bahinipati_2009_horizontal collaboration in semiconductor manufacturing industry supply chain

16
Horizontal collaboration in semiconductor manufacturing industry supply chain: An evaluation of collaboration intensity index Bikram K. Bahinipati * , Arun Kanda, S.G. Deshmukh Department of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi 110 016, India article info Article history: Received 5 June 2008 Received in revised form 14 November 2008 Accepted 4 March 2009 Available online 12 March 2009 Keywords: Horizontal collaboration Semiconductor industry supply chain AHP–FLM Compatibility test abstract This paper aims to provide a generic quantitative model to comprehensively assess the degree of collab- oration with individual horizontal collaboration initiatives with a view to check feasibility for satisfying the customer requirements. The analytic hierarchy process–fuzzy logic model (AHP–FLM) approach is chosen for developing the model, a method that is often used to tackle complex strategic decision making that calls for subjective judgment based on well-established logical reasoning, rather than on simple feel- ing and intuition. In the process, the complex and unstructured problem for ‘compatibility test’ is broken down into elements, and then a customized hierarchy structure is set up to demonstrate the relationship between different hierarchy levels and among these elements. Each element may have a different level of importance for the horizontal collaboration. A fuzzy rule based collaboration intensity index (CII) is developed to build up the relationships among these evaluation attributes. Synthesizing the generic rel- ative importance and the forecasted degree of collaboration, the proposed approach can determine the success of the collaboration initiative. An illustrative example of a semiconductor industry supply chain (SSC) member that intends to partner with a potential and competing candidate enterprise is developed to demonstrate the applicability of the proposed fuzzy strategic alliance selection framework and to mea- sure the effectiveness of a horizontal collaboration initiative. Ó 2009 Elsevier Ltd. All rights reserved. 1. Introduction Collaborations in supply network mean long-term relationships among members through reductions in transaction costs, and in- crease in resource sharing, learning, and sharing of knowledge (Cousins, 2002). Horizontal collaboration occurs between partners at the same level in the manufacturing process, where the benefits of collaborative manufacturing/purchasing include lower prices due to aggregated manufacturing/purchasing quantities, reduced supply risk, reduced administration cost due to centralized pur- chasing activities, and networking benefits as group members communicate and interact with each other (Tella & Virolainen, 2005). However, such an initiative may be subjected to a number of disadvantages, such as (a) loss of flexibility, as products pur- chased must have a high similarity among group members, (b) loss of control by individual SC members, (c) high coordination costs, as group members are competitors, (d) anti-trust problems, and (e) potential consolidation of the supply market in the long run. Fur- ther, the success of such an initiative depends on quality of leader- ship or coordination in the group, as the ability to negotiate contracts and coordinate members’ interests is critical. We define horizontal collaboration as a business agreement be- tween two or more companies at the same level in the supply chain (SC) or network in order to allow greater ease of work and cooperation towards achieving a common objective. This may be achieved through proper manipulation, utilization and sharing of appropriate resources, such as machinery, technology and man- power. Such initiative is highly desirable for the semiconductor industries to ensure global business opportunity (Cruijssen, Cools, & Dullaert, 2007). In their drive to be more efficient and competi- tive, the semiconductor industries have focused on the internal organization and processes. Due to more competitive pressure, these companies are now looking externally beyond the bound- aries of their own organizations and value chains for horizontal and vertical collaboration with supply chain partners to achieve the benefits in long run. Horizontal collaboration completely exploits the conceptualiza- tion of SCs as supply networks (Mason, Lalwani, & Boughton, 2007). One of the key ways to value creations is the effective deployment and sharing of resources. As the semiconductor indus- tries are involved in these concepts of multi-relations in a network, these are often managed not as a cohesive whole, but as self evolv- ing systems. This introduces the ideas behind reducing the number of contact points so that they can be better managed, and the seg- mentation of relationships to determine different levels of collabo- rative actions. However, if they can be conceptualized as a 0360-8352/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.cie.2009.03.003 * Corresponding author. Tel.: +91 9818183281. E-mail address: [email protected] (B.K. Bahinipati). Computers & Industrial Engineering 57 (2009) 880–895 Contents lists available at ScienceDirect Computers & Industrial Engineering journal homepage: www.elsevier.com/locate/caie

Upload: adrian-serrano-hernandez

Post on 29-Jan-2016

12 views

Category:

Documents


0 download

DESCRIPTION

Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

TRANSCRIPT

Page 1: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

Computers & Industrial Engineering 57 (2009) 880–895

Contents lists available at ScienceDirect

Computers & Industrial Engineering

journal homepage: www.elsevier .com/ locate/caie

Horizontal collaboration in semiconductor manufacturing industry supply chain:An evaluation of collaboration intensity index

Bikram K. Bahinipati *, Arun Kanda, S.G. DeshmukhDepartment of Mechanical Engineering, Indian Institute of Technology Delhi, New Delhi 110 016, India

a r t i c l e i n f o a b s t r a c t

Article history:Received 5 June 2008Received in revised form 14 November 2008Accepted 4 March 2009Available online 12 March 2009

Keywords:Horizontal collaborationSemiconductor industry supply chainAHP–FLMCompatibility test

0360-8352/$ - see front matter � 2009 Elsevier Ltd. Adoi:10.1016/j.cie.2009.03.003

* Corresponding author. Tel.: +91 9818183281.E-mail address: [email protected] (B.K

This paper aims to provide a generic quantitative model to comprehensively assess the degree of collab-oration with individual horizontal collaboration initiatives with a view to check feasibility for satisfyingthe customer requirements. The analytic hierarchy process–fuzzy logic model (AHP–FLM) approach ischosen for developing the model, a method that is often used to tackle complex strategic decision makingthat calls for subjective judgment based on well-established logical reasoning, rather than on simple feel-ing and intuition. In the process, the complex and unstructured problem for ‘compatibility test’ is brokendown into elements, and then a customized hierarchy structure is set up to demonstrate the relationshipbetween different hierarchy levels and among these elements. Each element may have a different level ofimportance for the horizontal collaboration. A fuzzy rule based collaboration intensity index (CII) isdeveloped to build up the relationships among these evaluation attributes. Synthesizing the generic rel-ative importance and the forecasted degree of collaboration, the proposed approach can determine thesuccess of the collaboration initiative. An illustrative example of a semiconductor industry supply chain(SSC) member that intends to partner with a potential and competing candidate enterprise is developedto demonstrate the applicability of the proposed fuzzy strategic alliance selection framework and to mea-sure the effectiveness of a horizontal collaboration initiative.

� 2009 Elsevier Ltd. All rights reserved.

1. Introduction

Collaborations in supply network mean long-term relationshipsamong members through reductions in transaction costs, and in-crease in resource sharing, learning, and sharing of knowledge(Cousins, 2002). Horizontal collaboration occurs between partnersat the same level in the manufacturing process, where the benefitsof collaborative manufacturing/purchasing include lower pricesdue to aggregated manufacturing/purchasing quantities, reducedsupply risk, reduced administration cost due to centralized pur-chasing activities, and networking benefits as group memberscommunicate and interact with each other (Tella & Virolainen,2005). However, such an initiative may be subjected to a numberof disadvantages, such as (a) loss of flexibility, as products pur-chased must have a high similarity among group members, (b) lossof control by individual SC members, (c) high coordination costs, asgroup members are competitors, (d) anti-trust problems, and (e)potential consolidation of the supply market in the long run. Fur-ther, the success of such an initiative depends on quality of leader-ship or coordination in the group, as the ability to negotiatecontracts and coordinate members’ interests is critical.

ll rights reserved.

. Bahinipati).

We define horizontal collaboration as a business agreement be-tween two or more companies at the same level in the supplychain (SC) or network in order to allow greater ease of work andcooperation towards achieving a common objective. This may beachieved through proper manipulation, utilization and sharing ofappropriate resources, such as machinery, technology and man-power. Such initiative is highly desirable for the semiconductorindustries to ensure global business opportunity (Cruijssen, Cools,& Dullaert, 2007). In their drive to be more efficient and competi-tive, the semiconductor industries have focused on the internalorganization and processes. Due to more competitive pressure,these companies are now looking externally beyond the bound-aries of their own organizations and value chains for horizontaland vertical collaboration with supply chain partners to achievethe benefits in long run.

Horizontal collaboration completely exploits the conceptualiza-tion of SCs as supply networks (Mason, Lalwani, & Boughton,2007). One of the key ways to value creations is the effectivedeployment and sharing of resources. As the semiconductor indus-tries are involved in these concepts of multi-relations in a network,these are often managed not as a cohesive whole, but as self evolv-ing systems. This introduces the ideas behind reducing the numberof contact points so that they can be better managed, and the seg-mentation of relationships to determine different levels of collabo-rative actions. However, if they can be conceptualized as a

Page 2: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 881

complete network and a single point of control, there is consider-able scope for holistic gains in the collaborative manufacturing/purchasing operations.

The emphasis on horizontal collaboration becomes more prom-inent, as technology infrastructure, collaborative potential, sup-plier sourcing, and information sharing potential are essential forpartnership evaluation criteria. Two mechanisms/initiatives thatare essential in the integration and coordination of the networkare collaboration and information technology (Lee & Whang,2000). Collaboration is necessary to share information and hori-zontally integrate the operations of the network. This is a challeng-ing management task as collaboration is largely a social processwhile information sharing is largely a technological process (Shore& Venkatachalam, 2003). So, the horizontal integration spans overtwo very different management domains.

The semiconductor industry sector is characterized by a num-ber of key and unique characteristics from the perspective of prod-uct features and the sector’s structure, where collaborationpractices developed in response to the economic pressures aredriving the evolution of the chain and encourage greater horizontaland vertical coordination (Kapur, Peters, & Berman, 2003). Withthe objective of supplying a specific product or component, orlocating new enterprises, all enterprises in the semiconductormanufacturing cooperate as synergetic unit to pursue for success(Zhang, Xu, & Wang, 2004). To maintain and improve the compet-itive power of these industries, it is a critical step to select agile,competent and compatible partners quickly and rationally duringits formation phase. Hence, an adaptable and reasonable mathe-matical model is necessary for the core enterprise to select the bestpartner(s) (Saen, 2007). Based on the model for ‘compatibility test’,the evaluation system can be developed. After the assessment of allthe critical factors by the experts, the system will provide theinformation regarding the degree of collaboration feasible withthe partnering organization(s) to facilitate decision making. Theoutcome of this work can help semiconductor industry practitio-ners to explore the degree of collaboration associated with jointventures, and to identify strategies to manage collaboration inthe operation of joint ventures. Although the SC members are eagerto identify the compatibility of the partners for horizontal collabo-ration, finding a quantitative approach to evaluate the comprehen-sive level of the degree of collaboration (defined as collaborationintensity Index, CII in this work) to support rational and objectivedecision making seems to be more significant at the formativestage of the joint venture.

2. Horizontal collaboration

It has been suggested in literature that supply chain collabora-tion is technology dependent (McCarthy & Golocic, 2002), and hasproved difficult to implement (Sabath & Fontanella, 2002). Typi-cally, the supply network fails to differentiate between whom tocollaborate with, i.e., a segmentation of potential buyers or suppli-ers (Sabath & Fontanella, 2002), and this occurs, fundamentally,due to a lack of trust between the business partners (Ireland &Bruce, 2000). However, the semiconductor industries have recog-nized the need for collaboration, emphasizing the establishmentof closer and long-term working relationships or partnerships withcompeting suppliers/buyers at various levels in the supply chain,so as to develop more efficient and responsive supply chains.

The technology collaborations through horizontal relationshipsare continuing agreements, where partners extend their expertisethrough sharing of skills and human resources. The objectives ofsuch collaborations may include increased knowledge of techno-logical threats and opportunities, and improved capabilities inproduct development and efficiency in manufacturing (Cruijssen

et al., 2007). The aims of technological collaboration includeimprovement in the innovation process and the various technolog-ical objectives of corporate business and manufacturing strategy,and public policy. They encompass the following:

1. Improving the development process, such as sharing the devel-opment of new knowledge, products, and processes.

2. Enhancing efficiency in manufacturing chain and manifestingitself in subcontracting relationships between large and smallfirms.

3. Merging previously discrete technologies and disciplines, suchas between mechanical and electronics engineering in the crea-tion of mechatronics.

4. Learning through information exchange about the potentials ofhorizontal collaboration and of particular partners, which maybe facilitated by the use of networks and databases designedto improve awareness of and access to technologicalknowledge.

5. Corporate strategies, such as to improve innovation, and may beconcerned with reducing the cost, risk and uncertainty of tech-nological innovation, merging the resource advantage of onemember with the behavioural advantage of the other, futurescope for mergers or acquisitions, globalization of activities,and improving manufacturing capability.

6. Public policies aimed at improving the comparative technolog-ical performance, and to enhance information flows betweenfirms.

2.1. Types of horizontal collaboration

Typically, the semiconductor industry supply chain encouragesthree forms of horizontal collaboration:

� Infrastructural collaboration may provide collective industrialresearch, with considerably more cooperation, where knowl-edge is jointly created and shared. For example, Microelectronicsand Computer Technology Corporation (MCC) in USA is a collab-orative research group of companies sponsoring research ofcommon interest.

� Contractual form of collaboration may take the shape of a jointventure, formed by two or more SC partners with shared equityinvestments. It could be a partnership linking firms on the basisof continued commitment to shared objectives without equitysharing, commonly known as strategic alliance. For example,IBM and Siemens are collaborating to develop next generationchips.

� R&D collaborations are the agreements that occur between firmssharing R&D efforts and firms may agree to exchange technolo-gies. R&D contracts can be considered as a form of horizontalcollaboration, if contract is flexible and alters with what isdiscovered.

The full range and scope of horizontal collaboration is too broadto allow full consideration in this paper. Instead a more restrictivedefinition is used which includes any activity where two or moreSC partners contribute differential resources and know-how toagreed complementary objectives. This may include: (1) Collabora-tive research programmes and consortia, (2) Joint ventures andstrategic alliances, and (3) Shared R&D and manufacturingcontracts.

2.2. Focus of horizontal collaboration

Collaboration is a feature of the high-technology industry. Stud-ies of horizontal collaboration across semiconductor industry show

Page 3: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

882 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

that a large numbers of enterprises are devoted to technological is-sues (Barratt, 2004), and a few of them involve in joint R&D effort.A large number of agreements are in the area of information andcommunication technology. Studies of various semiconductor elec-tronics industries and technologies, such as information technol-ogy, telecommunications, integrated circuits, computer systems,and semiconductors show a high level of collaborative initiativesamong enterprises (Lin & Chen, 2004).

If the focus of horizontal collaboration is pre-competitive, or isconcerned with building capabilities for the development of newproducts, then this form of strategic collaboration in core areas ap-pears to make more sense (Kapur et al., 2003). Further, the aims ofjoint development of individual products in which a market alreadyexists are described as ‘tactical’. But, if the industry, through one ormore tactical collaborations, manages to develop its capability forfuture with new product development, then these can be describedas ‘strategic’.

The focuses of horizontal collaboration changes over time, i.e.,these vary along with the industrial and technological develop-ment. The reasons can be suggested for technological collaborationfrom within an innovation, corporate, public policy and globaliza-tion perspective. New technologies are extremely expensive todevelop. A new semiconductor wafer fabrication plant can cost$500–$700 million. Collaboration can help share these high costs,although returns from them will also be shared (Caraynnis & Alex-ander, 2004). Cooperation can reduce the unnecessary duplicationof R&D efforts. Many new technologies may involve the diffusion ofpreviously discrete areas of knowledge, such as mechatronics anddevelopment of high temperature super conductors. So, horizontalcollaboration may be chosen by the large firm as a means of access-ing depth of knowledge. This will enable the scope of potentialproducts, both in the sense of improved technical capabilitiesand market applications. Collaboration provides an effective mech-anism for the joint creation and promotion of new technical stan-dards. In summary, horizontal collaboration in semiconductorindustry SC is a strategic tool, which may be used to minimizethe effect of competition (Lau, 2002), either by raising the scaleof resources developed to a project to deter other firms fromattempting to compete, or by trying in a partner with specific skillsso that competitors cannot gain access to them.

3. Motivation and objectives

Joining a global SC for semiconductor industries is a criticalstrategy to stay competitive for today’s business. While choosingthe best partner, the purchasing manager might be uncertainwhether the selection will satisfy completely the demands of theirorganizations (Bevilacqua & Petroni, 2002). The overall objective ofthe potential partner evaluation process is to reduce risk and max-imize overall value to the organization (Wang and Che, 2007).Thus, a methodology that can capture both the subjective andobjective evaluation measures is needed. The analytic hierarchyprocess (AHP) approach was suggested in the literature for supplierselection problems (Benyoucef & Canbolat, 2007). The applicationof AHP in more than 25 diverse areas to rank, select, evaluate,and benchmark decision alternatives are demonstrated by Vaidya(2006).

The strength of the AHP lies in its ability to structure a complex,and multi-attribute problem hierarchically, and then to investigateeach level of hierarchy separately, combining the results (Liu & Hai,2005). In the traditional formulation of the AHP, the human judg-ments are represented with crisp numbers. However, in manypractical cases, the human preference model is uncertain and deci-sion makers might be reluctant or unable to assign exact numericalvalues to the comparison judgment. For example, when evaluating

different partners, the decision makers are usually unsure abouttheir level of preference due to incomplete and uncertain informa-tion about potential partners and their performances. Since someof the evaluation criteria are subjective and qualitative, it is verydifficult for the decision maker to express the strength of his pref-erences and to provide exact pair-wise comparison judgments(Tolga, Demircan, & Kahraman, 2005). For this reason, a methodol-ogy based on AHP and fuzzy logic can help us to reach an effectivedecision. By this, we can deal with uncertainty and vagueness inthe decision process.

There is a very little research done within the framework of hor-izontal collaboration in SCM literature, i.e., to estimate a collabora-tion intensity index in the semiconductor industry SC based on thecollaboration attributes. These attributes are selected from differ-ent domains, such as industry characteristics, competitive advan-tages, internal parameters and external parameters, in relation tocollaborative procurement. It may be noted that the importance gi-ven to certain attributes may vary depending on the objective andthe process of the supply chain. There may be situations whendecisions regarding the implementation of horizontal collaborationare to be taken independently. The attributes identified can beevaluated independently to test the compatibility of the SC mem-bers for collaboration. As these attributes are considered to beindependent entities, the relative importance of these attributescan be determined by using analytic hierarchy process (AHP)(Saaty, 1990). The collaboration intensity index can be consideredanalogous to the alternatives defined in the AHP framework.

The methods mentioned in the contributions to AHP are unableto adequately represent the uncertainty of human judgment. Inaddition, some are too complicated to be operated by program-ming. Further, dealing with the inconsistency of judgment matrixis another problem. This method also ignores the effects resultedfrom interdependent attributes. Coping with the drawbacks ofthe previous researches, this paper aims to meet the followingessential practical requirements for the SC alliance compatibilitytest: (1) The decision attributes may include qualitative attributes,(2) The interdependency among attributes should be considered,(3) The resources available to obtain the information on the evalu-ation attributes are limited, and (4) The effectiveness of the collab-orative initiative should be measured.

Multi-criteria decision analysis method (Belton & Stewart,2002; Hwang, 2004) is applied to the scores and weights of mul-ti-level structures of the decision system (Saaty, 1990). Most ofthe systems do not seem to be appropriate for modeling the deci-sion problems based on information system. One major issue isthat the multiple objective programming applied to partnershipselections has limitations due to the inclusion of qualitative criteriathat are in fact highly important in this context. The theory of fuzzysets has extended traditional mathematical decision theories sothat it can cope well with any vagueness problem, which cannotadequately be treated by probability distributions (Murphy,1995). The impact and the relationships among the characteristicsin multi-criteria decision problems can be described only by vagueverbal descriptions. The reasons for increasing popularity of theapplication of the fuzzy system in partnership evaluation are (Chan& Kumar, 2007): (1) Fuzziness must be introduced so as to obtain areasonable model to solve this complex problem and (2) There is aneed to formulate human knowledge in a systematic manner andput it into mathematical model.

Fuzzy sets and fuzzy logic are powerful mathematical tools formodeling uncertain systems in industry and facilitates common-sense reasoning in decision making in the absence of completeand precise information. Their role is significant when applied tocomplex phenomena not easily described by traditional mathe-matical methods, especially when the goal is to find a good approx-imate solution. Many real-world applications cannot be described

Page 4: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 883

and handled by classical set theory (Chen & Pham, 2001). A fuzzyset is an extension of the crisp set. Crisp sets only allow full mem-bership or non-membership at all, whereas fuzzy sets allow partialmembership. In other words, an element may partially belong to afuzzy set. These elements may use values ranging from 0 to 1 forshowing the membership of the objects in a fuzzy set. Completenon-membership is represented by 0, and complete membershipas 1. Values between 0 and 1 represent intermediate degrees ofmembership. So, a fuzzy number is characterized by a given inter-val of real numbers, each with a grade of membership between 0and 1 (Nguyen and Walker (2000)).

Uncertainty of the information in the current problem alongwith inherent difficulties related to human knowledge make thedecision making highly complicated (Fu et al., 2006). The purposeof crisp measure is to capture the expert’s knowledge which is notpossible in traditional AHP to reflect the human thinking style(Kahraman, Cebeci, & Ululan, 2003). So, AHP methodology inte-grated with fuzzy logic can be adopted as an alternative to the con-ventional methods of weight estimation in AHP.

A semiconductor manufacturing industry can establish a newparadigm of horizontal collaborative partnership with competingfellow-buyers or suppliers by carefully applying the concepts ofmodularization and standardization to its products (Link & Marxt,2004), organization and supply chain. In this context, we define‘compatibility’ as the suitability of potential competing SC mem-ber(s) to engage in a collaborative relationship with another mem-ber. The partnership assessment process is important and routine,especially for these similar companies that manufacture productswith short life cycles (Chang, Wang, & Wang, 2006). Our researchfocuses on this critical aspect of horizontal collaboration betweencompeting semiconductor manufacturing industries (suppliers),i.e., developing a multi-attribute framework for establishing ‘com-patibility’ between partners willing to initiate resources or processtools sharing for horizontal collaboration. In this paper, we consid-ered a decision method (Prodanovic & Simonovic, 2003) defined asanalytic hierarchy process–fuzzy logic model (AHP–FLM), whichdetermines the weighting of subjective judgment for the scientificevaluation framework of the horizontal collaboration problem(HCP). To target the specific semiconductor industry of interest, acustomized configuration hierarchy (CCH) is identified with basicbelief acceptability values for all attributes assigned by industryexperts.

As different projects tend to expose the member of semiconduc-tor industry supply chain (SSC) to different degree of collaboration,and different collaborative initiatives have different levels of im-pact on the success of shared collaborations, the proposed AHP–FLM approach is one of the most optimal techniques to evaluatethe degree of collaboration and to provide comprehensive informa-tion to SC partners to make rational decisions. The proposed modelconsiders the parameters with different degrees of importance. Therelative importance is very much organization-specific, which de-pends on the abilities of the organizations to implement horizontalcollaboration. The effective implementation of collaboration de-pends on the impact of these attributes for the purpose. Some ofthese attributes are supported by subjective judgments to ensurecompatibility between industries for collaboration. One approachto overcome this difficulty is to assign uniform linguistic termsin a fuzzy environment and obtain judgments from semiconductorindustry experts. The judgment regarding these attributes may bevague and difficult to quantify. Fuzzy logic theory allows the natu-ral description, in linguistic terms, of the compatibility aspectrather than relying on precise numerical threshold values. Fuzzysystem models are flexible enough to accommodate these impre-cise and vague data (Ross, 1997). This advantage, dealing withthe complicated systems in a simple way, is the main reasonwhy fuzzy logic theory is applied in the present context. Its mul-

ti-attribute decision-making property applied for a range of lin-guistic terms has motivated to propose this fuzzy model for‘compatibility test’ (Shore & Venkatachalam, 2003). We also pres-ent the procedure of constructing a fuzzy inference system usingfuzzy logic to score the ‘compatibility test’.

4. The research methodology

The method consists of developing an analytic hierarchy pro-cess–fuzzy logic model (AHP–FLM) to conduct a ‘compatibility test’for horizontal collaboration in the semiconductor industry supplychain.

4.1. Identification and selection of focus group

The semiconductor industry was chosen for several reasons: (1)This industry is of economic significance to many countries and isone of the largest manufacturing activities in the world. (2) Thisindustry has experienced the diffusion of innovation like high per-formance workplace practices along the supply chain, and offerspotential to simultaneously examine different dimensions of col-laboration. (3) These are highly capital-intensive enterprises anddeal with short life cycle products.

As the literature on horizontal collaboration is very limited togive a sound conceptual foundation, focus group research wasused. This study selected focus groups rather than in-depth inter-view for two reasons: (1) The specific experiences of participantsin focus groups can be more utilized, i.e., the participants can buildon others’ opinions and come up with new ideas. (2) The resultsgenerated by focus groups could be more reliable and valid. Thisstudy started with selection of participants, as the objective wastoo specific and the participants in the focus groups have some-thing in common. As such, potential participants were chosen fromenterprises in electronics industry, which comprises semiconduc-tor manufacturing, consumer electronics, personal computer, andtelecommunication and information technology. The respondentspossessed an average experience of nearly 10 years in purchas-ing/supply management functions and were employed from eightdifferent enterprises operating in semiconductor, electronics andPC assembly. The respondents held a variety of positions, with pur-chasing manager being the most common.

The respondents have dealt with new product development insemiconductor and consumer electronics industry for at least 5years. The respondents are holding management positions andtheir knowledge are assumed to be sufficient to share with eachother. Data collection at the management level has been frequentlyused in dealing with short life cycle products and new productdevelopment studies. Further, the respondents have experiencedin developing and introducing re-innovative products in the past5 years. The respondents were also asked about their work experi-ences immediately prior to the position they are currently holding.The majority of the respondents had been in a supply position withtheir previous employer. Accordingly, the total number of respon-dents was 20. All respondents were assured anonymity, and toavoid some participants dominating the process, the respondentswere not encouraged to interact with each other. This has beenachieved by keeping the interaction with the individual respon-dents from the authors’ side only.

The assessment portfolio for each attribute was developed asfollows: The practice of ‘compatibility test’ for horizontal collabo-ration refers to the extent to which the supply chain membersshared their private information and resources for supply chainoperations over time. The four major attributes and 10 sub-attri-butes were identified through the review of previous studies andby panel of experts. The contribution of each attribute was as-

Page 5: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

Table 1Collaboration attributes for ‘compatibility test’ for horizontal collaboration.

A: Industry characteristics (IC) C: Internal parameters (IP)A1: Industry structure (IS) C1: Strategic parameters (SP)

A11: Decision-making ability C11: Outsourcing strategiesA12: Level, scope and time horizon. C12: Attitude of top managementA13: Previous partnership history C2: Tactical parameters (TP)

A2: Financial stability (FS) C21: CommunicationA21: Business performance C22: ICT integrationA22: Capital required/available C23: Speed of decision making

A3: Global reputation (GR) C24: Collaborative planningA31: Eco friendliness C3: Operational parameters (OP)A32: Brand image C31: Productivity

C32: FlexibilityC33: ControlC34: Lead timeC35: ReliabilityC36: Capacity utilizationC37: Product sizeC38: Inventory turnover

B: Competitive advantage (CA) D: External parameters (EP)B1: Product orientation (PO) D1: Product characteristics (PC)

B11: Quality of products and services D11: Demand variabilityB12: Product life cycles D12: Price elasticity

B2: General competitive edge (GCE) D13: Competitive pressureB21: Market share D2: Industry orientation (IO)B22: Customers orientation D21: Financial performanceB23: Technology standard/level D22: Profit potential

D23: Resource utilization

884 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

sessed on a five-point Likert scale. All responses were ranked from1 to 5, 1 representing minimum influence and 5 representing max-imum influence for compatibility towards horizontal collaboration.The ‘compatibility test’ was operationalized as the degree to whichthe SC members are involved in joint problem solving and resourcesharing at the planning and operational levels.

4.2. Selection of collaboration attributes

The first step of AHP–FLM approach was to identify the evalua-tion attributes that need to be considered in the strategic SC part-nership compatibility, based on brainstorming, expert opinion, andliterature survey. Based on these attributes, a set of customizedevaluation attributes related to the semiconductor industry wereidentified by experts or decision makers. Since variations existfor assuring the exactness of the information of the decision attri-butes, there may be a basic acceptability index assigned to everyevaluation attribute by the experts/decision makers. Then the opti-mal evaluation attributes model can be formulated based on thebasic acceptability. This issue has not been considered in this paperto simplify the research. However, the evaluation attributes wereselected by the semiconductor industry experts from the perspec-tive of their own industrial domain based on consensus.

At the formation stage, most collaboration attributes associatedwith the joint venture are not clear to the SC partners. These attri-butes, as described by the literature, are multifaceted and stronglyproject/component/product-related. Judgment of these attributesis normally vague and imprecise during formation stage. The SCmembers need to make a comprehensive assessment with respectto collaboration condition pertaining to the proposed joint venture.The prudent selection of evaluation attributes plays a critical rolein establishing the degree of collaboration between the SC mem-bers. Further, as the number of evaluation attributes increases,the chance of having interdependency also tends to increase. Theevaluation attributes in SSC problem inherit two important practi-cal characteristics: uncertainty and ignorance. Uncertainty existswhen the available information for decision making is incomplete,imprecise or unreliable, whereas ignorance results from lack ofinformation while making decisions (Beynon, Curry, & Morgan,2000). For example, if the decision maker is not completely certainto obtain exact information about the supply network performanceof the candidate SC, a value is assigned by the decision maker torepresent his confidence in the information or value of this evalu-ation attribute based on his past experience.

The brainstorming session with the semiconductor industry ex-perts and academicians, and review of literature and case studiesin semiconductor industry supply chain covered various aspectsof collaborative relationships pertaining to (1) finance, (2) humanresource management, (3) industrial characteristics, (4) knowl-edge/technology acquiring and management, (5) marketing, (6)organizational competitiveness, (7) product development, produc-tion and logistic management, and (8) relationship building andcoordination (Hajidimitriou & Georgiou, 2002; Harvey & Lusch,1995; lp, Huang, Yung, & Wang, 2003; Mikhailov, 2002). Evaluationattributes also include qualitative indices global reputation andcommunication openness (lp et al., 2003). The evaluation attri-butes also include subjective indices, such as business and culturalcoherence and shareholder’s favourability, and objective indices,such as corporate image and geographic coverage (Mikhailov,2002).

The sample of 20 respondents drawn from the semiconductorindustry business databases were directly involved in collaborativeinitiatives in their enterprises. A semistructured interview wasconducted for the ‘compatibility test’ for horizontal collaborationand to obtain adequate data for generalization of the findings.The conceptualization phase identified the key attributes for col-

laboration practices among similar services/goods providers. Todevelop a scale, recent literature was used to define the domainsof an attribute into a set of sub-attributes. Respondents were askedabout their perceptions in deciding partners for actual collabora-tive practices from the perspective of procurement. They wereasked questions about their relationships with the SC members.From this perspective, it is believed that the research has capturedthe essence of the overall view of each chain member that reflectskey attributes of current collaborative practices.

The attributes identified by the authors with the help of respon-dents from semiconductor manufacturing industries for the ‘com-patibility test’ during horizontal collaboration initiative aredepicted in Table 1. These attributes have been categorized intofour major groups while evaluating the collaborative intensityfunction. The attributes for the ‘compatibility test’ are categorizedas under:

A. Industry characteristics: This attribute was considered criti-cal, as the foremost activity to initiate partnership (in termsof collaboration) is to assess the industry structure, financialstability and global reputation for compatibility. Theresource sharing and revenue sharing ability of the similarservice providers is considered while initiating a collabora-tive venture. It must consider the aspects of industry struc-ture (IS), which is dependent on the decision-making ability,level, scope and time horizon, and the previous partnershiphistory of the SC members. Further, it must also considerthe aspects of financial stability (FS), which is dependenton business performance and capital required or availablewith the SC members, and the aspects of global reputation(GR), which depends on the ecofriendliness and the brandimage of the products of the SC members.

B. Competitive advantage: Another important attribute, whichwas considered critical while pursuing the ‘compatibilitytest’, is the competitive advantage of the competing mem-bers in the supply network. It is necessary to assess theproduct orientation (PO) of the organization in terms ofquality of products and services as well as the life cycle ofthe products. Further, the general competitive advantage

Page 6: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 885

(GCE) in the global market may be evaluated in terms ofmarket share, customer orientation and the existing technol-ogy level or standard.

C. Internal parameters: Another important attribute, the semi-conductor industry experts feel critical, was the internalparameters of the partnering enterprises. It is necessary toassess the strategic parameters (SP), such as out sourcingstrategies, and the attitude of the top management of theorganizations for horizontal collaboration. Tactical parame-ters (TP), such as communication technologies, ICT integra-tion, speed of decision making and collaborative planningis essential for the partnering organizations to foster thestrategic partnership. Nevertheless, the operational parame-ters (OP), such as productivity, flexibility, control measures,lead time, reliability, capacity utilization, product size, andinventory turnover are also true performance indicators tobe assessed during the selection of strategic partner.

D. External parameters: Another important parameter, whichwas considered critical while undertaking the ‘compatibilitytest’, is the external parameters, such as product characteris-tics as well as industry orientation. It is necessary to assessthe product characteristics (PC) in terms of demand variabil-ity, price elasticity and competitive pressure. Further, theindustry orientation (IO) aspect deals with the financial per-formance measure, profit potential as well as resource utili-zation capability.

It is difficult to consider all the attributes and their possibleimpact on the degree of collaboration, as these belong to differentdomain. Some of these attributes are supported by subjectivejudgments to ensure compatibility between industries for hori-zontal collaboration. One approach to overcome this difficulty isto assign a set of uniform linguistic terms in a fuzzy environmentand obtain judgments from semiconductor industry experts. Thisset is unique for all the attributes identified. The judgmentsregarding these attributes may be vague in nature and difficultto quantify. The proposed model considered these attributes withdifferent degrees of importance. The relative importance mayvary from organization to organization depending on their capa-bilities to implement horizontal collaboration. The relative impor-tance can be determined by AHP, but effective implementation ofcollaboration depends on the impact of these attributes for thepurpose.

It should also be noted that the AHP requires that a problem bedecomposed into levels, each of which is comprised of elements.The elements of a given level are mutually independent, but com-parable to the elements of the same level. This assumption wasfundamental to the proposed AHP–FLM approach. Hence, theway to categorize the collaboration attributes for ‘compatibilitytest’ and the relationship/nature of these attributes at the same le-vel will determine the effectiveness and validity of the method.These parameters are strongly correlated. An attribute at a partic-ular level may influence other factors at another level. During cat-egorization, the collaboration attributes may be stronglyconnected. As a result, an attribute at one particular level mayinfluence other attributes at another level. Hence, the indepen-dency among elements cannot be guaranteed. As a reference, whensetting up the hierarchy structure for the proposed fuzzy AHPmethod, the categorization method of various factors ensured themaximum independency among elements at the same level.

4.3. Design of customized configuration hierarchy for compatibilitytest

The interrelationships among various objectives are best ex-pressed by the use of a hierarchical structure. Under this type of

hierarchical structure, only the performance criteria in the bottomlevel are given by the experts; performance of the other criteria notin the bottom level can be obtained by the use of aggregation (Lee& Shib, 2001). To produce a generic hierarchy structure, the collab-oration attributes were sorted into a number of groups at the cri-teria level, with a few sub-parameters under each group, asshown in Fig. 1.

We claim that the customized configuration hierarchy hasadvantages of adaptability and flexibility. For adaptability, theoptimal evaluation attribute model was customized and adaptedto meet the special needs of the semiconductor industry supplychain. For flexibility, this paper relaxed the attribute dependency,which is a common phenomenon in the literature of multi-attri-bute decision-making problems. The major advantages of hierar-chical performance structure are:

1. Expressing the interrelationship among all the criteria in a briefand clear form.

2. Clustering criteria into a category according to their character-istics and to allocate them in a suitable position such that it ispossible to identify and review the criteria performance withinand between each category.

3. Obtaining the performance level by level; the evaluation resultscan give the actual achievement for each criterion, which can beused to improve the system.

4.4. Determination of weight vector by AHP

Analytic hierarchy process has been used for scaling the weightsof the elements in each level of the hierarchy with respect to theattributes of the next higher level. This was done by means ofpair-wise comparisons of the attributes to indicate the strengthwith which one attribute dominates another for the criterion underwhich they were compared. The reason lies in that the perfor-mance of a system is a result of the interaction of various attri-butes, but every attribute plays its own role and makescontribution to the system as a whole. The core of the AHP is theconformation of judgment matrix by semantic scale and theiraccording reciprocals (Maggie & Rao, 2001; Wang, Huang, &Dismukes, 2004). Such judgment often ignores the vagueness ofrespondents’ mind.

In many practical cases the pair-wise judgments of decisionmakers will contain some degree of uncertainty (Saaty, 1990).The decision maker may be certain about the ranking order ofthe comparison results but uncertain about the precise numericalvalues of his judgments. The classical AHP attempts to overcomethis problem by introducing a discrete linguistic set of comparisonjudgment. The pair-wise comparison was established using a nine-point scale that converts the human preferences between variousattributes, as shown in Table 2. The comparison was based on ex-pert judgment. All the 20 respondents were invited; each expert’sopinion was obtained and analyzed individually to determine theweight vector pertaining to one group of factors.

AHP was used to determine the comparative weights of the var-ious criteria. It decomposed a complicated system into a well-orga-nized structure, based on the hierarchy. By the use of pair-wisecomparison between various criteria, a more credible and moreobjective criterion weighting has been obtained. In order to avoidartificial errors and the contradiction of different factors, a consis-tency check was conducted until a satisfactory condition is ob-tained. A consistency check is a unique advantage of AHPcompared to other methods. After all 20 expert’s opinions wereanalyzed and the weight vector based on each expert’s judgmentwas worked out, the weight vector was determined by averagingthe sum of individual weight vectors. It should be noted that the

Page 7: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

Level 4: Decision Parameter

Level 2: Attribute

Level 3: Sub-attributes

EXTERNAL PARAMETERS

(D)

INDUSTRY ORIENTATION

(D2)

Collaboration Intensity Index (CII)

INDUSTRY CHARACTERSTICS

(A)

COMPETITIVE ADVANTAGE

(B)

INTERNAL PARAMETERS

(C)

FINANCIAL STABILITY

(A2)

GLOBAL REPUTATION

(A3)

PRODUCT ORIENTATION

(B1)

GENERAL COMPETITIVE

EDGE (B2)

OPERATIONAL PARAMETERS

(C1)

TACTICAL PARAMETERS

(C2)

STRATEGIC PARAMETERS

(C3)

PRODUCT CHARACTERISTICS

(D1)

EEvvaalluuaattiioonn ooff PPootteennttiiaall PPaarrttnneerrss ffoorrHHoorriizzoonnttaall CCoollllaabboorraattiioonn

INDUSTRY STRUCTURE

(A1)

Level 1: Decision Prob lem

Fig. 1. Customized configuration hierarchy for ‘compatibility test’.

Table 2Linguistic measures of importance (adapted from Saaty (1990)).

Scale Definition Explanation

1 Equal importance Two factors are contributing equally to theobjective

3 Weak importance There is evidence favouring one factor overanother but it is not conclusive

5 Strong importance Good evidence and logical criteria exist toshow one is more important

7 Very strong importance Conclusive evidence as to the importance ofone factor over another

9 Absolute importance Evidence in favour of one factor over anotheris of the highest possible

2, 4, 6, 8 Intermediate values When compromise is needed

886 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

invited experts possessed different levels of knowledge of the fac-tors associated with the ‘compatibility test’ for horizontal collabo-ration. Hence, their opinions were considered to have differentlevels of impact on the pair-wise comparison. A weight coefficientreflecting the expert difference can be incorporated while averag-ing the weights assigned by individual experts for various factors.However, this issue was not considered in this paper to simplifythe research.

4.5. Defuzzification

Defuzzification is the process by which a solution set is con-verted into a single crisp value. The fuzzy logic solution set is inthe form of a function, relating the value of the result to the degreeof membership. The entire range of possible solutions may be con-tained in the fuzzy solution set. So, defuzzification is a process toextract and easily comprehendible answer from the set. Since fuz-zy sets or fuzzy numbers represent many possible real numbers

(with different membership degrees), they do not always yield atotally ordered set. This makes comparison a non-trivial affair. Var-ious approaches have been proposed in the literature. Because ofits simplicity, intuitive appeal, and effectiveness, Chen andHwang’s crisp score method (Chen & Hwang, 1992) was adoptedto defuzzify the fuzzy sets.

Fuzzy set theory has provided a powerful approach to representand manipulate such non-quantitative descriptions through differ-ent operations of fuzzy sets and fuzzy numbers. Linguistic expres-sion were captured by appropriate fuzzy numbers and finallydefined by crisp (real) numbers. In fuzzy sets, the linguistic expres-sions were transformed to triangular fuzzy numbers (TFN). TFNconsists of three parameters, wherein each parameter representsa quantity of a linguistic value associated with a degree of mem-bership in the interval [0,1]. Further, these three parameters de-note the smallest possible quantity, the most promising quantity,and the largest possible quantity that describe the linguistic value.One of the advantages for applying fuzzy rule based approach toconnect evaluation attributes was to relax the following assump-tions in the weight-based approach (Mikhailov, 2002): (1) the eval-uation attributes were independent and (2) the trade-off ratebetween two attributes was a constant. So, the fuzzy rule basedcollaboration intensity function developed in this paper has im-proved the handling of the interdependency among evaluationattributes.

The degree or intensity of the collaboration is difficult to mea-sure due to the vagueness of the information regarding the param-eters identified for the ‘compatibility test’. Fuzzy based approachesare well suited in the cases where the information is not very pre-cise to establish the relationship between collaboration attributesand the degree of collaboration. In order to deal with the fuzzinessassociated with the use of linguistic terms for assessment and con-vert these fuzzy assessments by the managers into a meaningful

Page 8: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

0.2 0.4 0.6 0.8 1.00 Assessment Score

Degree of Membership

0

1

EL VL L ML AV MH H VH EH

Fig. 2. Triangular membership for the parameters of the ‘compatibility test’.

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 887

portfolio, fuzzy sets were used. The fuzzy data were then trans-formed into crisp values using fuzzy logic methodology. It is acumbersome task to assign linguistic terms to all the attributesof the ‘compatibility test’, because these attributes further dependon various sub-attributes; for example: the global reputation attri-bute depends on the eco friendliness and brand images of the col-laborating SC members. The values for these sub-parametersdepend on the business of the industry, types of products they pro-duce, availability of resources, investments in ICT, life cycle of theproducts, willingness of SC members for collaboration and adapt-ability of new technologies in quick time.

The evaluation of collaboration attributes of the ‘compatibilitytest’ for collaboration was done by defining a set of linguisticterms, and this set was unique for all the attributes. The uniformset of linguistic terms for all these attributes provided a standardscale to each parameter, which otherwise was difficult as theseattributes belong to different domains. The unique set establishedthe crisp number as was obtained from the same set of linguisticterms for all these parameters. The linguistic terms used were: ex-tremely low (EL), very low (VL), low (L), marginally low (ML), aver-age (AV), marginally high (MH), high (H), very high (VH), andextremely high (EH). The 10 attributes as identified earlier in fourbroad categories (Table 1) were perceived differently for the eval-uation of compatibility for the competing SC members. The linguis-tic terms were based on the performance improvement achievedby considering these attributes. So, a study of the semiconductorindustry helped in assigning linguistic terms based on the applica-tion as well as for testing compatibility by assessing these attri-butes. A thorough questionnaire covering the detailed aspects ofthese attributes helped in incorporating the degree of agreementof semiconductor industry experts.

The linguistic values were created to model the crisp thresholdranges. The linguistic assessment and judgments were vague and itwas not reasonable to represent it in terms of precise numbers. Therespondents were more confident in using interval judgments thanfixed value judgments. So, triangular fuzzy numbers were used todecide the priority of one decision variable over other in AHP com-bined with fuzzy logic (Chan & Kumar, 2007). It was convenient towork with triangular fuzzy number because of their computationalsimplicity, and they were useful in promoting representation andinformation processing in a fuzzy environment. Based on a reviewof the data and discussions with the experts, for the fuzzified scor-ing system of ‘compatibility test’, we have determined the mem-bership functions with characteristics of smoothness, symmetry,and zero on both extremes with a quick rise in the middle. A trian-gular membership function satisfied our requirements.

In crisp sets, which are collection of objects with the same prop-erties, the objects either belong to the set or not. In practice, thecharacteristics value for an object belonging to the set is codedas 1 and if it is outside the set then the coding is 0. The key ideain fuzzy logic is the allowance of partial belongings of any objectto different subsets of the universal set instead of belonging to asingle set completely. It is obvious that there are interferences be-tween the numbers because of fuzzy linguistic word approxima-tions. Likewise in fuzzy logic, values of variables are expressedby linguistic terms, the relationship is defined in terms of IF-THENrules, and the outputs are also fuzzy subsets which can be made‘‘crisp” using defuzzification techniques. First, the crisp values ofsystem variables were fuzzified to express them in linguistic terms.Fuzzification is a method for determining the degree of member-ship that a value has to a particular fuzzy set. This was determinedby evaluating the membership function of the fuzzy set for the va-lue (Lee & Shib, 2001). The fuzzy scores provided more sensitiveinformation regarding the status of physical assessment of collab-oration attributes. This soft scoring might act as a mechanism tostart a treatment intervention that will slow the loss of functional-

ity. A fuzzily formatted attribute value was represented through alinguistic term and an associated membership function (essentiallyby defining a fuzzy set).

4.6. Evaluation of collaboration intensity index

The attributes affect the extent of collaboration in semiconduc-tor industry between competing SC members. But these attributesshould not carry equal weightage. The extent of implementation ofthese attributes might be different for different pair of competingsuppliers/members. The relative importance of these parameterswas determined by assigning weights with the help of AHP (Liu& Hai, 2005). Even though the use of AHP has been used in varioustypes of decision-making processes as depicted in literature, itseemed to be first attempt to apply AHP with fuzzy logic in hori-zontal collaboration problem.

The weights were represented as WIC, WCA, WIP, and WEP forattributes, such as industry characteristics, competitive advantage,internal parameters, and external parameters, respectively. A pair-wise comparison was conducted to assign relative weights to theseparameters. These weights determined a score for the extent ofcollaboration between the competing SC members. These attri-butes were assessed from various perspectives, as has been givenin Table 1. Assessment of all these attributes was conducted byusing linguistic terms. Each linguistic term has a range of assess-ment score, which was described by a triangular membership func-tion (Fig. 2) rather than a single estimate. The managers of thesemiconductor and consumer electronics industries were comfort-able in describing these attributes in terms of a range rather than asingle estimate (value). For example, the term ‘high’ (the range is‘marginally high’ to ‘extremely high’), was be defined by a triangu-lar membership function with an assessment score of 0.7, assigneda membership grade of 1, and assessment score 0.6 and 0.8, as-signed a membership grade 0. Scores in the range 0.6–0.7 and0.7–0.8 have partial membership grades in the range of 0–1.

The assessment using a linguistic term or a range of terms wasconverted into crisp form, i.e., the process of defuzzification is per-formed. A crisp number C was estimated by using a trapezoidalmembership fuzzy Number F = (a,b,c,d) as depicted in Fig. 3 (Chen& Hwang, 1992):

C ¼ 12

d1� c þ d

þ bb� aþ 1

� �ð1Þ

It was observed from Fig. 4 that [b,c] interval represents a member-ship degree of 1 (one) for the fuzzy number (F). The values less thana and beyond d represent membership degree of 0 (zero) for F. Themembership degrees vary linearly in the range [a,b] and [c,d].When a range of terms was used, a trapezoidal equivalent of the tri-angular membership functions representing the range of linguisticterms was derived before the use of the conversion formula.

Page 9: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

a

H M

b c d

VH

Deg

ree

of

Mem

bers

hip

Fig. 4. Conversion of linguistic term to a crisp number.

a b c d

Deg

ree

of

Mem

bers

hip

Fig. 3. Trapezoidal membership function.

888 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

The crisp score was computed for all the attributes, i.e., CIC, CCA,CIP, and CEP. The overall score of the degree of collaboration wascomputed as follows:

CII ¼W IC:CICðCIS þ CFS þ CGRÞ þWCA:CCAðCPO þ CGCEÞþW IP:CIPðCSP þ CTP þ COPÞ þWEP:CEPðCPC þ CIOÞ ð2Þ

where CII was the final score for the degree of collaboration be-tween 0 and 1 for the competing enterprises. The consideration ofrelative weights (Wx) and fuzzy logic for crisp score (Cx) for all theattributes and sub-attributes (x) resulted in a framework for ‘com-patibility test’. The framework for proposed methodology is de-picted in Fig. 5.

The collaboration intensity index (CII) was split into linguisticterms, such as ‘‘low”, ‘‘medium” and ‘‘high” to represent ‘‘no collab-oration” (NC), ‘‘partial collaboration” (PC) and ‘‘complete collabora-tion” (CC) depending on the availability of data on the abovementioned linguistic terms and the relative weights. A suitablefuzzy rule was used to categorize the extent of collaboration underthese three linguistic terms. This will help in evaluating the extentof horizontal collaboration in the semiconductor industries. Fur-ther, the comparison between various semiconductor industries(similar service providers) can be done for selecting the most idealstrategic partner for horizontal collaboration, which has the largestscore. The proposed model has focused on the four critical attri-butes identified for compatibility among SC members during hori-zontal collaboration. In general, the model can be used to anynumber of attributes for the ‘compatibility test’.

5. Application of the proposed methodology

The complex SC activities and the diverse competing organiza-tions with varying capabilities of technology, knowledge, informa-tion and resources lead to difficulty in accomplishing the objective.

Further, the degree of implementation of the collaboration activi-ties may be different for competing organizations. So, a case studywas considered as one of the valid approach to understand theimplications of the fuzzy logic based approach for selecting thestrategic partner for horizontal collaboration. This may form theo-retical generalizations with the help of structured interview andanalytical procedure to determine the extent of collaborationsamong the competing members of the supply network in the semi-conductor industries. However, in the present context, the semi-structured interview with the respondents in the focus groupcomprised of various aspects of the four critical attributesidentified.

The objective of this section is to demonstrate, step by step, theapplicability of the proposed framework for evaluating the poten-tial candidates of SC partnership for viability of shared collabora-tive relationships. This study was undertaken on a globalcomponent supplier of a personal computer (PC) manufacturer.The main reason for selecting this component supplier were: (1)The supplier deals with high-value component (A-class) natureand also most expensive supply item within this category and (2)It has the longest lead time in all supply items for PC manufactur-ing. The company owns several factories globally with the coreindustrial capability in personal computer (PC) manufacturingand assembly. The company plans to join one strategic SC alliance(SC#1) in order to expand the company’s global market. The char-acteristics of the SC#1 are as follows:

The SC#1 contains a number of strongly affiliated companies.The manufacturing companies are located globally. Theseindustries are responsible for producing PC components includ-ing computer mother boards, memories, power suppliers andmonitors. SC#1 purchases key computer components includingCPUs, chipsets, and hard disks from factories located in Asia-pacific. To enhance competitiveness and customer responsive-ness, the assembly factories are located near the consumermarkets.

Based on the experience and preliminary research with respectto the factors of the ‘compatibility test’ for horizontal collaboration,the current study has focused on 10 attributes as identified earlier,and categorized them into four groups: industry structure, com-petitive advantage, internal parameters, and external parameters.These test attributes groups are in parallel rather than one govern-ing another. The sub-attributes under these categories are lesslikely to be influenced by other factors at the same level. Hence,this kind of classification method can minimize the correlationsbetween the test attributes at the same level, which satisfies thefundamental prerequisite of the proposed AHP-fuzzy logic ap-proach. The managers responded with their agreements regardingthe importance of the various aspects of the collaborative venturesamong competing members of the supply network. To overcomethe diversity of the views of the different members of semiconduc-tor industry supply network, the average of the weights given bythe managers was considered for analysis. These weights wereused to evaluate the importance of these critical attributes forthe horizontal collaboration and the linguistic terms to evaluatethe extent of collaboration based on these attributes.

The focus group of 20 experts with robust knowledge and expe-rience of the semiconductor industry supply chain were consultedto carry out the pair-wise comparison of the importance of testattributes at Level 3 and attribute groups at Level 2. The feedbackwas collected and analyzed to obtain the weight vector of each ex-pert’s judgment matrix. The maximum eigen value, consistency in-dex (CI) and consistency ratio (CR) pertaining to each expert’sjudgment matrix were calculated as well. It was evident that allthe values of CR were less than 0.1. Therefore, it can be concluded

Page 10: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

START

Horizontal Collaboration Problem

Generate Evaluation Attributes (Brainstorming and expert opinion)

Industry Characteristics (IC) 1. Industry Structure (IS) 2. Financial Stability (FS) 3. Global Reputation (GR)

Competitive Advantages (CA) 1. Product Orientation (PO) 2. General Competitive Edge (GCE)

Internal Parameters (IP) 1. Operational Parameters (OP) 2. Tactical Parameters (TP) 3. Strategic Parameters (SP)

External Parameters (EP) 1. Product Characteristics (PC) 2. Industry Orientation (IO)

Developing Customized Configuration Hierarchy

Define evaluation rule

STOP

AHP-FLM MODULE

Assign weights to the importance of the attributes (by experts)

Conversion of weights into importance factors

Assign weights to the attributes depending on the impact on horizontal collaboration (by experts)

Conversion of weights into linguistic factors

Establish and evaluate the pair wise comparison matrices to determine the relative priority weights of the attributes and sub-attributes

AHP

CONSISTENCY TEST Evaluate if the comparison matrix conforms with the

consistency ratio

No

Yes

Fuzzy Logic

• Collect data for collaboration attributes • Defuzzify data for the attributes • Define membership function • Estimate membership value of all linguistic terms for the attributes • Estimate Crisp numbers

Relative weights of the collaboration attributes (Make it the reference for evaluation)

Crisp Scores of all the collaboration attributes and sub-attributes

Aggregate the weightings and fuzzy ratings of the attributes to obtain the defuzzified collaboration intensity index (CII)

Linguistic Terms for CII

Fuzzify CII

Intensity of Collaboration

NC PC CC

Fig. 5. Flow chart depicting the overall logic of the proposed (AHP–FLM) model.

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 889

that a reasonable level of consistency has been achieved by usingexpert judgment. Then the average weight vectors of test attri-butes at Level 3 and attribute groups at Level 2 were determined.Further, based on a thorough investigation of semiconductor

industry supply chain and the proposed project of horizontal col-laboration (based on ‘compatibility test’), the above 20 expertswere consulted to evaluate the performance or the degree of com-patibility among the chosen competitor for viability of this project.

Page 11: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

Table 4Transition the degree of agreement into linguistic terms.

Degree of agreement Linguistic representation

0.0–1.0 Extremely low1.0–1.5 Very low1.5–2.0 Low2.0–2.5 Marginally low2.5–3.0 Average3.0–3.5 Marginally high3.5–4.0 High4.0–4.5 Very high4.5–5.0 Extremely high

Table 5Estimation of average importance of attributes and its impact on horizontalcollaboration.

Collaborationattributes

Average importance ofthe attributes

Impact on horizontalcollaboration

Linguisticterm

Critical attributes (Level 2) and sub-attributes (Level 3)IC 4.50 4.45 VHCA 4.10 3.90 HIP 4.90 4.85 EHEP 3.50 4.25 VH

IS 4.50 4.10 VHFS 4.90 4.75 EHGR 3.30 3.60 H

PO 3.50 4.65 EHGCE 4.70 4.45 VH

SP 4.70 4.70 EHTP 4.50 4.25 VHOP 3.70 3.70 H

PC 3.40 3.85 HIO 4.80 4.55 EH

890 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

The appraisal set of performance included nine grades, such as ex-tremely low, very low, low, marginally low, average, marginallyhigh, high, very high and extremely high.

5.1. Conversion of degree of agreement into relative importance factors

The managers of the semiconductor industries were asked togive importance to the attributes in the light of horizontal collab-oration among similar service providers. The responses were askedon the scale ranging from 1 to 5. The weights given by the expertslie in the range 3.2–5.0 for all the four critical attributes and sub-attributes. These values could not be used directly for finding therelative weights of the mechanisms. These were converted intothe relative importance factors, which were used to weigh eachattribute. The degrees were transformed into AHP subjective scalefor pair-wise comparisons. The difference in weights of any twoattributes helped to know the importance of one attribute over an-other for horizontal collaboration among competing SC members.The maximum difference between the degrees (or weights) as-signed to all the critical attributes and sub-attributes were 1.8,the minimum difference being 0.2. The difference between themaximum and minimum values was divided into nine intervals(i.e., from 0.2 to 1.8) with an interval difference of 0.2. The mini-mum difference of 0.2 has been assigned a relative importance of1 and the maximum difference of 1.8 has been assigned a relativeimportance of 9, as per the subjective scale of AHP. Similarly, for allthe values of difference between the degrees of agreement wereassigned the relative importance as shown in Table 3.

The relative importance of the attributes was not unique, as itwas dependent on the nature of business of the similar service pro-vides in the semiconductor industry supply network. The differ-ence between the minimum and the maximum weights weredivided into nine equal intervals. This transformation subjectivescale of AHP into weights helped in the pair-wise comparison ofcritical attributes. These weights and hence the difference betweenminimum and maximum weights was different for different mem-bers in the semiconductor industry supply network.

5.2. Conversion of degree of agreement into linguistic term

The importance of horizontal collaboration can be realized onlywhen there is some improvement in the performance of the semi-conductor industries by utilizing the critical attributes. As thequestions regarding the importance of these parameters and itsimpact on horizontal collaboration were asked to the experts usinguniform Likert scale, the impact on horizontal collaboration wascaptured by taking the average of the weights for a particular attri-bute. The degrees of agreement in the form of average weightswere mapped to assign linguistic terms as shown in Table 4. Theimportance given by the semiconductor industry experts to differ-ent attributes and hence the linguistic terms assigned to theseattributes in the light of horizontal collaboration are given below:

Table 3Conversion of differences in the degree of agreement into AHP weights.

Difference in the degree of agreement AHP Subjective scale

0.2 The parameter is equally im0.4 The parameter importance0.6 The parameter is weakly m0.8 The parameter importance1.0 The parameter is strongly m1.2 The parameter importance1.4 The parameter is very stron1.6 The parameter importance1.8 The parameter is absolutely

The managers were asked to rate the degree of agreement (in ascale of 1–5 regarding industry characteristics, competitiveadvantage, internal parameters, and external parameters.

The average weight of various attributes as has been evaluatedby the selected experts is depicted in the following Table 5. Afterknowing the importance, the degrees of agreement regarding thecompatibility for horizontal collaboration by using these parame-ters were evaluated by the respondents, which are also depictedin Table 5. The average weight of the assessment scores by themanagers was estimated and was converted into linguistic terms.

5.3. Estimation of collaboration intensity index

The transformation of degrees of agreement into relative impor-tance and linguistic terms are depicted in Table 5. Pair-wise com-parisons were made to estimate the relative weights of all the fourcritical attributes and all the sub-attributes of the ‘compatibilitytest’ for horizontal collaboration, an illustration of which is

Weight

portant to other parameter 1is a compromise between 1 and 3 to other parameter 2ore important to other parameter 3is a compromise between 3 and 5 to other parameter 4

ore important to other parameter 5is a compromise between 5 and 7 to other parameter 6gly more important to other parameter 7is a compromise between 7 and 9 to other parameter 8

more important to other parameter 9

Page 12: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

Table 6aPair-wise comparison of critical attributes.

Criticalattributes

IC CA IP EP Alternativepriority weights

Consistency check

IC 1 3 1/3 5 0.2650 kmax = 4.12CA 1/3 1 1/5 3 0.1234 CI = 0.04IP 3 5 1 7 0.5606 CR = 0.0444EP 1/5 1/3 1/7 1 0.0518

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 891

depicted in Tables 6a, 6b, 6c, 6d, 6e. These matrices have focusedon the estimation of priority weights of the critical attributes andsub-attributes, and evaluation of consistency ratios (CR) of thejudgment matrices for the case study. These estimates also helpedin determining the sensitivity of the potential partners with re-spect to the critical attributes.

Consistency means that the decision maker is exhibiting acoherent judgment in specifying the pair-wise comparison of thevarious attributes. The CR helps to determine the level of consis-tency in the pair-wise comparison table. As CR < 0.1, the level ofconsistency in the judgment of the managers was acceptable(Saaty, 1990). Otherwise, the pair-wise scores have to be revised.

The sensitivity of potential partner with respect to the attri-butes and sub-attributes are depicted in Fig. 6a and b. The final pri-ority weights of different attributes showed that the internalparameters (IP) of the partnering organization carried the highestpriority for horizontal collaboration, and it was followed by indus-

Table 6bPair-wise comparison of sub-attributes of industry characteristics.

Sub-attributes IS FS GR Alternativepriority weights

Consistency check

IS 1 1/3 6 0.2895 kmax = 3.0742FS 3 1 8 0.6464 CI = 0.0371GR 1/6 1/8 1 0.0641 CR = 0.0639

Table 6cPair-wise comparison of sub-attributes of competitive advantage.

Sub-attributes PO GCE Alternative priority weights Consistency check

PO 1 1/6 0.1428 kmax = 2GCE 6 1 0.8571 CI = 0

CR = 0

Table 6dPair-wise comparison of sub-attributes of internal parameters.

Sub-attributes SP TP OP Alternativepriority weights

Consistency check

SP 1 4 5 0.6652 kmax = 3.0866TP 1/4 1 3 0.2311 CI = 0.0433OP 1/5 1/3 1 0.1038 CR = 0.07468

Table 6ePair-wise comparison of sub-attributes of external parameters.

Sub-attributes PC IO Alternative priority weights Consistency check

PC 1 1/7 0.1250 kmax = 2IO 7 1 0.8750 CI = 0

CR = 0

try characteristics (IC), competitive advantage (CA), and externalparameters (EP). Under the internal parameters of the partneringorganization, the impact of strategic parameters (SP) is generallydecided by the outsourcing strategies and the attitude of the topmanagement. These factors impact most to the internal parameterscompared to tactical and operational parameters (TP and OP),therefore it has got a higher priority weight. Financial stability(FS) of the partnering organization depends on the business perfor-mance as well as the capital required or available for the collabo-rative initiative. As financial stability (FS) has a strong impact onthe industry characteristics (IC), it has been identified as the mostimportant element under the attribute of industry characteristics(IC), and carried higher priority weights in this particular study.Apart from this, general competitive edge (GCE) under the compet-itive advantage (CA) attribute (which is dependent on marketshare, customer orientation and technology standard/level), andindustry orientation (IO) under external parameters (EP) attribute(which is dependent on financial performance, profit potential andresource utilization) carried high priority weight, and it seemed tobe practically very relevant in the current political, economic, andbusiness environment.

Further, the range of linguistic terms was assigned and the crispscore for all ranges were estimated according to the Expression (1),which has been depicted in Table 7.

The values of relative weights and crisp scores of the attributesas per their linguistic term were substituted in Expression (2) toestimate the intensity of collaboration, which is depicted in Table8. For an ideal case of horizontal collaboration, all the attributescontribute equally to the performance of semiconductor industrysupply chain (SSC). The extent of collaboration for an ideal casecan be estimated by considering all the attributes to be equallyimportant with relative weights of 0.25 for each attribute andthe linguistic term to be assigned in an ideal case is being 0.864.So, the extent of collaboration in an ideal case was estimated tobe 4 � 0.25 � 0.864 = 0.864. We defined relative CII as the overallscore expressed as the percentage of the most ideal score. For theindustry considered in our case study, the overall score of 0.8207has a relative CII of (0.8207/0.864) i.e., 0.9499 (approximately95%), of the ideal case of collaboration.

The collaboration intensity index (CII) of 0.8207 is in the crispform, which was based on a scale of 0–1. The partner evaluationrule selects a partner whose crisp scores are ‘‘high” (relates to com-plete collaboration, CC). The selection criteria require comparisonwith the linguistic term ‘‘high”. To perform this comparison, thecrisp score was used to compute the membership degree (scale 0to 1) from fuzzy membership functions for CII for values ‘‘high”.Fig. 7 shows the membership function for fuzzy classification ofCII criteria into ‘‘low”, ‘‘medium” and ‘‘high”, which relates to nocollaboration (NC), partial collaboration (PC), and complete collab-oration (CC), respectively. The membership function forCII = ‘‘high” has a value 0 starting at CII score of 0.682 and reachesa maximum value of 1 (one) when CII score is 0.864. For CII score,the membership degree for the fuzzy function CII = ‘‘high” may becomputed. For any CII score below 0.682, the membership degreemay be computed by using membership equation x�0:682

0:182

� �, where

x represents the crisp CII score.For the case study on hand, the membership degree ¼

0:8207�0:6820:182

� �¼ 0:762. The semiconductor industries were inclined

for a collaborative partnership with the similar service providerswith a membership degree of more than or equal to 0.5. For amembership degree of 0.5, the crisp score can be computed as[0.5 � 0.182 + 0.682] = 0.773. The relative CII in this case wasestimated as [0.773/0.864] = 0.8946. So, the collaborative relation-ship with a relative CII value of 89.46% or more have a ‘‘high”probability of success, i.e., a successful complete collaboration(CC) relationship may be ensured.

Page 13: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

0

0.1

0.2

0.3

0.4

0.5

0.6

EPIPCAIC

Pri

orit

y W

eigh

ts

Attributes

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

IS FS GR PO GCE SP TP OP PC IO

Pri

orit

y W

eigh

ts

Sub-attributes

a b

Fig. 6. (a) Sensitivity of the potential partner with respect to major attributes and (b) sensitivity of the potential partner with respect to sub-attributes.

Table 7Crisp scores of the linguistic terms.

Linguistic terms Crisp score

Extremely low 0.136Very low 0.227Low 0.318Marginally low 0.409Average 0.500Marginally high 0.591High 0.682Very high 0.773Extremely high 0.864

Table 8Estimation of intensity or degree of horizontal collaboration.

Factors Weightsby AHP

Degree ofagreement forhorizontalcollaboration

LinguisticTerm

Crispscore

Degree ofcollaboration

By considering all the critical attributes and the sub-attributesIC 0.2650 4.45 VH 0.773 Collaboration

intensity index(CII) = 0.8207

CA 0.1224 3.90 H 0.682IP 0.5606 4.85 EH 0.864EP 0.0518 4.25 VH 0.773

IS 0.2895 4.10 VH 0.773FS 0.6464 4.75 EH 0.864GR 0.0641 3.60 H 0.682

PO 0.1428 4.65 EH 0.864GCE 0.8571 4.45 VH 0.773

SP 0.6652 4.70 EH 0.864TP 0.2311 4.25 VH 0.773OP 0.1038 3.70 H 0.682

PC 0.1250 3.85 H 0.682IO 0.8750 4.55 EH 0.864

892 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

Similarly, the membership function for CII = ‘‘low” has a value of1 starting at CII score of 0 and reaches a value of 0 when CII score is0.409. For CII score, the membership degree for the fuzzy functionCII = ‘‘low” may be computed. For any CII score below 0.409, themembership degree may be computed by using membership equa-tion 0:409�x

0:409

� �, where x represents the crisp CII score. The semicon-

ductor industries were not inclined for a collaborativepartnership with the similar service providers with a membershipdegree equal to 0. For a membership degree of 0, the crisp scorewas computed as [0.409 � 0 � 0.409] = 0.409. The relative CII inthis case was estimated as [0.449/0.864] = 0.4734. So, the collabo-

rative relationship with a relative CII value of 47.34% or less have a‘‘low” probability of success, i.e., no collaboration (NC) is possible.So, the nature of relationship can be predicted depending on therelative CII values for any two members, as depicted in Table 9.

5.3.1. Significance of collaboration intensity indexThe relative CII has certain implications for the industry under

consideration, which are as follows:

1. This parameter is a determining factor for the positioning of theindustry for long-term survival.

2. It establishes a perfect balance between the firms’ technologicaldevelopment and strategic orientation to facilitate acquisitionof competencies for horizontal collaboration.

3. It emphasizes organizational and managerial involvement to becritical as financial involvement in the creation of resources.

4. It demonstrates that the strong internal capabilities and strate-gic coherence with partner’s business policy tends to favourresource creation.

5. When new technologies increase in complexity, their develop-ment requires that the firms acquire external capabilities tocomplement its internal ones used in new productdevelopment.

6. In order to maximize output and resource creation, collabora-tive initiatives be organized in such a manner that all the SCmembers can be both responsible and autonomous, but are atthe same time united by a cohesive structure. The success ofsuch a structure requires that the managers be focused to part-ner firms’ financial and resource involvement and to their moti-vation to acquire competencies.

6. Discussions and managerial implications

The research highlights the attributes of collaborative practicesthat drive operational performance. Similar services/goods provid-ers differed in the degree of importance to the collaborative attri-butes. This was due to the differing priorities of the variousmembers of the supply chain. As collaborative efforts add costsand brings benefits differently to the chain members, the identifi-cation of most appropriate attributes would enable the chainmembers to understand each others’ concerns and find effectivesolutions that benefit all parties. The research also emphasizesthe importance of collaborative efforts in attaining better perfor-mance. As there are different expectations from collaborative prac-tices among the chain members, they need to share concerns inorder to clearly set objectives and devise plans to achieve those

Page 14: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0

1

Degree of Membership

0 Collaboration Intensity Index

NC PC CC

Fig. 7. Illustration of intensity of horizontal collaboration.

Table 9Range of CII for collaborative relationships between SC members.

Range of relative CII Relationship among SC partners

0–47.34% No collaboration (NC)47.35–89.45% Partial collaboration (PC)More than or equal to 89.46% Complete collaboration

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 893

objectives. Management from both sides must put together anintegrated strategy in order to initiate improvement in specific col-laborative practices.

The analysis of horizontal collaboration in the context of ‘com-patibility test’ in terms of strategic partner selection leads to anumber of managerial insights, which are as follows:

1. The framework will help management of the semiconductormanufacturing industries to enable a quick analysis regardinghorizontal collaboration with similar industries based on ‘com-patibility’ test. A desirability of a partnership can be evaluated.

2. Currently a few of these parameters are being used by thisindustry, which only enables informal collaboration; the pro-posed framework would be able to establish a strategic compat-ibility between partners willing to initiate resources or processtools sharing for horizontal collaboration.

3. This framework is not intended to perform a detailed analysispertaining to all the costs and benefits due to horizontal collab-oration through strategic compatibility test. It would provide aninsight for the strategic partners to determine and allocate costsand benefits to survive in the e-market.

4. It is proposed to use a multi-disciplinary group of decision mak-ers to reduce the subjective bias of one decision maker. The con-sistency of the judgments should also be checked.

5. The framework is able to track down why a certain element ispreferred above another. This will facilitate in determining therequired improvements in the practices necessary of the indus-tries for a strategic partnership.

Based upon the hybrid approach proposed for horizontal collab-oration in this paper, the following guidelines are proposed to helppractitioners to link a manager’s intuition and judgement with thehybrid approach:

(a) The hybrid approach focuses on improving the effectivenessof the horizontal collaboration strategy development pro-cess, helping managers in the decision-making process,and supporting the managerial judgement and intuition. Asexperienced managers have good judgement and intuition,they should always be an integral part of the strategy devel-opment process. Managers get support from the systemwithout the need to understand or develop analysis models.

Managers also retain reasonable control over the strategydevelopment process and the most appropriate criticalattributes.

(b) Generally, managers are flexible and creative. However,managerial judgement and intuition may be limited byexperience, background and social environments. Many ofthem lack strategic analysis skills. The proposed systemcan provide analysis aids and information processing sup-port. The system is unbiased and consistent, but rigid. Boththe system’s consistency and managerial flexibility shouldbe incorporated. A strong and balanced interplay betweenthe decision makers and the hybrid system should beattained to produce a total effect for the strategy develop-ment process through utilising the powers of both parties.

(c) The group decision-making process is intended to developgroup judgement and consensus. The fuzzification and theevaluation rules are intended to support and complementmanagerial judgement and expertise. So, a combined useof the proposed system’s general knowledge and the manag-ers’ specific knowledge about their products and markets inthe perspective of collaboration should always beencouraged.

The practical implication of this study is that AHP–FLM ap-proach will be more appropriate for high-value components wherestringent purchasing criteria are required. In contrast, AHP remainsan appropriate approach for relatively lower value components. Insummary, the novelty of this study lies in the application of a hy-brid approach to a real semiconductor industries scenario. Thisstudy has considered an important issue of supply chain manage-ment, providing a better decision for partnership selection duringhorizontal collaboration using appropriate quantitative tech-niques. In addition to exploring the insights from the perspectiveof managers, this research has consulted the academic literaturerelated to horizontal collaboration. By doing so, we were able tocross check our research findings, and make them more persuasiveand precise.

7. Concluding remarks

This paper aims to provide a generic quantitative model to com-prehensively assess the degree of collaboration that come withindividual relationships and to evaluate whether such a project isreally viable. This research is significant because the informationpresented provides a comprehensive collaboration assessment toolto all SC members concerned. It is concluded that while opportuni-ties are great, the compatibility of supply network members forhorizontal collaboration should be properly assessed, and the pro-posed AHP–FLM approach is suitable for such tasks. This approachfor ‘compatibility test’ for horizontal collaboration enables us tomake decisions based on vague or imprecise data. In this method,the values from pair-wise comparison of each criterion are ex-pressed with triangular fuzzy numbers. By this way, we can dealwith the uncertainties of decision problem by using the conceptsof fuzzy numbers and linguistic variables to evaluate the factorsin such a manner that the view points of an entire decision-makingbody can be expressed without any constraint.

Compared with AHP, the proposed hybrid approach combinessubjective analysis with quantitative analysis more reasonably,synthesizes group opinions more adequately. The results of thecase study for compatibility among members indicate that the pro-posed approach results in fair and reasonable conclusions. The out-come of this paper provides an effective method for thesemiconductor industry supply chain members to evaluate thesuccess of such shared collaborative systems in a structured and

Page 15: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

894 B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895

simple manner. The case study has demonstrated the thoughtful-ness, flexibility, and efficiency of the proposed model to directlytap the subjectivity and preferences of the decision makers. Theproposed method presented herewith should be equally useful inanalyzing and assessing the success of any such projects elsewhereby selecting industry-specific attributes and by involving expertsfrom those industries.

The present study also explains how SC members are willing todevelop collaborative practices through continuously improvinginternal and external processes, competitive advantage and indus-try characteristics. However, the chain members need to modifyand customize the approaches for collaborative practices that suittheir unique environment under which they operate. The contribu-tions to the literature on supply chain collaboration which resultedfrom this study are (1) The study proposed and tested 10 dimen-sions (attributes) for effective SC collaboration. (2) Collaborativeefforts among chain members should be encouraged in order toimprove operational performance. (3) As the SC members variedin their perception of the importance of these attributes that differ-entiate between high and low performing collaborative practices, itmay be recommended that the chain members should create anegotiation mechanism that encourages the collaborative practicesto achieve better performance. The contribution of this study pro-vides useful managerial insights into the development of horizon-tal collaborative practices in the supply network. Theoperationalization of four critical attributes and sub-attributescan be expanded based on focus groups or case studies, i.e., newattributes can be added to the present work to demonstrate moreeffective ‘compatibility test’.

Our study has certain limitations, which are as follows:

(1) The study did not consider the fierce competition from thesimilar SC members from a global perspective;

(2) The study did not consider the management structures ofcollaborative procurement initiative as a result of culturaldifferences from a global perspective;

(3) The study does not discuss whether a decision maker exertsany influence on mental cognition and experiential charac-teristics when rating the linguistic interval scale.

It may be observed that many conflicts arise at the point of devel-opment of the collaborative initiative. So, understanding the maindimensions of culture and how they affect conflict may assist in cre-ating a common culture for the collaborative initiatives and providesfor managing conflicts as they emerge. However, it seems that onlysevere circumstances, such as threat from global competition canmotivate the formation of large-scale horizontal collaboration.

Despite the advantages of the proposed hybrid approach (AHP–FLM) for partner evaluation problem, more research is certainlycalled for within the context of studying a more complex supplychain with multiple supply network and nodes as well as investi-gating other hybrid methods for optimum partner selection. Differ-ent alternative methodologies, such as fuzzy extended AHP (Chan& Kumar, 2007), fuzzy analytic network process (ANP), fuzzy TOP-SIS, and fuzzy electre can also be implemented to tackle the com-plex problem of partner evaluation (Sevkli, Koh, Zaim, Demirbagh,& Tatoglu, 2008). Further scope of this research is to develop multi-criteria optimization models for establishing contributions to bemade by compatible partners in terms of resources sharing for cer-tain joint product development and supply.

Acknowledgement

The authors are grateful to the reviewers and the editor-in-chieffor their valuable comments and suggestions on improving themanuscript.

References

Barratt, M. (2004). Understanding the meaning of collaboration in the supply chain.Supply Chain Management: An International Journal, 9(1), 30–42.

Belton, V., & Stewart, T. R. (2002). Multiple criteria decision analysis: An integratedapproach. Norwell, NA: Kluwer Academic Publishers.

Benyoucef, B., & Canbolat, M. (2007). Fuzzy AHP-based supplier selection in e-procurement. International Journal of Services and Operations Management, 3(2),172–192.

Bevilacqua, M., & Petroni, A. (2002). From traditional purchasing to suppliermanagement: A fuzzy logic-based approach to supplier selection. InternationalJournal of Logistics: Research and Applications, 5(3), 235–255.

Beynon, M., Curry, B., & Morgan, P. (2000). The Dempster–Shafer theory of evidence:An alternative approach to multicriteria decision modeling. Omega, 28(1),37–50.

Caraynnis, E. G., & Alexander, J. (2004). Strategy, structure and performance andissues of precompetitive R&D consortia: Insights and lessons learnt fromSEMATECH. IEEE Transactions on Engineering Management, 51(2), 226–232.

Chan, F. T. S., & Kumar, N. (2007). Global supplier development considering riskfactors using fuzzy extended AHP-based approach. Omega, 35(4), 417–431.

Chang, S. L., Wang, R. C., & Wang, S. Y. (2006). Applying fuzzy linguistic quantifier toselect supply chain partners at different phases of product life cycle.International Journal of Production Economics, 10, 348–359.

Chen, S. J., & Hwang, C. L. (1992). Fuzzy multiple attribute decision making-methodsand applications. Berlin: Springer.

Chen, G., & Pham, T. T. (2001). Introduction to fuzzy sets, fuzzy logic, and fuzzy controlsystems. Florida: CRC Press.

Cousins, P. D. (2002). A conceptual model for managing long-term inter-organisational relationships. European Journal of Purchasing and SupplyManagement, 8(2), 71–82.

Cruijssen, F., Cools, M., & Dullaert, W. (2007). Horizontal cooperation in logistics:Opportunities and impediments. Transportation Research, 43(2), 129–142.

Fu, H. P., Ho, Y. C., Chen, R. C. Y., Chang, T. H., & Chien, P. H. (2006). Factors affectingthe adoption of electronic marketplaces. International Journal of Operations andProduction Management, 26(12), 1301–1324.

Hajidimitriou, Y. A., & Georgiou, A. C. (2002). A goal programming model for partnerselection decisions in international joint ventures. European Journal ofOperational Research, 138(3), 649–662.

Harvey, M. G., & Lusch, R. F. (1995). A systematic assessment of potentialinternational strategic alliance partners. International Business Review, 4(2),195–212.

Hwang, H.-S. (2004). Web-based multi-attribute analysis model for engineeringproject evaluation. International Journal of Computers & Industrial Engineering,46(1), 669–678.

Ireland, R., & Bruce, R. (2000). CPFR: Only the beginning of collaboration. SupplyChain Management Review(September/October), 80–88.

Kahraman, C., Cebeci, U., & Ululan, Z. (2003). Multi criteria supplier selection usingfuzzy AHP. Logistics Information Management, 16(6), 382–394.

Kapur, V., Peters, J., & Berman, S. (2003). High-tech 2005: The horizontal,hypercompetitive future. Strategy & Leadership, 31(2), 34–47.

Lau, R. S. (2002). Competitiveness factors and their relative importance in the USelectronics and computer industries. International Journal of Operations &Production Management, 22(1), 125–135.

Lee, E. S., & Shib, H. S. (2001). Fuzzy and multi-level decision making. An interactiveComputational Approach. London: Springer-Verlag.

Lee, H. L., & Whang, S. (2000). Information sharing in a supply chain. InternationalJournal of Technology Management, 20(3 & 4), 373–387.

Lin, C-W. R., & Chen, H-Y. S. (2004). A fuzzy strategic alliance selection frameworkfor supply chain partnering under limited evaluation resources. Computers inIndustry, 55(2), 159–179.

Link, P., & Marxt, C. (2004). Integration of risk- and chance management in thecooperation process. International Journal of Production Economics, 90(1),71–78.

Liu, F. H., & Hai, H. L. (2005). The voting analytic hierarchy process method forselecting supplier. International Journal of Production Economics, 97, 308–317.

lp, W. H., Huang, M., Yung, K. L., & Wang, D. (2003). Genetic algorithm solution for arisk-based partner selection problem in a virtual enterprise. Computers andOperations Research, 30(2), 213–231.

Maggie, C. Y., & Rao, V. M. (2001). An application of the AHP in vendor selection of atelecommunications system. Omega, 29(2), 171–182.

Mason, R., Lalwani, C., & Boughton, R. (2007). Combining vertical and horizontalcollaboration for transport optimization. Supply Chain Management: AnInternational Journal, 12(3), 187–199.

McCarthy, S., & Golocic, S. (2002). Implementing collaborative planning to improvesupply chain performance. International Journal of Physical Distribution &Logistics Management, 32(6), 431–454.

Mikhailov, L. (2002). Fuzzy analytical approach to partnership selection information of virtual enterprises. Omega, 39, 393–401.

Murphy, C. K. (1995). Limits on the analytic hierarchy process from its consistencyindex. European Journal of Operational Research, 65, 138–139.

Nguyen, H. T., & Walker, E. A. (2000). A First Course in Fuzzy Logic (2nd ed.). BocaRaton, FL: Chapman & Hall, CRC.

Prodanovic, P., & Simonovic, S. P. (2003). Fuzzy compromise programming for groupdecision making. IEEE Transactions on Systems, Man, and Cybernetics, Part A, 33,358–365.

Page 16: Bahinipati_2009_Horizontal Collaboration in Semiconductor Manufacturing Industry Supply Chain

B.K. Bahinipati et al. / Computers & Industrial Engineering 57 (2009) 880–895 895

Ross, T. J. (1997). Fuzzy logic with engineering applications. New York, NY: McGrawHill.

Saaty, T. L. (1990). Multicriteria decision making: The analytic hierarchy process.Pittsburgh, PA: RWS Publications.

Sabath, R., & Fontanella, J. (2002). The unfulfilled promise of supply chaincollaboration. Supply Chain Management Review(July/August), 24–29.

Saen, R. F. (2007). Supplier selection in the presence of both cardinal and ordinaldata. European Journal of operational Research, 183(2), 741–747.

Sevkli, M., Koh, S. C. L., Zaim, S., Demirbagh, M., & Tatoglu, E. (2008). Hybridanalytical hierarchy process model for supplier selection. Industrial Managementand Data Systems, 108(1), 122–142.

Shore, B., & Venkatachalam, A. R. (2003). Evaluating the information sharingcapabilities of supply chain partners: A fuzzy logic model. International Journalof Physical Distribution and Logistics Management, 33(9), 804–824.

Tella, E., & Virolainen, V.-M. (2005). Motives behind purchasing consortia.International Journal of Production Economics, 93 & 94, 161–168.

Tolga, E., Demircan, M., & Kahraman, C. (2005). Operating system selection usingfuzzy replacement analysis and analytic hierarchy process. International Journalof Production Economics, 97, 89–117.

Vaidya, O. S. (2006). Analytic hierarchy process: An overview of applications.European Journal of Operational Research, 169, 1–29.

Wang, H. S., & Che, Z. H. (2007). An integrated model for supplier selection decisionsin configuration changes. Expert Systems with Applications, 32(4), 1132–1140.

Wang, G., Huang, S. H., & Dismukes, J. P. (2004). Product-driven supply chainselection using integrated multi-criteria decision-making methodology.International Journal of Production Economics, 91(18), 1–15.

Zhang, D. Y., Xu, X. F., & Wang, G. (2004). Process and method of model reuse foragile virtual enterprise. Computer Integrated Manufacturing Systems, 10(1),23–29.