHP. The customers’ voice is not heard in this case too.They have considered the “priority votes” for the 8 criteria from 60respondents. This is, again, another example of tackling the selec-tion decision with a group decision-making technique. As statedabove, this allows induction of higher vagueness in the decision.Moreover, no illustrative selection decision has been indicated byLiu and Hai [51] though the first supplier’s (i.e., S1) rating has onlybeen calculated using their methodology. This does not validate themethodology proposed by Liu and Hai [51]. No indication on how touse cost factor components, if the company would like to introducethem, has been illustrated.

Under these circumstances, this paper devises the hierarchi-

cal QFD methodology and subsequently validates the model withthe datasets adapted from Yahya and Kingsman [89]. A clear pic-ture of the ranking as well as obtained scores has been elucidated.Hierarchical QFD is a scoring model that enables DMs to clas-

: square root; constant: 0].

Mean sqaure F-Value p-Value Prob > F

9.142E−003 6.72 0.1368a

4.444E−003 3.27 0.21240.029 21.47 0.04350.015 11.12 0.07945.026E−003 3.70 0.19450.015 10.94 0.08050.011 7.95 0.10626.014E−003 4.42 0.17025.006E−003 3.68 0.19515.333E−004 0.39 0.59510.016 11.81 0.07534.733E−003 3.48 0.20314.239E−003 3.12 0.21952.779E−003 2.04 0.28911.360E−003

1024 A. Bhattacharya et al. / Applied Soft Computing 10 (2010) 1013–1027

etwe

sptapfm[uoocaplhhet

Fig. 11. 3D plot for the interaction b

ify the information according to the needs of customers. Theroposed model allows the incorporation of the voice of the cus-omers. Moreover, it trades-off as well as combines both cardinalnd ordinal factors. The ‘objective factor decision weight’ plays aivotal role in combining both the subjective factors and the costactor components. The characteristic plot helps a DM in deter-

ining the most prioritised candidate-supplier. Like Liu and Hai51], there have been 10 candidate-suppliers who are evaluatedsing the hierarchical QFD model. In the proposed model, the rolef AHP is to make the supplier selection problem a hierarchicalne. It is a standard practice in AHP that the criteria and sub-riteria are placed and built hierarchically based on the numbernd sub-divisions of such criteria and sub-criteria. In the sup-lier selection decision there are four levels of hierarchies, each

evel having a specific target. The utility values for each level of

ierarchy have been indicated by comparing the sub-criteria. Theierarchy contains miscellanea of both customer requirements andngineering requirements as illustrated in Table 1. It is to be notedhat the datasets allowed the consistency indices of all the matri-

Fig. 12. 3D plot for the interaction betwe

en factors ‘B’ and ‘C’ for response 1.

ces to remain within the permissible limit prescribed by Saaty[69]. Moreover, the traditional approaches face difficulties to tacklemultiple, conflicting-in-nature criteria having incommensurableunit of measurements wherein customers’ voices are heard andjudged subsequently. In this juncture a hierarchical MCDM toollike AHP is best suited to deal with the conflicting-in-nature cri-teria.

Liu and Hai [51] have tackled the same problem of Yahya andKingsman [89] utilising a different model. The previous two worksdo not consider the importance of cost factor measures. As the sup-plier selection problem as posed by Yahya and Kingsman [89] is areal-life problem, the solutions using the hierarchical QFD modelhave been organised in two parts, one without CFMs and the otherwith CFMs. These two different sets of solutions allow investiga-tion of the behaviour of the ranking of the candidate-suppliers if

GFSB had also considered the CFMs. Thus, the proposed hierar-chical QFD model is capable of handling datasets both with CFMsand without CFMs. Another benefit of using the QFD matrix is thatthe planning process utilises a central relationship matrix distinctly

en factors ‘B’ and ‘C’ for response 2.

oft Co

dE

6

aTImIicfitbmfMmrpCoA

aismaceatn

7

isyprmaslaa

psitmstrd

gn

[

[

[

[

[

[

[

[

[

[

A. Bhattacharya et al. / Applied S

emonstrating the fundamental relationships among the CReqs andReqs.

.1. Future trends of research

A diversified range for future direction of research is avail-ble for the presented methodology for supplier selection decision.he presented methodology involves 8 criteria and 13 sub-criteria.t is, thus required to investigate the behaviour of the proposed

odel when the selection problem involves more than 8 criteria.f such a situation exists, the level of the hierarchy needs to bencreased in order to suitably adapt as well as allocate the newriteria in the hierarchy. Moreover, the components of the costactor measures can be optimised using OR techniques like lex-cography goal programming. If activities are established amonghe cost components, a technique called activity based costing maye suitable for integrating with the methodology. The presentedodel does not elucidate the level-of-satisfaction of the DM. There-

ore, it is another issue to address as a further research work.oreover, experimentation with ANP in the proposed integratedodel is another issue which would provide scope for further

esearch. Accordingly, a different kind of the supplier selectionroblem should be used during experimentation with a ANP-QFD-FM model so as to judge the inter-relationship among the clusterf elements connected in a network while using the features ofNP.

Sometimes, vagueness, risks and uncertainties are introduced insupplier selection decision thereby involving a complex scenario

n such decisions. In such situations fuzzy techniques are founduitable to address vague information. The proposed model can beade more flexible by integrating the self-adaptive capability usingproperly validated and appropriate neuro-fuzzy hybrid model. A

omprehensive Bayesian approach and Dempster–Shafer theory ofvidence may be fruitful to integrate with the presented model insuitable manner when the supplier selection decision encoun-

ers a very high range of risk factors in a complicated value-chainetwork.

. Conclusions

The devised concurrent engineering methodology allows thell-structured supplier selection problem to be tackled in a moretructured hierarchical manner. It is evident from the results, anal-sis and the discussion outlined in the previous sections that therocess presented in this paper is a feasible, useful and practical foranking candidate-suppliers in real-world situations. The proposedethodology is unique in the sense that an OR approach has been

ssociated with QFD so as to hear the voice of the customers in aupply chain scenario. The strength of the proposed methodologyies in allowing DMs to avoid other cumbersome techniques whensimple, purchaser/supplier-friendly efficient algorithm is readilyvailable.

The hierarchical QFD method enhances the effectiveness of theurchasing decisions for the company. It assists the purchaser inolving the right problem. The integrated method is so flexible thatt can accommodate more criteria and sub-criteria if desired. Fur-her, the method provides for the space for new hierarchies that

ay result from the inclusion of any new and relevant criteria andub-criteria for purchasing decisions. But adequate care must beaken while adding up criteria and sub-criteria so that issues of rank

eversal and preference reversal within the hierarchical frameworko not arise.

The proposed methodology allows the purchaser to segre-ate the criteria according to the problem needs. This filters outon-relevant criteria and sub-criteria while making purchasing

[

[

mputing 10 (2010) 1013–1027 1025

decisions and enable the DM to check redundancy of criteriaand candidate-alternatives. Moreover, the proposed model aidsthe purchaser to model the decision situation more preciselyconsidering specifically both the intangible and tangible factors.Incorporation of the groups’ decisions is one of the flexibilities ofthe proposed methodology. The model allows the DM to visualisethe decision variables’ responses to a slight variation in the deci-sion weights. Enhancement of the computation time by automatedcomputation and faster analysis of decision-making information ispossible with the proposed model. Finally, the proposed hierarchi-cal QFD methodology facilitates efficient communication betweenthe suppliers and purchasers.

Acknowledgement

This research has been funded by the Embark Initiative Post-doctoral Fellowship Scheme from the Irish Research Council forScience, Engineering and Technology under grant No. PD/2007/9.

References

[1] Y. Akao, G.H. Mazur, The leading edge in QFD: past, present and future, Inter-national Journal of Quality & Reliability Management 20 (1) (2003) 20–35.

[2] R.L. Armacost, P.J. Componation, M.A. Mullens, W.W. Swart, An AHP frame-work for prioritizing customer requirements in QFD: an industrialized housingapplication, IIE Transactions 26 (4) (1994) 72–79.

[3] G. Barbarosoglu, T. Yazgac, An application of the analytic hierarchy process tothe supplier selection problem, Production and Inventory Management Journal1st Quarter (1997) 14–21.

[4] S.B. Barla, A case study of supplier selection for lean supply by using a mathe-matical model, Logistics Information Management 16 (6) (2003) 451–459.

[5] O. Bayazit, Use of analytic network process in vendor selection decisions,Benchmarking: An International Journal 13 (5) (2006) 566–579.

[6] O. Bayazit, B. Karpak, An AHP application in vendor selection, in: ISAHP 2005,Honolulu, Hawaii, July 8–10, 2005, pp. 1–24.

[7] M.Y. Bayrak, N. Celebi, H. Taskin, A fuzzy approach method for supplier selec-tion, Production Planning & Control 18 (1) (2007) 54–63.

[8] W. Bei, S. Wang, J. Hu, An analysis of supplier selection in manufacturing supplychain management, International Conference on Service Systems and ServiceManagement 2 (2006) 1439–1444.

[9] M. Bevilacqua, F.E. Ciarapica, G. Giacchetta, A fuzzy-QFD approach to supplierselection, Journal of Purchasing & Supply Management 12 (2006) 14–27.

10] M. Bevilacqua, A. Petroni, From traditional purchasing to supplier manage-ment: a fuzzy logic-based approach to supplier selection, International Journalof Logistics Research and Applications 5 (3) (2002) 235–255.

11] A. Bhattacharya, B. Sarkar, S.K. Mukherjee, Integrating AHP with QFD for robotselection under requirement perspective, International Journal of ProductionResearch 43 (17) (2005) 3671–3685.

12] K.S. Bhutta, F. Huq, Supplier selection problem: a comparison of the total costof ownership and analytic hierarchy process approaches, Supply Chain Man-agement: An International Journal 7 (3) (2002) 126–135.

13] L. de Boer, L.L.M. van der Wegen, Practice and promise of formal supplierselection: a study of four empirical cases, Journal of Purchasing & Supply Man-agement 9 (3) (2003) 109–118.

14] L. de Boer, E. Labro, P. Morlacchi, A review of methods supporting supplierselection, European Journal of Purchasing & Supply Management 7 (2) (2001)75–89.

15] E. Bottani, A. Rizzi, Strategic management of logistics service: a fuzzy QFDapproach, International Journal of Production Economics 103 (2006) 585–599.

16] E. Burke, J.M. Kloeber Jr., R.F. Deckro, Using and abusing QFD scores, QualityEngineering 15 (1) (2002) 9–21.

17] F. Cebi, D. Bayraktar, An integrated approach for supplier selection, LogisticsInformation Management 16 (6) (2003) 395–400.

18] F.T.S. Chan, Interactive selection model for supplier selection process: ananalytical hierarchy process approach, International Journal of ProductionResearch 41 (15) (2003) 3549–3579.

19] F.T.S. Chan, N. Kumar, M.K. Tiwari, H.C.W. Lau, K.L. Choy, Global supplier selec-tion: a fuzzy-AHP approach, International Journal of Production Research 46

(14) (2008) 3825–3857.

20] Y.-M. Chen, P.-N. Huang, Bi-negotiation integrated AHP in suppliers selection,Benchmarking: An International Journal 14 (5) (2007) 575–593.

21] Y.-M. Chen, P.-N. Huang, Bi-negotiation integrated AHP in suppliers selection,International Journal of Operations & Production Management 27 (11) (2007)1254–1274.

1 oft Co

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

[

[

[

[

[

026 A. Bhattacharya et al. / Applied S

22] S.-Y. Chou, C.-Y. Shen, Y.-H. Chang, Vendor selection in a modified re-buysituation using a strategy-aligned fuzzy approach, International Journal of Pro-duction Research 45 (14) (2007) 3113–3133.

23] K.L. Choy, W.B. Lee, H. Lau, D. Lu, V. Lo, Design of an intelligent supplier relation-ship management system for new product development, International Journalof Computer Integrated Manufacturing 17 (8) (2004) 692–715.

24] P.-T. Chuang, Combining the analytic hierarchy process and quality functiondeployment for a location decision from a requirement perspective, Interna-tional Journal of Advanced Manufacturing Technology 18 (2001) 842–849.

25] R. Davidrajuh, Modelling and implementation of supplier selection proceduresfor e-commerce initiatives, Industrial Management & Data Systems 103 (1)(2003) 28–38.

26] E.A. Demirtas, Ö. Üstün, Analytic network process and multi-period goalprogramming integration in purchasing decisions, Computers & IndustrialEngineering 56 (2) (2009) 677–690.

27] E.A. Demirtas, Ö. Üstün, An integrated multiobjective decision making processfor supplier selection and order allocation, Omega 36 (1) (2008) 76–90.

28] G. Dickson, An analysis of vendor selection systems and decisions, Journal ofPurchasing 2 (1) (1966) 5–17.

29] I. Dogan, U. Sahin, Supplier selection using activity-based costing and fuzzypresentworth techniques, Logistics Information Management 16 (6) (2003)420–426.

30] R. Dulmin, V. Mininno, Supplier selection using a multi-criteria decision aidmethod, Journal of Purchasing & Supply Management 9 (2003) 177–187.

31] E.H. Forman, S.I. Gass, The analytic hierarchy process: an exposition, OperationsResearch 49 (4) (2001) 469–486.

32] C. Gencer, D. Gürpinar, Analytic network process in supplier selection: a casestudy in an electronic firm, Applied Mathematical Modelling 31 (11) (2007)2475–2486.

33] M.E. Gonzaılez, G. Quesada, C.A.M. Monge, Determining the importance of thesupplier selection process in manufacturing: a case study, International Journalof Physical Distribution & Logistics Management 34 (6) (2004) 492–504.

34] S.H. Ghodsypour, C. O’Brien, A decision support system for supplier selectionusing an integrated analytic hierarchy process and linear programming, Inter-national Journal of Production Economics 56–57 (1998) 199–212.

35] C.P.M. Govers, QFD not just a tool but a way of quality management, Interna-tional Journal of Production Economics 69 (2) (2001) 151–159.

36] A.F. Guneri, A. Yucel, G. Ayyildiz, An integrated fuzzy-LP approach for a sup-plier selection problem in supply chain management, Expert Systems withApplications 36 (5) (2009) 9223–9228.

37] S.H. Huang, H. Keskar, Comprehensive and configurable metrics for sup-plier selection, International Journal of Production Economics 105 (2) (2007)510–523.

38] O. Jadidi, T.S. Hong, F. Firouzi, R.M. Yusuff, N. Zulkifli, TOPSIS and fuzzymulti-objective model integration for supplier selection problem, Journal ofAchievements in Materials and Manufacturing Engineering 31 (2) (2008)762–769.

39] V. Jain, S. Wadhwa, S.G. Deshmukh, Supplier selection using fuzzy associa-tion rules mining approach, International Journal of Production Research 45(6) (2007) 1323–1353.

40] J.-M. Ju, S.-G. Hwang, A study on supplier evaluation and selection methodbased on dependence, Journal of Advanced Computational Intelligence andIntelligent Informatics 8 (4) (2004) 415–420.

41] C. Kahraman, U. Cebeci, Z. Ulukan, Multi-criteria supplier selection using fuzzyAHP, Logistics Information Management 16 (6) (2003) 382–394.

42] G. Kannan, A.N. Haq, Analysis of interactions of criteria and sub-criteria for theselection of supplier in the built-in-order supply chain environment, Interna-tional Journal of Production Research 45 (17) (2007) 3831–3852.

43] A. Kokangul, Z. Susuz, Integrated analytical hierarch process and mathematicalprogramming to supplier selection problem with quantity discount, AppliedMathematical Modelling 33 (3) (2009) 1417–1429.

44] T.J. Kull, S. Talluri, A supply risk reduction model using integrated multicriteriadecision making, IEEE Transactions on Engineering Management 55 (3) (2008)409–419.

45] C.K. Kwong, H. Bai, Determining the importance weights for the customerrequirements in QFD using a fuzzy AHP with an extent analysis approach, IIETransactions 35 (7) (2003) 619–626.

46] C.K. Kwong, W.H. Ip, J.W.K. Chan, Combining scoring method and fuzzy expertsystems approach to supplier assessment: a case study, Integrated Manufac-turing Systems 13 (7) (2002) 512–519.

47] M.-S. Lee, Y.-H. Lee, C.-S. Jeong, A high-quality-supplier selection model forsupply chain management and ISO 9001 system, Production Planning & Control14 (3) (2003) 225–232.

48] A.H.I. Lee, A fuzzy supplier selection model with the consideration of benefits,opportunities, costs and risks, Expert Systems with Applications 36 (2) (2009)2879–2893, Part 2.

49] H.-L. Li, L.-C. Ma, Detecting and adjusting ordinal and cardinal inconsisten-cies through a graphical and optimal approach in AHP models, Computers &Operations Research 34 (2007) 780–798.

50] R.J. Linn, F. Tsung, L.W.C. Ellis, Supplier selection based on process capability

and price analysis, Quality Engineering 18 (2) (2006) 123–129.

51] F.-H.F. Liu, H.L. Hai, The voting analytic hierarchy process method for selectingsupplier, International Journal of Production Economics 97 (2005) 308–317.

52] A. Mandal, S.G. Deshmukh, Vendor selection using interpretive structural mod-elling (ISM), International Journal of Operations & Production Management 14(6) (1994) 52–59.

[

[

[

mputing 10 (2010) 1013–1027

53] M. Marufuzzaman, K.B. Ahsan, K. Xing, Supplier selection and evaluationmethod using analytical hierarchy process (AHP): a case study on an apparelmanufacturing organisation, International Journal of Value Chain Management3 (2) (2009) 224–240.

54] C. Masella, A. Rangone, A contingent approach to the design of vendor selectionsystems for different types of co-operative customer/supplier relationships,International Journal of Operations & Production Management 20 (1) (2000)70–84.

55] H. Min, International supplier selection: a multi-attribute utility approach,International Journal of Physical Distribution & Logistics Management 24 (5)(1994) 24–33.

56] M.R.S. Moghadam, A. Afsar, B. Sohrabi, Inventory lot-sizing with supplier selec-tion using hybrid intelligent algorithm, Applied Soft Computing, in press.

57] M. Ni, X. Xu, S. Deng, Extended QFD and data-mining-based methods forsupplier selection in mass customization, International Journal of ComputerIntegrated Manufacturing 20 (2–3) (2007) 280–291.

58] R. Ohdar, P.K. Ray, Performance measurement and evaluation of suppliers insupply chain: an evolutionary fuzzy-based approach, Journal of ManufacturingTechnology Management 15 (8) (2004) 723–734.

59] M.S. Ozdemir, T.L. Saaty, The unknown in decision making: what to do aboutit, European Journal of Operational Research 174 (2006) 349–359.

60] S. Percin, An application of the integrated AHP-PGP model in supplier selection,Measuring Business Excellence 10 (4) (2006) 34–49.

61] W.-N. Pi, C. Low, Supplier evaluation and selection via Taguchi loss functionsand an AHP, International Journal of Advanced Manufacturing Technology 27(5–6) (2006) 625–630.

62] M.E. Porter, Competitive Advantage: Creating and Sustaining Superior Perfor-mance, Free Press, New York, 1985.

63] R. Ramanathan, Supplier selection problem: integrating DEA with theapproaches of total cost of ownership and AHP, Supply Chain Management:An International Journal 12 (4) (2007) 258–261.

64] J. Razmi, H. Rafiei, M. Hashemi, Designing a decision support system to eval-uate and select suppliers using fuzzy analytic network process, Computers &Industrial Engineering 57 (4) (2009) 1282–1290.

65] T.L. Saaty, Decision making with the analytic hierarchy process, InternationalJournal of Services Sciences 1 (1) (2008) 83–98.

66] T.L. Saaty, Decision-making with the AHP: why is the principal eigenvectornecessary, European Journal of Operational Research 145 (2003) 85–91.

67] T.L. Saaty, How to make a decision: the analytic hierarchy process, Interfaces24 (6) (1994) 19–43.

68] T.L. Saaty, The Analytic Hierarchy Process, revised edn, University of PittsburghPress, Pittsburgh, PA, 1988.

69] T.L. Saaty, The Analytic Hierarchy Process, McGraw-Hill, New York, 1980.70] T.L. Saaty, A scaling method for priorities in hierarchical structures, Journal of

Mathematical Psychology 15 (1977) 234–281.71] T.L. Saaty, Decision Making in Complex Environments: The Analytic Network

Process for Decision Making with Dependence and Feedback, RWS Publications,Pittsburgh, PA, 2001.

72] J. Sarkis, S. Talluri, A model for strategic supplier selection, The Journal of SupplyChain Management: A Global Review of Purchasing and Supply (Winter) (2002)18–28.

73] SCE Limited, SCOR Metrics—Total Supply Chain Management Cost, 2008,http://www.scelimited.com/totalsupplychainmanagementcost.html, accessed07.08.08.

74] M. Sevkli, S.C.L. Koh, S. Zaim, M. Demirbag, E. Tatoglu, Hybrid analytical hier-archy process model for supplier selection, Industrial Management & DataSystems 108 (1) (2008) 122–142.

75] M. Sonmez, A review and critique of supplier selection process and practices,in: Business School Occasional Papers Series, Paper 2006:1, ISBN 1 85901 1977,Loughborough University, United Kingdom, 2006.

76] Sreekumar, S.S. Mahapatra, A fuzzy multi-criteria decision making approachfor supplier selection in supply chain management, African Journal of BusinessManagement 3 (4) (2009) 168–177.

77] F. Tahriri, M.R. Osman, A. Ali, R.M. Yusuff, A. Esfandiary, AHP approach for sup-plier evaluation and selection in a steel manufacturing company, Journal ofIndustrial Engineering and Management 01 (02) (2008) 54–76.

78] F. Tahriri, M.R. Osman, A. Ali, R.M. Yusuff, A review of supplier selection methodsin manufacturing industries, Suranaree Journal of Science and Technology 15(3) (2008) 201–208.

79] T.-L. Tseng, C.-C. Huang, F. Jiang, J.C. Ho, Applying a hybrid data-miningapproach to prediction problems: a case of preferred suppliers predic-tion, International Journal of Production Research 44 (14) (2006) 2935–2954.

80] G. Tuzkaya, A. Ozgen, D. Ozgen, U.R. Tuzkaya, Environmental performanceevaluation of suppliers: a hybrid fuzzy multi-criteria decision approach,International Journal of Environmental Science and Technology 6 (3) (2009)477–490.

81] R.J. Vokurka, J. Choobineh, L. Vadi, A prototype expert system for the evalua-tion and selection of potential suppliers, International Journal of Operations &Production Management 16 (12) (1996) 106–127.

82] J.-W. Wang, C.-H. Cheng, H. Kun-Cheng, Fuzzy hierarchical TOPSIS for supplierselection, Applied Soft Computing, in press.

83] G.S. Wasserman, On how to prioritize design requirements during the QFDplanning process, IIE Transactions 25 (3) (1993) 59–65.

84] C.A. Weber, J.R. Current, W.C. Benton, Vendor selection criteria and methods,European Journal of Operational Research 50 (1) (1991) 2–18.

oft Co

[

[

[

[

[

[

[

[92] M.A. Youssef, M. Zairi, B. Mohanty, Supplier selection in an advanced manu-facturing technology environment: an optimization model, Benchmarking for

A. Bhattacharya et al. / Applied S

85] C.A. Weber, J.R. Current, A. Desai, Non-cooperative negotiation strategies forvendor selection, European Journal of Operational Research 108 (1) (1998)208–223.

86] C.A. Weber, L.M. Ellram, Supplier selection using multi-objective program-ming: a decision support system approach, International Journal of PhysicalDistribution & Logistic Management 23 (2) (1993) 3–14.

87] C. Wu, Total supply chain cost model. MBA & MS (Electrical Engineering) the-sis, Sloan School of Management & Department of Electrical Engineering and

Computer Science, MIT, 2005.

88] W. Xia, Z. Wu, Supplier selection with multiple criteria in volume discountenvironments, Omega 35 (5) (2007) 494–504.

89] S. Yahya, B. Kingsman, Vendor rating for an entrepreneur development pro-gramme: a case study using the analytic hierarchy process method, Journal ofOperational Research Society 50 (1999) 916–930.

[

mputing 10 (2010) 1013–1027 1027

90] C.-C. Yang, B.-S. Chen, Supplier selection using combined analytical hierar-chy process and grey relational analysis, Journal of Manufacturing TechnologyManagement 17 (7) (2006) 926–941.

91] I.H. Yigin, H. Taskin, I.H. CedImoglu, B. Topal, Supplier selection: anexpert system approach, Production Planning & Control 18 (1) (2007) 16–24.

Quality Management & Technology 3 (4) (1996) 60–72.93] D. Zhang, J. Zhang, K.-K. Lai, Y. Lu, An novel approach to supplier selection based

on vague sets group decision, Expert Systems with Applications 36 (5) (2009)9557–9563.