A novel approach to incorporate customer preference and perception into product configuration: A case study on smart pads

(1)

A novel approach to incorporate customer preference and perception into product

con

ﬁguration: A case study on smart pads

Chih-Hsuan Wang

⁎

, One-Zen Hsueh

Department of Industrial Engineering & Management, National Chiao Tung University, Hsinchu, Taiwan

a b s t r a c t

a r t i c l e i n f o

Article history: Received 3 August 2012

Received in revised form 4 December 2012 Accepted 4 January 2013

Available online 4 February 2013 Keywords: AHP DEMATEL Kano model Preference segmentation Concept evaluation

This paper proposes a hybrid framework combining AHP (analytical hierarchy process), KM (Kano model), with DEMATEL (decision making trial and evaluation laboratory) to incorporate customer preference and perception into the process of product development. Initially, AHP is applied to respondents to form a basis of market segmentation. Thereafter, with respect to identiﬁed segments, AHP and KM are employed to extract customer preference for design attributes (DAs) and customer perception of marketing require-ments (MRs), respectively. Finally, by means of DEMATEL, the causal relationships between MRs and DAs are systematically recognized to uncover new ideas of next-generation products.

1. Introduction

In a traditional“supply-push” driven era, manufacturing compa-nies merely considered offering products with high quality, low cost, functioning performance and courteous after-sales service to satisfy market majorities[12]. Nowadays, owing to the concept of mass customization, customer satisfaction has become a growing concern to dominate the competing paradigm[2,6]. In order to sur-vive in today's “demand-pull” environments, modern companies need to conceive attractive products/services to acquire different market segments and even for“customized” individuals[18]. Never-theless, customers are too diverse, too heterogeneous, and too widely scattered in their preferences, perceptions, shopping behaviors, life-styles, and their psychological demographics[27]. Thus, irrespective of the fact that high product variety does signiﬁcantly stimulate prod-uct sales, most manufacturing companies are inevitably facing the trade-offs between increasing product variety and controlling manufacturing complexity[14,29]. In practice, to respond to dynam-ically changing customer desire, awareness of customer preference/ perception is becoming much more imperative than ever before dur-ing the process of product development[21].

To tackle the aforementioned issues, one of the most famous schemes originated from the discipline of strategic marketing is a so-called“STP” approach (segmentation-targeting-positioning), which has been widely adopted among academic researchers and industrial practitioners[16]. Specifically, the step of “segmentation” allows mar-keters to divide the entire market into ad-hoc segments in which cus-tomers demonstrate similar patterns within a group but behave heterogeneously between groups. Secondly, the step of “targeting” helps afirm assess each segment's attractiveness, profitability, and then be able to select one or more segments to run their business. Final-ly, the step of“positioning” emphasizes differentiating a firm from com-petitors through offering attractive alternatives.

Apparently, market segmentation is the most critical step to achieve the success of the entire process of STP. According to Wang [28], there are several commonly used variables for market segmen-tation, including demographic variables (i.e. age, gender, race, and salary), psychographic variables (i.e. social class, lifestyle and per-sonality) and behavioral variables (i.e. user preference, usage pat-tern, and loyalty status). Theoretically, market segmentation assumes that groups of customers with similar profiles or patterns are likely to demonstrate a homogeneous response to specific prod-uct promotion and marketing programs[9]. In this study, the pur-pose of market segmentation is to form a launch pad for generating and assessing potential product alternatives, particularly with re-spect to those identified niche segments. To effectively divide the whole market, customers' perceived importance degrees of design attributes are treated as input variables to carry out market segmentation.

⁎ Corresponding author at: 1001 University Road, Hsinchu 30003, Taiwan. Tel.: +886 3 5712121 57310; fax: +886 3 5722392.

http://dx.doi.org/10.1016/j.csi.2013.01.002

Contents lists available atSciVerse ScienceDirect

Computer Standards & Interfaces

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c s i

(2)

Furthermore, to help an enterprise better understand and capture dynamically changing customer desire, this paper attempts to incor-porate customer preference as well as customer perception into the process of product development, especially to diminish the gap be-tween customers' requirements and manufacturers' alternatives. Consequently, a market-oriented framework which integrates AHP, KM with DEMATEL is proposed and several key issues are addressed below:

● Learning which design attributes (DAs) are more representative to segment the entire market,

● Examining customer preference for multi-leveled DAs for generating concepts and assessing product alternatives in a customer-driven manner,

● Eliciting customer perception of marketing requirements (MRs) to form a launch pad for discovering new ideas of the next-generation products,

● Identifying the complicated interrelationships between DAs and MRs to help product managers better understand their inherent dynamics.

The rest of this paper is structured as follows.Section 2brieﬂy overviews classical techniques for eliciting customer preference and Section 3introduces the proposed framework. An industrial example regarding conﬁguring product varieties of smart pads is illustrated in Section 4. Finally, conclusions and future studies are drawn in Section 5.

2. Review of classical techniques to elicit customer preference In an era of mass customization, companies need to deliberately understand what customers want and need in order to avoid fatal mistakes before implementing their product strategies[9,12]. New product development (NPD), defined as a process of transforming an identified market opportunity into profitable product(s) for sale, usually consists of a sequence of steps in which an enterprise employs to conceive, design, and commercialize product alternatives[3]. As a matter of fact, NPD is an interdisciplinary activity involving market-ing, operation, manufacturmarket-ing, and requires sustainable commitment from the top level of management. Therefore, various disciplines in-cluding marketing research, consumer behavior, and concurrent engi-neering, attempt to contribute to different stages of NPD [18]. Currently, recent publications have witnessed emerging growth of the front-end issues such as customer relationship management and customer requirement management[14].

In fact, the capability of concept generation and concept evalua-tion for different segments has been recognized as one of the key de-terminants for many firms to survive in an extremely uncertain business environment[3,4,9,19]. Nevertheless, without incorporating customer preference or customer perception into the process of con-cept generation/evaluation, the objective of customer satisfaction is difficult to be fulfilled[2,6,21]. To the best of our knowledge, several techniques which are widely applied to various industries like quality function deployment (QFD), conjoint analysis (CA), and Kano model (KM), are shortly overviewed later.

2.1. Quality function deployment (QFD)

Quality function deployment[1]is a well-known scheme that pro-vides a structural framework to translate customers' voices into tan-gible product design. Typically, the conventional QFD consists of the following four phases: phase one translates marketing requirements into design attributes; phase two translates design attributes into part characteristics; phase three translates part characteristics into manufacturing operation, and phase four translates manufacturing operation into production requirements[17]. By considering the in-terdependences between MRs and DAs and the correlations among

themselves, QFD is capable to derive the priorities of DAs in terms of the weights of MRs[29,30]: Rij0¼ Xn k¼1 Rik γkj Xn k¼1 Xn k¼1 Rik γkj ; ð1Þ WtDAj¼ Xm i¼1 WtMRiRij0; ð2Þ

where WtMRiand WtDAjrepresent the weight of MRiand DAj, respec-tively. Here, m marketing requirements and n design attributes are assumed to characterize the QFD, Rij' is the normalized dependence between MRi and DAj, and λik and γkj denote the correlations among MRs and DAs, respectively.

Nevertheless, QFD has been criticized by insufﬁcient customer in-volvement (i.e. customer preference and customer satisfaction) when generating the weights of MRs or DAs. In addition, QFD is deﬁcient in generating/assessing product concepts, especially when a product is functionally decomposed into various design attributes associated with multi-levels. To enhance its applicability, several researchers suggest to combine the QFD with conjoint analysis (CA) or Kano model (KM)[8,21,24,26].

2.2. Conjoint analysis (CA)

Conjoint analysis[20]is one of the most popular techniques to measure diverse customer preference among multi-attributed prod-ucts or services. When a product is decomposed into independent at-tributes associated with their corresponding levels, a respondent's overall utility could be decomposed into his/her part-worth values [8,25]. For reference, a general mathematical form of CA can be modeled as follows[13]: Uk¼ β0þ Xm i¼1 Xn j¼1 uijk; ð3Þ

where Ukmeans alternative k's overall utility,β0denotes a regular-ized constant, uijkrepresents alternative k's part-worth utility corre-sponding to attribute i associated with level j, m is the number of attributes and n is the number of associated levels for attribute i. To derive the importance degree of various attributes, it is widely ac-cepted that an attribute having a wider range of part-worth values should have greater impact on the overall utility of a product.

For convenience, let's illustrate a simple example. Suppose that a smart pad is characterized by six attributes (A1–A6) associated with multi-levels (e.g. A1(3), A2(2), A3(3), A4(2), A5(2), and A6(2)), intu-itively, a maximal number of 144 (32_{* 2}4_{) combinations might be} possibly generated. To derive their part-worth utilities of six attri-butes, it is impossible to ask an evaluator to prioritize 144 alternatives at a time. Hopefully, by means of fractional factorial design, the entire process could be signiﬁcantly simpliﬁed and reduced to rank only 16 orthogonal alternatives. Obviously, CA treats a multi-attributed prod-uct on a single layer and thus it cannot process a functional hierarchy structure.

2.3. Kano model (KM)

The basic idea of KM[15]is using a nonlinear way to measure cus-tomers' asymmetric perceptions of two sides: positive delight when an attribute is present and negative disgust when an attribute is

(3)

absent, respectively. Referring toFig. 1, various Kano categories are brieﬂy explained as follows[21,24]:

● Must-be (M): the attribute which belongs to this category consists of the basic criteria of a product since customers are extremely dissatisfied if it is not fulfilled. However, its fulfillment cannot sig-nificantly increase satisfaction level since customers take them for granted,

● One-dimensional (O): the presence of an attribute will increase customer satisfaction level while its absence will proportionally decrease satisfaction level. This category enhances customer loyal-ty for companies,

● Attractive (A): an attribute classified in this category usually acts as a weapon to differentiate companies from their competitors since its fulfillment generates absolutely positive satisfaction while customers are not dissatisfied at all when it is unfulfilled, ● Reverse (R): an attribute falling into this category should be

re-moved from a product since its presence is harmful to customer satisfaction while its absence is beneﬁcial,

● Indifferent (I): an attributes falling into this category do not con-tribute much to customer satisfaction regardless whether they are present or absent in a product,

● Questionable (Q): this outcome indicates that either the question-naire was incorrectly described or an illogical response was sent by an evaluator.

Traditionally, the Kano categories are prioritized in an order of M_{≻O≻A, indicating that the “must-be” category should be conﬁgured} ﬁrst, followed by the “one-dimensional” category, and then the “at-tractive” category. Apparently, both indifferent and reverse categories should be excluded because they cannot enhance satisfaction level at all but also incur extra manufacturing cost. To track the research trend of Kano model, interested readers could refer to a state-of-art review[22].

Although numerous studies [8,19,21,24,26] have fused various techniques to tackle different problems, most of them cannot

efﬁciently decompose a functional product hierarchy as well as effec-tively facilitate customer involvement into the process of product de-velopment. After carefully review several classical techniques, an overall comparison among them is shown inTable 1to demonstrate their relative strengths and weaknesses. Despite CA is capable to gen-erate design concepts, it is not considered in our paper because of its deﬁciency of identifying the interdependences between MRs and DAs. In addition, it is found that respondents are often impatient to com-plete the CA questionnaires when requiring them to balance the trade-offs among design attributes. To concurrently address the aforementioned issues, the AHP is incorporated into our hybrid framework.

3. The proposed hybrid framework

As indicated byFig. 2, a hybrid framework which combines AHP, KM, with DEMATEL is presented and its details are operated as follows:

● Representative MRs and DAs for characterizing a smart pad are listed after surveying product speciﬁcations and consulting focus groups,

● AHP is initially utilized to extract customers' perceived impor-tance degrees of DAs to carry out market segmentation through the K-means clustering,

● With respect to those identiﬁed segments, AHP and KM are employed to extract customer preference for DAs and customer perception of MRs, respectively,

● Based on the results obtained in the previous steps, competitive product alternatives are generated and evaluated in a market-oriented manner,

● By virtue of DEMATEL, the causal relationships between MRs and DAs are systematically identiﬁed to uncover new ideas of the next-generation products.

3.1. Use of AHP to extract customers' preferences for DAs

AHP (analytic hierarchy process) was originally proposed by Saaty [23]to tackle the problem of scarce resource allocation for the mili-tary. It is a simple, intuitive, yet powerful methodology to determine the importance degrees of the evaluation criteria and the priorities of competitive alternatives[3,4]. Today, AHP has been successfully ap-plied to various domain problems[11,19]. Generally, the AHP com-prises the following steps:

● Constructing a hierarchy of the decision problem: followed by a top-down approach, the hierarchy is usually decomposed into multi-levels which consist of main criteria, associated sub-criteria and competing alternatives.

Fulfillment Reverse One-dimensional Indifferent Delight Disgust Unfulfillment Attractive Must-be

Fig. 1. Kano model for displaying customer perception.

Table 1

An overall comparison between QFD, CA, and KM.

QFD CA KM

Identifying relationships between MRs and DAs Yes No No Handling multi-leveled product attributes No Yes No Extracting subjective customer preference Limited Yes Limited Eliciting vague customer perception No No Yes Performing concept generation/evaluation Limited Yes Limited

Practical feasibility High Low High

Apply AHP to respondents to derive the importance weights of design attributes

Utilize K-means clustering to segment the entire market into ad-hoc niche segments

Use AHP to extract customer preference for design attributes

Use KM to elicit customer perception of marketing requirements

Use DEMATEL to identify the causal relationships between marketing requirements and design attributes

(4)

● Employing pairwise comparisons between evaluation criteria (alternatives): Saaty [20] recommended using a 9-point rating scale to express human preference among criteria, like equally, weakly, moderately, strongly, and extremely preferred with scales of 1, 3, 5, 7, and 9, respectively. Values of 2, 4, 6, and 8 are the inter-mediate values for the preference scales (seeTable 2).

● Computing the maximal eigenvalue (see Eq.(4)) and its associated eigenvector (see Eq.(5)) to derive their relative weights of n evalu-ation criteria:

A−λI

j j ¼ 0; ð4Þ

A−λI

ð ÞX ¼ 0; ð5Þ

where I denotes an identity matrix, A means a n × n pairwise com-parison matrix generated by n main criteria and its element aij rep-resents the preference degree of criterion i over criterion j. In particular, when the maximal eigenvalue of matrix A (_λmax) is extracted, the weights of n criteria (W) could be obtained via ﬁnd-ing its correspondﬁnd-ing eigenvector (AW=λmaxW).

● Checking the decision quality of using the AHP: it is related to ex-amine whether respondents demonstrated consistency during the process of pairwise comparisons. For example, the property of tran-sitivity implies that“if A is preferred to B, and B is preferred to C, then A should be preferred to C”. Hence, the consistency index (CI) and consistency ratio (CR) deﬁned as:

CI¼λmax−n

n₋₁ ; ð6Þ

CR¼CI_RI; ð7Þ

where CI that is more closer to zero indicates its greater consistency, and RI is a random index (seeTable 3) suggested by Saaty[23]. When the value of CR is less than the threshold of 0.1, the decision process might be considered to be highly consistent.

In this study, the relative weights of DAs of respondents are used to form a basis to carry out preference-based market segmentation. Meanwhile, AHP is also applied to the identiﬁed segments to extract their aggregated preference for the associated multi-levels of DAs. 3.2. Use Kano model (KM) to elicit customer perception of

marketing requirements

The Kano questionnaire[15], as shown inTable 4, provides a quan-titative approach to investigate asymmetric customers' perceptions: positive delight for functional fulfillment and negative disgust for dys-functional unfulfillment (see Fig. 1. again). Initially, a respondent needs to select one of the followingfive linguistic terms, such as “like”, “must-be”, “neutral”, “live-with”, and “dislike”, to reflect his/her perception of the above-mentioned two scenarios. As indicated by Table 5, 25 possible combinations of assessments are classified into one out of the six Kano categories, namely, “attractive” (A), “one-dimensional” (O), “must-be” (M), “indifferent” (I), “reverse” (R), and“questionable” (Q).

In additional to obtaining Kano categories for product attributes, it is difﬁcult to be equipped with quantitative assessments in prac-tical implementations[22]. Based on[5,24], positive delight dð iþÞ, negative disgust dð i−Þ and the importance weight of attribute i are slightly modiﬁed as follows:

diþ¼ Aiþ Oi−Ri Aiþ Oiþ Miþ Riþ Ii; ð8Þ di−¼ − Oiþ Mi−Ri Aiþ Oiþ Miþ Riþ Ii; ð9Þ wi¼ Gi ∑ i Gi; Gi¼ di þ_−d i−; ð10Þ

where Ai, Oi, Mi, Ri, and Iirepresent corresponding percentages of re-sponses among various Kano categories and the relative weight (wi) of attribute i can be obtained through normalizing its range (Gi) de-ﬁned by positive delight less negative disgust. Here, the KM is used to elicit customer perception of MRs and to derive their relative pri-orities for different segments. Thus, in conjunction with previously extracted customer preference for DAs, design concepts of smart pads could be systematically generated and assessed to suit cus-tomers' needs.

3.3. Use DEMATEL to identify the causal relationships between MRs and DAs

DEMATEL (decision making trial and evaluation laboratory), de-veloped by the science and human affairs program of the Battelle Me-morial Institute of Geneva Research Centre[10], is able to visualize the complex relationship among the interdependent factors. Through converting the causal relationship of the whole system into a struc-ture model, the DEMATEL could distinguish all factors into two dis-tinct groups: the dispatcher group and the receiver group. Its details are described as follows[29]:

● Generating the direct-relation matrix: based on a nine-point rat-ing scale (seeTable 2again), domain experts are invited to com-plete influence measures between factors. Suppose there are n factors and then, a n × n influence matrix A in which its element aijdenotes the impact of factor i on factor j, displays the mutual in-fluences between these two factors.

● Normalizing the direct-relation matrix: the normalized matrix B can be obtained through Eqs.(11)–(12), in which the diagonal elements of matrix B are set zeros.

B¼ k A ð11Þ

Table 2

A nine-point rating scale used in AHP and DEMATEL.

Scale Preference measure Inﬂuence measure

1 Equally preferred Slightly inﬂuenced

3 Weakly preferred Weakly inﬂuenced

5 Moderately preferred Moderately inﬂuenced

7 Strongly preferred Strongly inﬂuenced

9 Extremely preferred Extremely inﬂuenced 2, 4, 6, 8 Intermediate scale between two adjacent degrees

Table 3

Random index (RI) used by the AHP. Order of matrix (number of criteria)

n 2 3 4 5 6 7 8

RI 0 0.58 0.90 1.12 1.24 1.32 1.41

Table 4

A sample of Kano questionnaires. How do you feel

about this attribute?

I like it that way It must be that way I am neutral I can live with it I dislike it that way Attribute 1 Functional √ Dysfunctional √ Functional Dysfunctional Attribute n Functional √ Dysfunctional √

(5)

k¼ Min 1 max 1≤i≤n Xn j¼1 aij ; 1 max 1≤j≤n Xn i¼1 aij 0 B B B B @ 1 C C C C A: ð12Þ

● Generating the total-relation matrix: the total-relation matrix M can be derived via Eq.(13), where I denotes an identity matrix. M¼ B þ B2

þ B3

þ ⋯ ¼ B I−Bð Þ−1: ð13Þ

● Computing a causal diagram through distinguishing the trans-mitter group from the receiver group: Here, notice that Ti means factor i's total dispatched inﬂuence (e.g. the sum of rows in the total-relation matrix) while Rj means factor j's

total received impacts (the sum of columns in the

total-relation matrix). Ti¼ Xn j¼1 Mij; ð14Þ Rj¼ Xn i¼1 Mij: ð15Þ

By portraying the dataset comprising (T + R, T−R), a causal dia-gram is visualized, where the horizontal axis represents“T+R” and the vertical axis denotes “T−R”. Specifically, the “T+R” named “prominence” reveals how significant the factor is. On the other hand, the “T−R” named “influence” separates a factor into either the cause group or the effect group. In simple words, a factor fallen in the cause group is acting as a“dispatcher” since it tends to impact

on other factors. By contrast, a factor fallen in the effect group is act-ing as a_{“receiver” because it is affected by other factors.}

4. Empirical results and discussion

In this section, an industrial example was realized in a middle scale Taiwanese company which manufactures various types of consumer electronics, such as mobile phones, LCD monitors, and notebooks. Recently, in order to extend its product lines, this company planned to design varieties of smart pads to meet diverse customers' needs. According to its marketing survey, the boundary between smart pads, smart phones, and smart cameras is now becoming more and more blurred and this implies that these products might be functionally re-placeable to some extent[7]. To diminish the gap between customer ex-pectation and customer perception, the company's top management decides to carry out a cross-functional project to incorporate customer involvement into the process of product development. Prior to describ-ing its details of the whole project, six representative MRs and DAs as-sociated with multi-levels are highlighted by domain experts and listed inTable 6. For reference, a simpliﬁed questionnaire is abbreviated and illustrated inTable 7.

4.1. Segmenting the entire market based on customers' perceived importance degrees

Initially, 120 customers are examined to investigate their per-ceived importance degrees of DAs by using the AHP questionnaire (also seeTable 7). After completing the process of customer survey, their results are processed by the AHP and then passed to the K-means clustering for the purpose of market segmentation. Howev-er, the number of segments, or equivalently the value of K needs to be specified in advance. To determine an optimal number of segments, a metrics called F score is adopted in this study. In simple words, F score is defined by the ratio of “mean square error between groups” divided by“mean square error within groups”. To seek an optimal number of segments, F ratio needs to pass a significance test among all segmen-tation variables (like DAs in our example). Through a try-and-error process, K = 3 is optimally determined since all DAs have passed a Table 5

Evaluation summary for Kano classiﬁcation. Functional

presence

Dysfunctional absence

Like (L) Must-be (M) Neutral (N) Live-with (W) Dislike (D)

Like (L) Q A A A O

Must-be (M) R I I I M

Neutral (N) R I I I M

Live-with (W) R I I I M

Dislike (D) R R R R Q

*A = attractive, I = indifferent, M = must-be, O = one-dimensional, R = reverse, and Q = questionable.

Table 6

Symbols of MAs and DAs for characterizing a smart pad. MR (marketing

requirements)

DA (design attributes) Associated levels R1 User interface A1 CPU (type) a11— Atom

a12— Dual a13— Quad R2 System performance A2 ROM capacity (GB) a21— 8 GB a22— 4 GB R3 Response speed

(boot/networking)

A3 Operating system a31— Apple iOS a32— Google android a33— MS window R4 Multi-media

performance

A4 Screen size (inch) a41— 9–10 in. a42— 7–8 in. R5 Durability A5 Front/back camera

(mega pixels)

a51— 200 M/500 M a52— 130 M/300 M R6 Portability A6 Battery capacity (mAH) a61— 6000mAH

a62— 35000mAH

Table 7

Illustration of simpliﬁed questionnaires.

Schemes Corresponding questions Scales Respondents AHP ● How much degree is DAi

preferred to DAj? ● How much degree is level i

preferred to level j for a design attribute?

Numeric (9-point) Customers

KM ● How do you feel about MRjif it is fulﬁlled? ● How do you feel about

MRjif it is unfulﬁlled?

Linguistic (5-point) Customers

DEMATEL ● How much inﬂuence does DAiexert on MRj?

Numeric (9-point) Experts

Table 8

Average importance weights of DAs for different segments.

S1 (home) S2 (business) S3 (entertainment)

A1 0.217 0.377 0.241 A2 0.058 0.146 0.065 A3 0.325 0.279 0.059 A4 0.253 0.084 0.332 A5 0.049 0.038 0.117 A6 0.098 0.075 0.186 Count 45 36 39

(6)

statistical testing. As indicated byTable 8, three distinct segments are named as S1 (home), S2 (business), and S3 (entertainment), respec-tively. And their top three signiﬁcant DAs are particularly marked to display their differences between segments.

Speciﬁcally, the home segment (S1) presents an order of A3≻A4≻A1 while the patterns of A1≻A3≻A2 and A4≻A1≻A6 are demonstrated by the business segment (S2) and the entertainment segment (S3), respectively. Not surprisingly, attribute A1 (CPU) is concurrently critical to three segments. By contrast, attribute A3 (op-erating system) is signiﬁcant to both segments of S1 and S2 whereas attribute A4 (screen size) is important to both segments of S1 and S3. Apparently, customers in the business segment concern much more about DAs which might impact on“system performance” and similar explanations could be generalized to other segments.

4.2. Extracting customer preference and customer perception for identiﬁed segments

Based on three identiﬁed segments, AHP and KM are utilized to extract customer preference for associated levels of DAs and custom-er pcustom-erception of MRs, respectively. Aftcustom-er looking into the details of Table 9, it is observed that two DAs involving A3 (operating system) and A4 (screen size) are diversely scattered among segments. To con-clude, Android OS (a31) is favored by both home (S1) and entertain-ment (S3) segentertain-ments while the business segentertain-ment (S2) prefers iOS (a32). Similarly, both S1 and S3 desire large screen size (a41), but small screen size (a42) is favored by S2 for the consideration of por-tability. Again, with the consideration of six DAs associated with their multi-levels, there might possibly generate up to 144 (32_{× 2}4₎ concepts of smart pads. Consecutively, to perform concept evaluation in a market-oriented manner, customer preference needs to be coupled into the entire process and the results are depicted in Table 10. Here, for simpliﬁcation, only the top three priorities among 144 alternatives are demonstrated and their transitions arisen from balancing the trade-offs between DAs are also indicated.

Meanwhile, with the aid of KM, customers' perceptions of MRs are elicited with respect to two scenarios: positive delight for functional fulfillment and negative disgust for dysfunctional unfulfillment (see Eqs.(8)–(9)). Similar to the concept of conjoint analysis, the range which is defined by “delight less disgust” is used to derive the priorities of MRs for three segments (see Eq.(10)). As shown inTable 11, the pattern of R6≻R2≻R3 consists of the top three MRs for S1 while R2≻R3≻R5 and R4≻R3≻R1 are presented by S2 and S3, respectively. Obviously, thesefindings imply that Kano model is effective to reveal customers' vague perceptions of MRs and also present the relative priorities for different segments. Hence, it might be suitable for gath-ering and tracking new ideas of the next-generation products. 4.3. Identifying the causal relationships between MRs and DAs

Referring toTable 7again, several cross-functional managers are invited tofill out their assessments on the interdependences between MRs and DAs (also seeFig. 3). After consulting all experts, their eval-uation results are aggregated as a 12 × 12 direct-relation matrix. Then, by virtue of DEMATEL, four main scores of all factors could be calcu-lated and shown inTable 12(also see Eqs.(11)–(15)). In brief, the “active” score of a factor denotes the sum of its dispatched impact on other factors and the“passive” score represents the sum of its re-ceived influence sent from other factors. Intuitively, the “prominence” score de_{fined by adding the “active” score to the “passive” score} Table 9

Extracted customer preference of DAs for different segments.

Attributes Speciﬁcations S1 S2 S3

A1 CPU a11— Quad 0.108 0.226 0.108

a12— Dual 0.065 0.113 0.084 a13— Atom 0.043 0.038 0.048 A2 ROM capacity a21— 8 GB 0.038 0.103 0.037 a22— 4 GB 0.020 0.044 0.027 A3 operating system a31— Android 0.137 0.117 0.031

a32— iOS 0.130 0.139 0.023

a33— Window 0.059 0.022 0.006 A4 screen size a41— 9–10 in. 0.129 0.039 0.196 a42— 7–8 in. 0.124 0.046 0.136 A5 front/back camera a51— 200 M/500 M 0.032 0.023 0.082 a52— 130 M/300 M 0.017 0.015 0.035 A6 battery capacity a61— above 5000 mAH 0.059 0.049 0.126 a62— below 5000 mAH 0.039 0.026 0.060

Table 10

The top three priorities of smart-pad alternatives. S1 (home) S2 (business) S3 (entertainment) #1 #2 #3 #1 #2 #3 #1 #2 #3

A1 a11 a11 a11 a11 a11 a11 a11 a11 a11

A2 a21 a21 a21 a21 a21 a21 a21 a21 a22

A3 a31 a31 a32 a32 a32 a32 a31 a32 a31

A4 a41 a42 a41 a42 a41 a42 a41 a41 a41

A5 a51 a51 a51 a51 a51 a52 a51 a51 a51

A6 a61 a61 a61 a61 a61 a61 a61 a61 a61

Table 11

Elicited customer perception of MRs for different segments.

A M O R I Q Delight Disgust Range Rank

S1 home R1 20% 47% 33% 0.530 −0.800 1.330 3 R2 45% 20% 35% 0.800 −0.550 1.350 2 R3 36% 18% 38% 8% 0.740 −0.560 1.300 6 R4 30% 26% 38% 6% 0.680 −0.640 1.320 4 R5 28% 35% 37% 0.650 −0.720 1.370 1 R6 28% 40% 32% 0.600 −0.720 1.320 4 S2 business segment R1 15% 55% 30% 0.450 −0.850 1.300 5 R2 13% 42% 45% 0.580 −0.870 1.450 1 R3 18% 40% 42% 0.600 −0.820 1.420 2 R4 36% 23% 32% 9% 0.680 −0.550 1.230 6 R5 14% 46% 40% 0.540 −0.860 1.400 3 R6 27% 25% 42% 6% 0.690 −0.670 1.360 4 S3 entertainment segment R1 18% 48% 34% 0.520 −0.820 1.340 3 R2 34% 23% 38% 5% 0.720 −0.610 1.330 5 R3 25% 35% 40% 0.650 −0.750 1.400 2 R4 8% 51% 41% 0.490 −0.920 1.410 1 R5 15% 41% 39% 5% 0.540 −0.800 1.340 3 R6 31% 30% 32% 7% 0.630 −0.620 1.250 6

(7)

reflects the importance degree of a factor. By contrast, the “influence” score defined by subtracting the “active” score from the “passive” score indicates the causality of a factor.

Furthermore, based onTable 12, the complicated interrelationships among all factors (e.g. MRs and DAs) could be visualized and portrayed in Fig. 4. Recall that the coordinates consist of (T+R, T−R), where “T+R” represents the horizontal axis and “T−R” denotes the vertical axis. Apparently, all DAs (denoted by the red_{“squares”) are categorized} into the“cause” (dispatcher) group because of having “positive” influ-ence. Conversely, all MRs (denoted by the blue“diamonds”) are classified into the“effect” (receiver) group due to having “negative” influence. More importantly, this structural diagram successfully assists product managers in separating MRs and DAs into two distinct groups and hence product engineers couldfind clues for enhancing specific MRs through improving corresponding DAs.

5. Conclusion and future research

For eventual survival and continuous growing of an enterprise, product planners or project managers spend most of their time to make crucial decisions under uncertain business environments. In a globally customized economy, meeting customers' requirements and going beyond their expectations still capture the focal point to achieve successful new product development. For instance, afirm might want to know who are its potential customers and which char-acteristics do they own? Meanwhile, afirm might be also interested in understanding what are its target segments and which product va-rieties should be offered tofit these segments? In practice, identifying profitable niche segments and configuring potential product alterna-tives with respect to these segments are vitally important to fulfill the aforementioned goals. In this paper, a hybrid framework combining AHP, KM, with DEMATEL is proposed to incorporate customer prefer-ence and customer perception into the decision-making process of concept generation and product evaluation.

To validate the applicability of our proposed approach, an industrial example regarding conﬁguring and prioritizing smart pads is de-monstrated for distinct segments. Based on our experimental results, this paper contributes to this domain by demonstrating the following: (1) learning which design attributes are crucial to segment the entire

market, (2) generating and evaluating competitive product alternatives in a market-oriented manner, (3) understanding customer perception of marketing requirements to uncover new ideas of the next-generation products, and (4) recognizing the complicated interrelationships be-tween marketing requirements and design attributes to offer managerial insights for industrial practitioners. To further provide decision supports on managing various product lines, we might integrate the current framework with other data mining techniques to explore how customer preference and/or customer perception impacts on customers' eventual selection of substitute products (i.e. ultrabooks or tablets) in future studies.

Acknowledgments

The authors would like to thank for two anonymous referees' helpful comments. This paper isﬁnancially supported by the Taiwan National Science Council under grant NSC-101-2410-H-009-002. References

[1] Y. Akao, Quality Function Deployment: Integrating Customer Requirements into Product Design, Springer, Cambridge, MA, 1990.

[2] R.G. Askin, D.W. Dawson, Maximizing customer satisfaction by optimal speciﬁca-tion of engineering characteristics, IIE Transacspeciﬁca-tions 32 (1) (2000) 9–20. [3] Z. Ayağ, An integrated approach to evaluating conceptual design alternatives in a

new product development environment, International Journal of Production Re-search 43 (4) (2005) 687–713.

[4] Z. Ayağ, A fuzzy AHP-based simulation approach to concept evaluation in a NPD environment, IIE Transactions 37 (4) (2005) 827–842.

[5] C. Berger, R. Blauth, D. Boger, C. Bolster, G. Burchill, W. DuMouchel, F. Pouliot, R. Richter, A. Rubinoff, D. Shen, M. Timko, D. Walden, Kano's method for under-standing customer-deﬁned quality, Center for Quality of Management Journal 2 (4) (1993) 3–35.

[6] T.K. Chien, C.T. Su, Using the QFD concept to resolve customer satisfaction strate-gy decisions, International Journal of Quality and Reliability Management 2 (3) (2003) 345–359.

[7] Y.F. Chang, C.S. Chen, H. Zhou, Smart phone for mobile commerce, Computers Standards & Interfaces 31 (2009) 740–747.

[8] A. Chaudhuri, M. Bhattacharyya, A combined QFD and integer programming framework to determine attribute levels for conjoint study, International Journal of Production Research 47 (23) (2009) 6633–6649.

[9] F. Fogliatto, J.C. Giovani, D. Siverira, Mass customization: a method for market segmentation and choice menu design, International Journal of Production Eco-nomics 111 (2008) 606–622.

[10] E. Fontela, A. Garbus, The DEMATEL Observer, Battelle Institute, Geneva Research Centre, 1976.

[11] G. Işıklar, G. Büyüközkan, Using a multi-criteria decision making approach to evaluate mobile phone alternatives, Computers Standards & Interfaces 29 (2007) 265–274.

[12] J. Jiao, Q. Ma, M.M. Tseng, Towards high value-added products and services: mass customization and beyond, Technovation 14 (3) (2001) 173–185.

[13] J. Jiao, Y. Zhang, Product portfolio planning with customer-engineering interac-tion, IIE Transactions 37 (2005) 801–814.

[14] J. Jiao, C.H. Chen, Customer requirement management in product development: a review of research issues, Concurrent Engineering 14 (3) (2006) 173–185. [15] N. Kano, Attractive quality and must-be quality, The Journal of Japanese Society

for Quality Control 14 (2) (1984) 39–48.

[16] P. Kotler, K. Keller, Framework for Marketing Management, Prentice Hall, NY, 2012.

[17] C.K. Kwong, H. Bai, A fuzzy AHP approach to the determination of importance weights of customer requirements in quality function deployment, Journal of In-telligent Manufacturing 13 (2002) 367–377. MR1 . . . . MRm DA1 . . . . DAn MR1 . MRm m

m× zero matrix m×n zero matrix

DA1

.

DAn

m

n× causal matrix n×n zero matrix

Fig. 3. A schematic representation for the input matrix of the DEMATEL.

Table 12

Visualizing a causal diagram between MRs and DAs via the DEMATEL. Active score Ti Passive score Rj Prominence score Ti+ Rj Inﬂuence score Ti−Rj R1 0 0.533 0.533 −0.533 R2 0 0.733 0.733 −0.733 R3 0 0.600 0.600 −0.600 R4 0 0.800 0.800 −0.800 R5 0 0.333 0.333 −0.333 R6 0 0.667 0.667 −0.667 A1 0.867 0 0.867 0.867 A2 0.667 0 0.667 0.667 A3 1.000 0 1.000 1.000 A4 0.600 0 0.600 0.600 A5 0.200 0 0.200 0.200 A6 0.333 0 0.333 0.333

Cause and effect diagram

-1.000 -0.500 0.000 0.500 1.000 1.500 0.000 0.500 1.000 1.500 Prominence Influence MRs DAs A5 A6 A4 A2 A1 A3 R5 R1 R3 R6 R2 R4

(8)

[18] M.C. Lin, C.C. Wang, T.C. Chen, A strategy for managing customer-oriented prod-uct design, Concurrent Engineering 14 (3) (2006) 231–244.

[19] M.C. Lin, C.C. Wang, M.S. Chen, C.A. Chang, Using AHP and TOPSIS approaches in customer-driven product design process, Computers in Industry 59 (2008) 17–31.

[20] R.D. Luce, J.W. Turkey, Simultaneous conjoint measurement: a new type of funda-mental measurement, Journal of Mathematical Psychology 1 (1964) 1–27. [21] K. Matzler, H.H. Hinterhumber, How to make product development projects more

successful by integrating Kano's model of customer satisfaction into quality func-tion deployment, Technovafunc-tion 18 (1) (1998) 25–38.

[22] M.M. Rashid, A review of state-of-art on Kano model for research direction, International Journal of Engineering, Science and Technology 2 (12) (2010) 7481–7490.

[23] T.L. Saaty, The Analytic Hierarchy Process, McGraw-Hill, New York, 1980. [24] Y. Sereli, P. Kauffmann, E. Ozan, Integration of Kano's model into QFD for multiple

product design, IEEE Transactions on Engineering Management 54 (2) (2007) 380–390.

[25] J. Song, T. Jang, S.Y. Sohn, Conjoint analysis for IPTV service, Expert Systems with Applications 36 (2009) 7860–7864.

[26] K.C. Tan, X.X. Shen, Integrating Kano's model in the planning matrix of quality function deployment, Total Quality Management 11 (8) (2000) 1141–1151. [27] C.H. Wang, Outlier identiﬁcation and market segmentation using kernel based

clus-tering techniques, Expert Systems with Applications 36 (2) (2009) 3744–3750. [28] C.H. Wang, Apply robust segmentation to the service industry using kernel

in-duced fuzzy clustering techniques, Expert Systems with Applications 37 (12) (2010) 8395–8400.

[29] C.H. Wang, J.N. Chen, Using quality function deployment for collaborative product design and optimal selection of module mix, Computers and Industrial Engineer-ing 63 (4) (2012) 1030–1037.

[30] G.S. Wasserman, On how to prioritize design requirements during the QFD plan-ning process, IIE Transactions 25 (3) (1993) 59–65.

Chih-Hsuan Wang received his Ph.D. degree at the Na-tional Taiwan University in 2005. Prof. Wang also re-ceived his B.S. degree in electronic engineering from the National Chiao Tung University in Taiwan in 1993 and M.S. degree in opto-electrical engineering from the National Taiwan University in 1995. He has been a visit-ing scholar at Texas A&M University and a research asso-ciate at the University of Tennessee. He is now an assistant professor at the Department of Industrial Engi-neering & Management of National Chiao Tung Universi-ty and his research interests include data mining, product development, business intelligence, and service design. He already published several journal articles in IIE transactions, International Journal of Production Re-search, Expert Systems with Applications, Journal of Intelligent Manufacturing, and Com-puters & Industrial Engineering. Since 2007, he also serves as the session chair for IEEE and APIEMS.

One-Zen Hsueh is currently a PhD candidate at the In-stitution of Industrial Engineering and Management, National Chao Tung University, Taiwan. Her research interests are Operation Management, Supply Chain Management, Performance Evaluation, Information Management and Business Intelligence, etc. For the past 15 years she has been working most of the time in the computer manufacturing industry. At the present she is the director of the Division of Operation Architecture and Analysis at Shuttle Inc. as well as the Special Assis-tant to the Shuttle headquarter General Manager.