Expert Systems with Applications 39 (2012) 3404–3412
Contents lists available at SciVerse ScienceDirect
Expert Systems with Applications journal homepage: www.elsevier.com/locate/eswa
Evaluation of new service concepts using rough set theory and group analytic hierarchy process Changyong Lee a,1, Hakyeon Lee b,2, Hyeonju Seol c,3, Yongtae Park a,⇑ a
Seoul National University, San 56-1, Shillim-Dong, Kwanak-Gu, Seoul 151-742, South Korea Seoul National University of Technology, 172 Gongreuing 2-dong, Nowon-gu, Seoul 139-746, South Korea c Korea Air Force Academy, 335-1, Ssangsu-ri, Namil-myeon, Cheongwon-gun, Chungbuk 363-849, South Korea b
a r t i c l e
i n f o
Keywords: Service engineering New service development (NSD) Concept evaluation New service concept (NSC) Analytic hierarchy process (AHP) Rough set theory (RST)
a b s t r a c t One of the most crucial stages in the new service development (NSD) process is concept selection, which is characterized by newly generated alternatives and vaguely defined concept evaluation criteria. Although a number of factors have been found to be influential, a lacuna remains as to how to make a strategic decision based on influential factors. This study proposes a systematic approach to evaluation of new service concepts (NSCs) by integrating the merit of group analytic hierarchy process (AHP) in modeling multi-criteria decision making (MCDM) problems and the strength of rough set theory (RST) in handling subjectivity in concept evaluation. The suggested approach is designed to be executed in four discrete stages. First of all, a hierarchical AHP model for the evaluation of NSCs is constructed in terms of strategy, finance, market, technology, and implementation. Second, pairwise comparisons are made among criteria and sub-criteria, and preferences to NSCs with respect to the sub-criteria are obtained by domain experts. Third, the individual judgments obtained at the preceding stage are aggregated into group judgments. Finally, the NSCs are prioritized based on risk propensity of decision makers. A case study of the video game service is presented to illustrate the suggested approach. We believe that our method can promote consensus building on the promising NSCs. Ó 2011 Elsevier Ltd. All rights reserved.
1. Introduction The strategic importance of new service development (NSD) is more apparent than ever as the market shifts rapidly and the intensity of competition in the service sector is ever intensified (Lundkvist & Yakhlef, 2004). It has become the norm for successful firms to have to consistently develop new services if they are to gain or maintain a competitive edge in such a turbulent environment (Mathieu, 2001; Meyer & Detore, 2001; Oliva & Kallenberg, 2003); consequently, there is a growing body of knowledge on NSD (Johne & Storey, 1998). In recent years, parallel to the concept of NSD that is strictly market-oriented discipline, service engineering has emerged as an engineering-centric one, and has been paid increasing attention from academia and practice alike (Lee, Kim, & Park, 2010). It is concerned with systematic development of services using suitable models, methods, and tools (Bullinger, Fahnrich, & Meiren, 2003).
⇑ Corresponding author. Tel.: +82 2 878 8358; fax: +82 2 878 8560. E-mail addresses:
[email protected] (C. Lee),
[email protected] (H. Lee),
[email protected] (H. Seol),
[email protected] (Y. Park). 1 Tel./fax: +82 2 878 3511. 2 Tel.: +82 2 970 6469; fax: +82 2 974 2849. 3 Tel.: +82 43 290 6492; fax: +82 43 297 0479. 0957-4174/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2011.09.028
One of such areas is concept selection. As is in the new product development (NPD) (Krishnan & Ulrich, 2001), the NSD process is comprised of a series of decision making. Among others, selecting new service concepts (NSCs) is regarded as one of the most crucial decisions because it directly influences the direction of remaining activities (Alam & Perry, 2002); concept evaluation is a core prerequisite activity for promoting consensus building on the most promising NSCs. Contrary to the extensive body of literature on concept evaluation in NPD, however, little attention has been paid to the evaluation of NSCs. It has been noted that methodological implications have rarely been investigated despite its importance to the overall process (Davison, Watkins, & Wright, 1989). Most firms have also been observed to use informal procedures and qualitative methods that primarily hinge on human intuition and experience (Easingwood, 1986; Edgett, 1993). Evaluating NSCs differs from other stages in the NSD process such as concept testing and evaluation of the service quality for the following reasons. Firstly, concept testing is similar to concept evaluation in that both stages aim to further narrow down the set of service concepts, but distinct in that concept testing focuses on market potential (Meiren & Burger, 2010). A set of factors affecting the selection of NSCs should be taken into account in the concept evaluation stage to determine whether the service concept is feasible and reasonable to develop and run the business. Secondly, the evaluation of service quality is aimed at determining whether
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
customers are satisfied or not based on service outcomes and customer responses while concept evaluation is performed to promote consensus building on the promising NSCs among newly generated alternatives (Goldstein, Johnston, Duffy, & Rao, 2002). There usually lacks sufficient and adequate supplementary information to aid decision making in the evaluation of NSCs (Johne & Storey, 1998). Under such considerations, the concept evaluation problem needs to address the following two methodological issues. The first issue is how to manipulate the subjective perceptions. Concept evaluation is a vague and subjective task because it is difficult to describe the exact concepts and benefits of services (Edvardsson, Haglund, & Mattsson, 1995; Mohammed-Salleh & Easingwood, 1993). There also lack sufficient and adequate supplementary information such as prototypes (or concept sketches) and annotations of key technical features in NPD (Johne & Storey, 1998; Ulrich & Eppinger, 2003). As a consequence, the concept evaluation in NSD involves subjective judgments, instead of numerical expressions and objective decisions. In this respect, the weights of evaluation criteria and performance ratings of NSCs are often not available and have to be assessed subjectively by domain experts. The second issue is data sample size. The prototyping and simulation approach in NPD may generate voluminous and various types of data. However, due to nature of the service concepts, it is usually difficult to obtain a large amount of data in the early stage of NSD process. The suggested approach can therefore be applied when the data set is small in size, as contrary to the conventional methods that require several statistical assumptions on data sample size. Put those things together, the primary purpose of this study is to propose a systematic approach to the evaluation of NSCs that meets the issues mentioned above. The concept evaluation in NSD is still in its infancy, but methods and tools are not brand-new but the existing ones in the more established research fields of industrial engineering, operations research, and computer science (IBM Research, 2004). In this respect, King and Sivaloganathan (1999) grouped the methods for concept evaluation into five categories: utility theory, analytic hierarchy process (AHP), graphical tools such as matrices, quality function deployment, and fuzzy set theory. Recent years have also seen a huge increase in integrated use of those methods. Among others, the fuzzy group AHP has received the most attention because of the abilities in handling the subjective human ideas and modeling multi-criteria decision making (MCDM) problems (Ayag˘ & Özdemir, 2009; Kuo, Tzeng, & Hwang, 2007; Saridakis & Dentsoras, 2008). However, the fuzzy group AHP is suffers the limitations that stem from pre-determined fuzzy membership function, as clarified next. First of all, the way of selecting the membership function has not yet been thoroughly established. It relies on the subjective and heuristic decisions of domain experts, and severely affects the performance of concept evaluation (Walczak & Massart, 1999). Some systematic approaches such as neural network can be utilized to tune the membership function, but they are not feasible in the evaluation of NSCs due to complexity and small data sample size. Second, the boundary interval that denotes the degree of subjectivity is fixed with respect to the types of membership functions. This is not be true in reality, because the subjective perceptions vary across decision makers (Zhai, Khoo, & Zhong, 2008). As a remedy, this study proposes a systematic approach to the evaluation of NSCs using group AHP and rough set theory (RST). First, the group AHP is one of the most widely adopted MCDM methods that is effective in structuring group decisions and manipulating the qualitative and quantitative criteria, but ineffective when applied to an ambiguous problem (Lin, Lee, & Chen, 2009). Second, the RST is a mathematical tool capable of dealing with the imprecise and subjective judgments of domain experts by overcoming the previously noted membership function-related limitations of fuzzy set theory. It only relies on the original
3405
judgments without any assumptions on membership function and auxiliary information, and can be utilized even though the data set is small in size (Pawlak, 1982). By integrating the strength of RST in handling the subjectivity and the merit of group AHP in modeling MCDM problems, the suggested rough group AHP measures the feasibility of NSCs in terms of strategy, finance, market, technology, and implementation by taking the subjective perceptions in the concept evaluation into account. Most of the previous studies were limited to identifying factors that affect the selection of NSCs (Goldstein et al., 2002). Although a number of factors have been found to be influential, a lacuna remains in the literature as to how to make a strategic decision based on influential factors with the help of a systematic and quantitative approach. We believe that the suggested approach can promote consensus building about the promising NSCs. It is also expected that our method can be employed in various ambiguous problem, such as R&D project selection and technology selection. The remainder of this paper is organized as follows. A general background of RST and group AHP is presented in Section 2. The suggested approach is explained in Section 3 and illustrated with a comparative case study of the video game service in Section 4. This paper ends with conclusions in Section 5. 2. Methodological background 2.1. Rough set theory (RST) The rough set theory (RST), originally introduced by Pawlak (1982), is a mathematical tool capable of dealing with subjective and imprecise concepts. The distinct strengths of RST, vis-à-vis related other methods, lie in handling the subjective information without any assumptions and additional adjustments. Recent years, therefore, have seen a huge increase in the use of RST. In NPD, the RST has been employed for various problems such as concept evaluation (Zhai, Khoo, & Zhong, 2009), concept design (Shao, Chu, Qiu, Gao, & Yan, 2009; Tseng & Huang, 2008), and quality evaluation (Zhai, Khoo, & Fok, 2002). In the RST, any vague concept can be represented as a pair of precise concepts based on the lower and upper approximations (Pawlak, 1982), as depicted in Fig. 1. The lower approximation of X is the set of all objects that can be certainly included in X while the upper approximation set of X consists of the elements which cannot be characterized with certainty as belong or not to X (Greco, Matarazzo, & Slowinski, 2001). The difference of upper and lower approximation is a boundary region of X in U, which is composed of the elements that can neither be ruled in nor ruled out as member of the target set. According to Zhai et al. (2008), the basic concept of RST can be extended to manipulate imprecise and subjective human ideas. Assume that there is a set of n classes of human ideas, R = {C1, C2, . . . , Cn}, ordered in the manner of C1 < C2 < < Cn and Y is an arbitrary objects of U, then the lower approximation of Ci, upper approximation of Ci, and boundary region are defined as:
Lower approximation : PXðC i Þ ¼ [fY 2 U=RðYÞ 6 C i g
ð1Þ
Upper approximation : PXðC i Þ ¼ [fY 2 U=RðYÞ P C i g
ð2Þ
Boundary region : BR ¼ [fY 2 U=RðYÞ – C i g ¼ fY 2 U=RðYÞ < C i g [ fY 2 U=RðYÞ > C i g
ð3Þ
The human ideas can be represented by rough numbers on the basis of lower (Lim(Ci)) and upper limit (Lim(Ci)) that are referred to as the mean of elements in the lower approximation and upper approximation, respectively. In this regard, the interval of boundary region denotes the degree of preciseness; a rough
3406
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
Fig. 1. Schematic demonstration of rough set theory.
number with a larger interval of boundary region is interpreted as vaguer or less precise one. The human idea and interval of boundary region are expressed by following equations:
Human idea : RNðC i Þ ¼ ½LimðC i Þ; ðLimðC i ÞÞ
ð4Þ
Interval of boundary region : IBRðC i Þ ¼ LimðC i Þ LimðC i Þ
ð5Þ
The arithmetic operations of interval analysis can be applied to rough numbers. If RN1 = (L1, U1) and RN2 = (L2, U2) are two rough numbers and k is a nonzero constant, then the arithmetic operations are given by:
RN 1 þ RN 2 ¼ ðL1 ; U 1 Þ þ ðL2 ; U 2 Þ ¼ ðL1 þ L2 ; U 1 þ U 2 Þ
ð6Þ
RN 1 k ¼ ðL1 ; U 1 Þ k ¼ ðkL1 ; kU 1 Þ RN 1 RN 2 ¼ ðL1 ; U 1 Þ ðL2 ; U 2 Þ ¼ ðL1 L2 ; U 1 U 2 Þ
ð7Þ ð8Þ
2.2. Group analytic hierarchy process The AHP, first introduced by Saaty (1980), is one of the most widely adopted MCDM methods. It decomposes a problem into several levels making up a hierarchy where each decision element is considered to be independent. The distinct strength of method lies in the effective manipulation of quantitative criteria as well as qualitative ones (Lin et al., 2009). The AHP has been widely used in the group decision making context because it helps structure group decisions so that the discussion centers on objectives rather than on alternatives (Lai, Wong, & Cheung, 2002). The process of AHP in the group decision making context is composed of major four steps: (1) Model construction: The problem is decomposed into a hierarchical structure that consists of goal, criteria, sub-criteria, and alternatives. (2) Pairwise comparisons: The elements are compared pairwisely with respect to the importance to the goal, importance to the criterion, and importance to the sub-criterion. The relative importance values are determined on a scale of 1–9, where a score of 1 indicates equal importance between the two elements and 9 represents the extreme importance of one element compared with the other one. A reciprocal value is assigned to the inverse comparison; that is, aji = 1/ aij where aij denotes the importance of the ith element compared with the jth element. To test consistency of a pairwise comparison, a consistency ratio (CR) can be introduced with consistency index (CI) and random index (RI). If the CR is less than 0.1, the pairwise comparison is considered acceptable (Liu & Shih, 2005). (3) Group priority vectors: Once all the pairwise comparisons are made at the individual level, group priority vectors are generated by aggregating the individual judgments. (4) Final priorities: The finial priorities of alternatives are derived by multiplying the group priority vectors of criteria, sub-criteria, and alternatives.
The most stages of AHP procedure in the group decision making context follow the conventional AHP procedure. The difference is only in step 3 where it uses methods for deriving the group priority vectors. There are two basic ways of deriving the group priority vectors from individual judgments: aggregating individual judgments (AIJ) and aggregating individual priorities (AIP). The former is appropriate for group members that act together as a unit while the other is appropriate for separate individuals (Forman & Peniwati, 1998). A new pairwise comparison matrix at the group level is developed based on the individual judgments in the AIJ procedure. The group priority vectors, on the other hand, are derived from individual priority vectors in the AIP procedure. For more detailed information, see the text by Saaty (1980). 3. Proposed approach In this section, we examine the overall process of the proposed approach, giving a brief explanation of each stage at the same time. The suggested approach is comprised of four consecutive stages, as depicted in Fig. 2. Firstly, a hierarchical AHP model for the evaluation of NSCs is constructed based on the extensive literature review. Secondly, pairwise comparisons are made among criteria and sub-criteria, and preferences to NSCs are obtained by decision makers. Thirdly, individual judgments are transformed into group judgments by using rough numbers. Finally, the NSCs are prioritized based on the risk propensity of decision makers. 3.1. Construction of AHP model Even if an NSC seems to be attractive from customers’ viewpoints, a service firm may not be able to develop the service concept due to the problems such as lack of required technology or knowledge, resource unavailability, violation of regulations,
Fig. 2. Overall process.
3407
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
and so on. It should be considered at the very first time that it is possible or reasonable to implement the chosen service concepts and run the service business. Reviewing the relevant literature revealed that some of the criteria for concept evaluation in NSD are the same as those of NPD while others are specific to services. As in the case of NPD, profit, sales, and market share estimates are typical criteria for evaluating NSCs (Martin & Horne, 1993). Strategic fitness (Kelly & Storey, 2000), image of service firms (Easingwood, 1986), cannibalization of existing services (Johne, 1993), and breach of law (Kelly & Storey, 2000) were also mentioned in the previous studies. The aforementioned criteria address strategic, financial, and market issues well, but there are other issues that should be dealt with in the evaluation of NSCs. One of them is technology. Recently, academic researchers have recognized the critical importance of technology in the development and delivery of services (Meuter, Ostrom, Roundtree, & Bitner, 2000). Technological innovation has led to creation of new services and provided opportunities to deliver existing services in more enhanced ways (Pateli & Giaglis, 2005). The assimilation approach to service innovation emphasizes the role of technology in service innovation. The second issue is coordination along the whole NSD process. The NPD process can be broadly divided into two stages: fuzzy front-end and execution-oriented back-end (Khurana & Rosenthal, 1997; Moenaert, DeMeyer, Souder, & Deschoolmeester, 1995). In the NSD context, the front-end stage includes idea generation and concept development/evaluation while the back-end stage consists of activities involved in actually implementing the chosen concept. However, it has been pointed out that there is a lack of linkage between the back-end and the front-end (Menor, Tatikonda, & Sampson, 2002). To coordinate the two stages, implementation and operational issues in the back-end stage should be mirrored in concept selection. The AHP model has been developed on the basis of the relevant literature and the above arguments as shown in Fig. 3. The objective of the model is to evaluate the NSCs, located at the bottom of model, as alternatives. The model consists of five criteria: strategy, finance, market, technology, and implementation. Each criterion includes three to five of sub-criteria, as described in Table 1.
1 indicates the equal importance between two elements and 9 represents the extreme importance of one element compared with the other one. However, the preferences to NSCs in terms of subcriteria are not obtained from pairwise comparisons, but from individual ratings due to increase of complexity. When there are n alternatives and m sub-criteria, the number of paired comparisons required is mn(n 1)/2. If ten concepts are evaluated together with the sub-criteria employed, hundreds of pairwise comparisons are required, which is too time-consuming. For this reason, the preferences to NSCs are evaluated independently with a scale of 1–9. For the negative items whose value is preferred to be low, the reverse value is assigned. 3.3. Derivation of group judgments The AIJ method is employed to derive group priorities since the participants involved in the evaluation of NSCs share a common objective. In the AIJ method, four basic ways have been proposed to aggregate pairwise comparison matrices at the individual level: consensus, vote or compromise, geometric mean of individual judgments, and separate models (Dyer & Forman, 1992). Among others, the geometric mean is adopted to synthesize individual judgments since it preserves the reciprocal property of pairwise comparison matrices without violation of the Pareto principle (Forman & Peniwati, 1998). The pairwise comparisons among criteria and sub-criteria and the preferences to NSCs at individual level are transformed into rough numbers at the group level by using the following equations:
Lower approximation : PXðGMij Þ ¼
fY 2 U ij jRðYÞ 6 GM ij Þ
ð9Þ
d¼1
Upper approximation : PXðGMij Þ ¼
D [
fY 2 U ij jRðYÞ P GMij Þ ð10Þ
d¼1
Lower limit : LimðGMij ÞÞ ¼
Mij Y
!1=Mij xij
ð11Þ
m¼1
Upper limit : LimðGMij Þ ¼
3.2. Pairwise comparisons and individual ratings
D [
Nij Y
!1=Nij yij
ð12Þ
n¼1
The evaluation of NSCs starts with the proposed AHP model. Pairwise comparisons are carried out among criteria with respect to the importance to the goal and among sub-criteria with respect the importance to the criteria using a scale of 1–9, where a score of
Rough number : RNðGMij Þ ¼ ðLimðGMij Þ; LimðGMij ÞÞ
ð13Þ
In the equation, D and GMij refer to the number of domain experts and geometric mean of relative importance of ith element compared pairwisely with jth element with respect to the goal or
Fig. 3. AHP model for the evaluation of NSCs.
3408
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
Table 1 Concept evaluation criteria in NSD.
a
Criteria
Subcriteria
Description
Strategy
FST IMA CAN()a LOY
The The The The
Finance
PRO ICO() ORE()
The expected profits of a NSC The expected cost to fully implement a NSC The expected operation and maintenance expenses of a NSC
Market
EMS PGR MCO() EIS
The The The The
Technology
TAV OPP LCT
The degree to which a company has required technologies, if not, technologies are available and affordable via buying and licensing The degree to which the corresponding technology provides more innovation opportunities to the company The level of capability of a company to develop the required technology
Implementation
EXP RIS() LPR() RAV TLM()
The The The The The
degree degree degree degree
to to to to
which which which which
a a a a
NSC NSC NSC NSC
is fit with company’s strategy contributes to positive company image cannibalizes existing services provided by the company enhances customer relationship and loyalty
expected market size of a NSC expected potential growth of the market of a NSC degree to which the target market is competitive degree of easiness to imitate a NSC
degree of experience of company in developing and providing relevant services size of risk in implementing a NSC degree to which a NSC causes a legal problem degree to which the required materials and facilities for a NSC are available amount of time to launch a NSC to market
(): negative item.
corresponding criterion. xij and yij are the elements of lower and upper approximation for GMij; Mij and Nij denote the number of elements in lower and upper approximation. An aggregated pairwise comparison matrix is developed as given below:
0
1 B . e Aggregated pairwise comparison matrix : P ¼ B @ .. pc1
1 . . . p1c .. .. C C . . A 1 ð14Þ
where pij is the rough number that is composed of lower (Lim(GMij)) and upper limit (Lim(GMij)); c is the number of elements to be pairwisely compared. To derive the group priorities from an aggregated pairwise comparison matrix, the degree of satisfaction for the judgments (l) needs to be estimated. The larger value of this index denotes the higher degree of satisfaction for judgments. Among the various indexes, a linear convex combination method is employed, as depicted in the following equation (Ayag˘ & Özdemir, 2009).
Hurwicz principle : aðLimðresultÞÞ þ ð1 aÞðLimðresultÞÞ where 0 6 a 6 1
e The degree of satisfaction : P ¼ lLimðGMij Þ þ ð1 lÞLimðGMij Þ;
ria, and aggregated preference to NSCs through using Eqs. (6)–(8). However, the scores are rather rough numbers that contain the minimum and maximum values, which leads to the difficulties in decision making in concept selection. For this reason, we provide ranking rules of NSCs to cope with such a problem. A number of methods have been introduced and employed for the situation of decision making under uncertainty. Of previous methods, the three most widely used techniques are adopted with respect to the risk propensity of decision makers: maximin, maximax, and Hurwicz principle. Firstly, the maximin principle is based on the pessimistic view that the worst possible outcome will occur. The NSC that maximizes the minimum value is thus preferred. Secondly, the NSCs maximizing the maximum value are favored in the maximax principle. However, these two approaches are limited to only considering the extremely optimistic or pessimistic situations. Finally, the Hurwicz principle allows the selection of NSCs based on the risk propensity of decision makers with the index of optimism (a) ranging from 0 to 1, as given by:
where 0 6 l 6 1
ð15Þ
Finally, the eigenvector is calculated by fixing the index l. The preference to NSCs at the group level is also derived by using Eqs. (9)–(13). In the equation, GMij refers to the geometric mean of individual assessments for NSC i with respect to the sub-criterion j. The aggregated preference to NSCs for the group is shown as below:
0
r 11 B . e B Aggregated preference to NSCs : R ¼ @ .. rm1
1 r1n .. C C . A r mn ... .. .
ð17Þ
ð16Þ
where rij is the rough number that is composed of lower (Lim(GMij)) and upper limit (Lim(GMij)); m and n are the number of NSCs and sub-criteria, respectively. 3.4. Prioritization of new service concepts The final evaluation scores are calculated by multiplying the group priority vector of criteria, group priority vectors of sub-crite-
4. Case study In this section, a comparative case study of the video game service is presented to illustrate the proposed approach. The video game industry has seen market shifting rapidly and a seemingly unlimited proliferation of technologies, resulting in life cycles becoming shorter. In this situation, the selection of new game serTable 2 Summary of video game service concepts. NSC Concept Concept Concept Concept Concept Concept Concept Concept Concept Concept
Description 1 2 3 4 5 6 7 8 9 10
Mobile-based two dimensional romance-simulation game Console-based three dimensional management-RPG game Console-based three dimensional fantasy-adventure game Mobile-based two dimensional SF-PVP game PC-based three dimensional horror-adventure game PC-based three dimensional SF-shooting game (person) Mobile-based two dimensional education-puzzle game Mobile-based two dimensional war-PVP game Mobile-based three dimensional fantasy-adventure game PC-based three dimensional romance-simulation game
3409
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412 Table 3 Pairwise comparison matrix among criteria at the individual level.
Strategy Finance Market Technology Implementation
Strategy
Finance
Market
Technology
Implementation
1.00 0.14 0.20 0.33 0.33
7.00 1.00 7.00 0.20 5.00
5.00 0.14 1.00 0.33 1.00
3.00 5.00 3.00 1.00 3.00
3.00 0.20 1.00 0.33 1.00
Table 4 Pairwise comparison matrix among sub-criteria with respect to the STR at the individual level.
FST IMA CAN LOY
FST
IMA
CAN
LOY
1.00 2.00 0.33 0.50
0.50 1.00 0.20 0.25
3.00 5.00 1.00 2.00
2.00 4.00 0.50 1.00
Finally, the ratings of game service concepts with respect to each sub-criterion were obtained as depicted in Table 5. 4.2. Derivation of group judgments
vice concepts is becoming more important, and at the same time, a systematic approach is required to aid decision making in concept selection. The case study has been conducted together with the Korean firm A. The detailed report of firm is omitted due to the privacy regulation. A total of 10 game service concepts have been generated from our previous work, as described in Table 2 (Lee, Song, & Park, 2009). An expert group was organized from a project team of A for the evaluation of service concepts. The expert group consisted of nine individuals, whose expertise is strategic planning, system design, graphic design, and implementation. The work experience of expert group ranged from five to ten years, and average work experience was seven years and five months. The generation of those concepts will not be included in this paper since this study is limited to the concept evaluation in NSD. We will only illustrate how the proposed approach can be applied to the given concepts.
4.1. Pairwise comparisons and individual ratings Firstly, pairwise comparisons were made among criteria with respect to contribution to the goal. The pairwise comparison matrix for the expert 1 is exemplified in Table 3. Secondly, pairwise comparisons were made among sub-criteria with respect to contribution to the corresponding criteria. The pairwise comparison matrix with respect to the strategy for the expert 1 is shown in Table 4.
The pairwise comparison matrix among criteria and sub-criteria were aggregated, and the group priority vectors were derived by substituting l = 0.5, as shown in Tables 6 and 7. The aggregated preference to NSCs with respect to the sub-criteria was derived as shown in Table 8. 4.3. Prioritization of new service concepts The evaluation scores of game service concepts were obtained by multiplying the group priority vector of criteria, group priority vectors of sub-criteria, and aggregated preference of NSCs as shown in Table 9. The following summarizes the rankings in terms of risk propensity of decision makers: Maximin principle (pessimistic): NSC 1 > NSC 9 > NSC 4 > NSC 3 > NSC 10 > NSC 5 > NSC 8 > NSC 6 = NSC 7 > NSC 2. Maximax principle (optimistic): NSC 1 > NSC 3 > NSC 6 > NSC 4 > NSC 7 = NSC 9 > NSC 8 = NSC 5 > NSC 10 > NSC 2. Hurwicz principle (moderate with a = 0.5): NSC 1 > NSC 3 > NSC 4 > NSC 9 > NSC 6 > NSC 5 > NSC 8 > NSC 10 > NSC 7 > NSC 2. The results indicate that NSC 1 has a maximum priority, on the other hand, NSC 2 was found to be less attractive. The rankings of some concepts such as NSC 3, NSC 4, and NSC 9 are dependent on the risk propensity of decision makers. 4.4. Comparison with fuzzy group AHP The evaluation of video game service concepts using symmetrical triangular fuzzy number has also been carried out for compar-
Table 5 Example of preferences to video game service concepts at the individual level.
FST IMA CAN LOY PRO ICO ORE EMS PGR MCO EIS TAV OPP LCT EXP RIS LPR RAV TLM
NSC 1
NSC 2
NSC 3
NSC 4
NSC 5
NSC 6
NSC 7
NSC 8
NSC 9
NSC 10
7 9 3 7 5 3 9 9 7 5 5 7 3 9 3 9 7 1 7
1 1 9 3 3 3 7 5 1 7 5 5 9 5 1 5 3 5 3
7 5 1 1 3 7 7 7 9 5 1 3 3 7 5 5 1 7 9
3 5 9 5 7 1 5 1 1 3 1 1 5 7 1 5 9 9 7
1 5 1 5 3 7 1 7 5 1 7 1 7 7 5 7 7 1 7
5 7 1 5 9 3 3 3 7 3 3 7 3 1 9 9 3 3 3
9 1 1 5 7 9 5 1 1 7 3 3 3 5 9 7 7 5 1
7 7 9 3 3 7 7 1 3 7 9 9 9 7 1 5 3 7 1
5 3 1 7 9 5 7 5 3 7 3 7 5 5 7 7 5 7 5
7 5 3 1 7 3 5 5 9 5 1 3 7 3 7 9 1 3 3
3410
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
Table 6 Group priority vector of criteria.
Strategy Finance Market Technology Implementation
Strategy
Finance
Market
Technology
Implementation
Priority
(1.00, 1.00) (0.27, 2.00) (0.20, 3.16) (0.33, 3.46) (0.29, 4.00)
(0.50, 3.74) (1.00, 1.00) (3.46, 7.00) (0.77, 5.00) (0.41, 5.00)
(0.32, 5.00) (0.14, 0.29) (1.00, 1.00) (0.24, 0.41) (0.24, 1.00)
(0.29, 3.00) (0.20, 3.87) (2.45, 6.00) (1.00, 1.00) (0.32, 3.00)
(0.25, 3.46) (0.20, 2.45) (1.00, 4.24) (0.33, 3.16) (1.00, 1.00)
0.24 0.12 0.30 0.16 0.17
Table 7 Group priority vectors of sub-criteria. Criteria
Strategy (=0.24)
Sub-criteria
FST
IMA
CAN
LOY
Finance (=0.12) PRO
ICO
ORE
Market (=0.30) EMS
PGR
MCO
EIS
Technology (=0.16) TAV
OPP
LCT
Implementation (=0.17) EXP
RIS
LPR
RAV
TLM
Priority
0.31
0.46
0.13
0.10
0.66
0.20
0.14
0.34
0.30
0.23
0.13
0.35
0.35
0.30
0.20
0.19
0.04
0.12
0.44
Table 8 Aggregated preferences to video game service concepts.
FST IMA CAN LOY PRO ICO ORE EMS PGR MCO EIS TAV OPP LCT EXP RIS LPR RAV TLM
NSC 1
NSC 2
NSC 3
NSC 4
NSC 5
NSC 6
NSC 7
NSC 8
NSC 9
NSC 10
(5.00, 7.00) (5.00, 9.00) (3.00, 5.00) (1.00, 7.94) (5.00, 7.94) (1.00, 4.58) (7.00, 9.00) (5.00, 9.00) (1.00, 4.58) (5.00, 7.94) (3.87, 9.00) (3.00, 5.92) (3.87, 9.00) (9.00, 9.00) (3.00, 5.00) (3.00, 9.00) (5.00, 7.94) (1.00, 3.00) (7.00, 9.00)
(1.00, 9.00) (1.00, 7.00) (3.00, 9.00) (1.00, 3.00) (3.00, 5.92) (3.00, 5.00) (5.00, 7.00) (1.00, 5.92) (1.00, 3.00) (1.00, 4.58) (1.00, 5.92) (5.00, 7.00) (3.00, 7.94) (1.00, 3.87) (1.00, 5.00) (3.00, 5.00) (1.00, 4.58) (1.00, 5.00) (3.00, 3.00)
(7.00, 9.00) (5.00, 7.00) (1.00, 9.00) (1.00, 7.00) (3.00, 7.00) (1.00, 4.58) (3.00, 7.00) (1.00, 7.00) (3.87, 9.00) (1.00, 3.87) (1.00, 5.92) (3.00, 5.92) (3.00, 7.00) (3.00, 7.94) (5.00, 9.00) (1.00, 5.92) (1.00, 7.00) (1.00, 5.92) (3.00, 9.00)
(3.00, 5.00) (5.00, 7.00) (3.87, 9.00) (3.87, 9.00) (7.00, 9.00) (1.00, 5.00) (5.00, 5.00) (1.00, 5.92) (1.00, 5.00) (3.00, 7.94) (1.00, 5.92) (1.00, 7.00) (5.00, 7.00) (1.00, 7.94) (1.00, 6.71) (1.00, 3.87) (5.00, 7.94) (1.00, 6.71) (7.00, 9.00)
(1.00, 9.00) (3.00, 5.92) (1.00, 7.94) (5.00, 9.00) (3.00, 7.00) (3.00, 5.92) (1.00, 3.00) (5.00, 7.00) (1.00, 3.87) (1.00, 5.00) (1.00, 7.94) (1.00, 5.00) (5.00, 7.94) (1.00, 5.92) (3.00, 5.00) (7.00, 9.00) (3.00, 5.92) (1.00, 5.00) (1.00, 5.92)
(1.00, 5.00) (1.00, 7.00) (1.00, 7.00) (5.00, 9.00) (1.00, 9.00) (1.00, 3.87) (1.00, 3.00) (3.00, 5.92) (1.00, 7.94) (3.00, 5.92) (3.00, 9.00) (5.00, 7.00) (3.00, 9.00) (1.00, 7.00) (3.00, 9.00) (3.00, 9.00) (3.00, 7.00) (1.00, 3.87) (3.00, 5.92)
(3.87, 9.00) (1.00, 5.00) (1.00, 5.00) (3.87, 9.00) (1.00, 5.92) (3.00, 9.00) (3.87, 9.00) (1.00, 3.00) (1.00, 5.92) (3.00, 7.00) (3.87, 9.00) (3.00, 9.00) (3.00, 3.00) (5.00, 7.00) (1.00, 5.20) (1.00, 5.92) (7.00, 9.00) (5.00, 5.00) (1.00, 9.00)
(5.00, 7.00) (1.00, 5.92) (5.00, 7.94) (3.00, 9.00) (1.00, 3.87) (7.00, 9.00) (5.00, 7.00) (1.00, 4.58) (3.00, 1.00) (1.00, 7.00) (3.00, 7.94) (5.00, 7.94) (1.00, 9.00) (1.00, 7.94) (1.00, 9.00) (5.00, 7.94) (3.00, 7.00) (5.00, 7.00) (1.00, 7.00)
(5.00, 7.94) (3.00, 5.00) (1.00, 5.92) (7.00, 9.00) (7.00, 9.00) (5.00, 9.00) (3.00, 7.00) (3.00, 5.00) (1.00, 3.00) (1.00, 7.00) (3.00, 5.00) (7.00, 7.00) (3.87, 9.00) (1.00, 5.92) (7.00, 7.00) (3.00, 7.00) (5.00, 7.00) (3.00, 7.00) (3.00, 5.92)
(3.00, 7.00) (5.00, 5.00) (3.00, 7.00) (1.00, 4.58) (3.00, 7.94) (3.00, 3.00) (3.00, 5.92) (1.00, 5.92) (1.00, 9.00) (5.00, 5.00) (1.00, 3.00) (1.00, 3.87) (3.00, 5.92) (1.00, 3.00) (3.00, 7.00) (3.00, 9.00) (1.00, 7.00) (3.00, 7.94) (3.00, 7.00)
Table 9 Evaluation scores and rankings of game service concepts. Service concepts
Evaluation scores
Pessimistic
Optimistic
Moderate
NSC NSC NSC NSC NSC NSC NSC NSC NSC NSC
(4.38, 7.68) (1.93, 5.72) (2.99, 7.31) (3.22, 6.97) (2.47, 6.42) (2.23, 7.10) (2.23, 6.48) (2.39, 6.42) (3.64, 6.48) (2.63, 6.15)
4.38 1.93 2.99 3.22 2.47 2.23 2.23 2.39 3.64 2.63
7.68 5.72 7.31 6.97 6.42 7.10 6.48 6.42 6.48 6.15
6.03 3.82 5.15 5.10 4.44 4.66 4.36 4.40 5.06 4.39
1 2 3 4 5 6 7 8 9 10
(1) (10) (4) (3) (6) (8) (8) (7) (2) (5)
(1) (10) (2) (4) (7) (3) (5) (7) (5) (9)
(1) (10) (2) (3) (6) (5) (9) (7) (4) (8)
ative analysis. To compare the results, the evaluation scores are shown in Fig. 4. We can summarize the similarities and differences between rough group AHP and fuzzy group AHP approach as follows. The Spearman correlation coefficient is 0.806 which is statistically significant at the level of 0.01. This supports the fact that the rough group AHP and the fuzzy group AHP measure the same thing. However, there are differences in the ranks and intervals. The reasons are twofold. First, the same level of importance ratings is denoted by fuzzy group AHP for different judgments, but different by rough numbers. For the ratings of three experts, ‘1’, ‘3’, ‘9’ and ‘3’, ‘3’, ‘3’ are transformed into a same
fuzzy number (2, 4), but transformed into different rough numbers (1.73, 5.20) and (3, 3), respectively. Second, regarding to the first issue, the evaluation scores using fuzzy numbers have the same intervals of boundary region because of the pre-determined fuzzy membership function. This shows that the fuzzy group AHP cannot consider the individual’s subjective perceptions; it may lead to unrealistic and distorted evaluation results. By contrast, in the rough group AHP, the different level of importance ratings and different intervals of boundary region are gained across the individual judgments. Put those things together, the proposed approach is advantageous over existing fuzzy group AHP in terms of manipulation of subjective perceptions.
5. Conclusions This study proposed a rough group AHP approach to evaluation of NSCs by focusing on feasibility analysis at the concept development stage. We suggested the use of AHP and RST, and developed four consecutive stages: construction of AHP model, pairwise comparisons and individual ratings, derivation of group judgments, and prioritization of NSCs. By integrating the strength of RST in handling subjectivity and the merit of AHP, the suggested approach measures the feasibility of NSCs in terms of strategy, finance, market, technology, and implementation, taking the subjectivity involved in the concept development. A
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
3411
Acknowledgments This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2009-0085757).
References
Fig. 4. Results of comparisons between rough group AHP and fuzzy group AHP.
case study of the video game service has been presented to illustrate the suggested approach. The contribution and potential utilities of this research are twofold. Firstly, this study theoretically contributes to the NSD research, by proposing a systematic approach to evaluation of NSCs. Most of the previous studies were limited to identifying factors that affects the selection of NSCs. Although a number of factors have been found to be influential, a lacuna still remains in the literature as to how to make a strategic decision based on influential factors with the help of a systematic and quantitative approach. The suggested rough group AHP measures the feasibility of NSCS by taking the subjectivity involved in the evaluation of NSCs into account. The core advantages of the suggested approach lie in its systematic process and analytic power. It is expected that our method can promote rapid and efficient consensus building about the most promising NSCs, and can be utilized especially when the conventional methods, such as conjoint analysis and statistics, cannot be used due to the problem of data sample size. Second, regarding the methodology, this research is explanatory in that a rough group AHP approach is first proposed to overcome the limitations of fuzzy group AHP. The subjectivity is measured from distribution of individual judgments without external information and additional adjustments. The focus of this study is not limited to how to evaluate and select NSCs. Rather, this research emphasizes the details of rough group AHP and its strength for MCDM problems which is characterized by subjectivity. The proposed approach can be effectively employed for many real world problems such as new business opportunity analysis and new technology creation, which can be characterized by newly generated alternatives and vaguely defined criteria. Despite all the possibilities offered by the proposed approach, this study is subject to several limitations that stand in the way of our future research plans. First, the concept evaluation criteria employed are by no means fixed and exhaustive. Important criteria may differ across the context of NSD. Therefore, the proposed approach should be modified by adding or removing some criteria from the model upon the context of NSD. Secondly, the interdependencies among evaluation criteria are not considered, which originate from the inherent characteristics of AHP. To overcome the limitation, the proposed approach can be extended by integrating RST together with group analytic network process (ANP). Finally, the validity of our method necessitates more testing work, which is indispensable for gaining external validity. These can be fruitful areas for future research.
Alam, I., & Perry, C. (2002). A customer-oriented new service development process. Journal of Services Marketing, 16(2), 515–534. Ayag˘, Z., & Özdemir, R. G. (2009). A hybrid approach to concept selection through fuzzy analytic network process. Computers and Industrial Engineering, 56(1), 368–379. Bullinger, H. J., Fahnrich, K. P., & Meiren, T. (2003). Service engineering-methodical development of new service products. International Journal of Production Economics, 85(3), 275–287. Davison, H., Watkins, T., & Wright, M. (1989). Developing new personal financial products – Some evidence of the role of market research. International Journal of Bank Marketing, 7(1), 8–15. Dyer, R. F., & Forman, E. H. (1992). Group decision support with the analytic hierarchy process. Decision Support Systems, 8(2), 99–124. Easingwood, C. J. (1986). New product development for service companies. Journal of Product Innovation Management, 3(4), 264–275. Edgett, S. (1993). Developing new financial services within UK building societies. International Journal of Bank Marketing, 11(3), 35–43. Edvardsson, B., Haglund, L., & Mattsson, J. (1995). Analysis, planning, improvisation and control in the development of new services. Journal of Service Management, 6(2), 24–35. Forman, E., & Peniwati, K. (1998). Aggregating individual judgments and priorities with the analytic hierarchy process. European Journal of Operational Research, 108(1), 165–169. Goldstein, S. M., Johnston, R., Duffy, J., & Rao, J. (2002). The service concepts: The missing link in service design research? Journal of Operations Management, 20(2), 121–134. Greco, S., Matarazzo, B., & Slowinski, R. (2001). Rough set theory for multicriteria decision analysis. European Journal of Operational Research, 129(1), 1–47. IBM Research (2004). Services science: A new academic discipline. . Johne, A. (1993). Insurance product development: Managing the changes. International Journal of Bank Marketing, 11(3), 5–14. Johne, A., & Storey, C. (1998). New service development: A review of the literature and annotated bibliography. European Journal of Marketing, 32(3/4), 184–251. Kelly, D., & Storey, C. (2000). New service development: Initiation strategies. Journal of Service Management, 11(1), 45–62. Khurana, A., & Rosenthal, S. R. (1997). Integrating the fuzzy front end of new product development. Sloan Management Review, 38(2), 103–120. King, A. M., & Sivaloganathan, S. (1999). Development of a methodology for concept selection in flexible design strategies. Journal of Engineering Design, 10(4), 329–349. Krishnan, K., & Ulrich, K. T. (2001). Product development decisions: A review of the literature. Management Science, 47(1), 1–21. Kuo, M., Tzeng, G., & Hwang, W. (2007). Group decision-making based on concepts of ideal and anti-ideal points in a fuzzy environment. Mathematical and Computer Modelling, 45(3/4), 573–581. Lai, V. S., Wong, B. K., & Cheung, W. (2002). Group decision making in a multiple criteria environment: A case using the AHP in software selection. European Journal of Operational Research, 137(1), 134–144. Lee, C., Song, B., & Park, Y. (2009). Generation of new service concepts: A morphology analysis and genetic algorithm approach. Expert Systems with Applications, 36(10), 12454–12460. Lee, H., Kim, C., & Park, Y. (2010). Evaluation and management of new service concepts: An ANP-based portfolio approach. Computers and Industrial Engineering, 58(4), 535–543. Lin, C., Lee, C., & Chen, W. (2009). Using fuzzy analytic hierarchy process to evaluate service performance of a travel intermediary. The Service Industries Journal, 29(3), 281–296. Liu, D. R., & Shih, Y. Y. (2005). Integrating AHP and data mining for product recommendation based on customer lifetime value. Information and Management, 42(3), 387–400. Lundkvist, A., & Yakhlef, A. (2004). Customer involvement in new service development: A conversational approach. Managing Service Quality, 14(2/3), 249–257. Martin, C. R., & Horne, D. A. (1993). Services innovation: Successful versus unsuccessful firms. Journal of Service Management, 4(1), 48–64. Mathieu, V. (2001). Service strategies within the manufacturing sector: Benefits, costs, and partnerships. Journal of Service Management, 12(5), 451–475. Meiren, T., & Burger, T. (2010). Testing of service concepts. The Service Industries Journal, 30(4), 621–632. Menor, L., Tatikonda, M. V., & Sampson, S. E. (2002). New service development: Area for exploitation and exploration. Journal of Operations Management, 20(2), 135–157. Meuter, M. L., Ostrom, A. L., Roundtree, R. I., & Bitner, M. J. (2000). Self-service technologies: understanding customer satisfaction with technology-based service encounters. Journal of Marketing, 64(3), 50–64.
3412
C. Lee et al. / Expert Systems with Applications 39 (2012) 3404–3412
Meyer, M. H., & DeTore, A. (2001). Creating a platform-based approach for developing new services. The Journal of Product Innovation Management, 18(3), 188–204. Moenaert, R. K., DeMeyer, A., Souder, W. E., & Deschoolmeester, D. (1995). R&D/ marketing communication during the fuzzy front-end. IEEE Transactions on Engineering Management, 42(3), 243–258. Mohammed-Salleh, A., & Easingwood, C. (1993). Why European financial institutions do not test-market new consumer products. International Journal of Bank Marketing, 11(3), 23–28. Oliva, R., & Kallenberg, R. (2003). Managing the transition from products to services. Journal of Service Management, 14(2), 160–172. Pateli, A. G., & Giaglis, G. M. (2005). Technology innovation-induced business model change: A contingency approach. Journal of Organizational Change Management, 18(2), 167–183. Pawlak, Z. (1982). Rough sets. International Journal of Computing and Information Sciences, 11(5), 341–356. Saaty, T. L. (1980). The analytic hierarchy process. New York: McGraw-Hill. Saridakis, K. M., & Dentsoras, A. J. (2008). Soft computing in engineering design – A review. Advanced Engineering Informatics, 22(2), 202–221.
Shao, X., Chu, X., Qiu, H., Gao, L., & Yan, J. (2009). An expert system using rough sets theory for aided conceptual design of ship’s engine room automation. Expert Systems with Applications, 36(2), 3223–3233. Tseng, T., & Huang, C. (2008). Design support systems: A case study of modular design of the set-top box from design knowledge externalization perspective. Decision Support Systems, 44(4), 909–924. Ulrich, K. T., & Eppinger, S. D. (2003). Product design and development. New York: McGraw-Hill. Walczak, B., & Massart, D. (1999). Rough sets theory. Chemometics and Intelligent Laboratory Systems, 47(1), 1–16. Zhai, L., Khoo, L., & Fok, S. (2002). Feature extraction using rough set theory and genetic algorithms – An application for the simplification of product quality evaluation. Computer and Industrial Engineering, 43(4), 661–676. Zhai, L., Khoo, L., & Zhong, Z. (2008). A rough set enhanced fuzzy approach to quality function deployment. International Journal of Advanced Manufacturing Technology, 37(5/6), 613–624. Zhai, L., Khoo, L., & Zhong, Z. (2009). A rough set based QFD approach to the management of imprecise design information in product development. Advanced Engineering Informatics, 23(2), 222–228.
