A Model for Selecting Technologies in New Product Development

Due to fast changing technologies, shortening product lifecycles, and increased global competition, companies today often need to develop new products continuously and faster. Successful introduction and acceleration of new product development NPD is important to obtain competitive advantage for companies. Since technology selection for NPD involves complex decision makings that are critical to the profitability and growth of a company, the selection of the most appropriate technology for a new product requires the use of a robust decision-making framework capable of evaluating several technology candidates based on multiple criteria. This paper presents an integrated model that adopts interpretive structural modeling ISM and fuzzy analytic network process FANP to evaluate various different available technologies for NPD. The ISM is used to understand the interrelationships among the factors, and the FANP is to facilitate the evaluation process of decision makers under an uncertain environment with interrelated factors. A case study of a flat panel manufacturer is performed to examine the practicality of the proposed model. The results show that the model can be applied for group decision making on the available technology evaluation and selection in new product development.


Introduction
Products life cycle is shortening continuously under the rapid transition of industrial structure and technology advancement.In order to excel in the competitive markets, new product introduction is important to get new sales and profits.Thus, a company has to keep developing new products to attract customers.In today's highly competitive high-tech industry, technology is developed rapidly, and the adoption of new technology is the only means to gain competitive advantage.While there are various technologies that may be used in a new product, different aspects must be considered to determine which technology is most suitable under the circumstances.Thus, a good objective model for evaluating and selecting the right technology to be adopted in the new product is required.
The demands for technology evaluation have increased with the flourishing development of technology licensing, technology transaction, or joint venture on the one hand and the pressing needs of new product introduction on the other hand 1 .In the evaluation of technologies, objective factors, such as cost, profit, revenue and time of completion, and subjective factors, such as flexibility, learning, and capacity increment, all must be considered 2 .Therefore, the problem is a multiple-criteria decision-making MCDM in nature.Some MCDM methods are proposed in literature.Punniyamoorthy and Ragavan 2 proposed a deterministic decision making approach, which adopted analytic hierarchy process AHP and the Brown-Gibson model EBG , for technology selection.Since real world knowledge may be fuzzy rather than precise, a useful decision-making model may need to have the ability to handle fuzzy assessments.Chuu 3 proposed a group decision-making model based on fuzzy multiple attributes analysis to assess the suitability of advanced manufacturing technology alternatives, and the fusion of fuzzy assessment data is performed by maximum entropy ordered weighted averaging MEOWA operators.Hsu et al. 4 applied fuzzy analytic hierarchy process FAHP to find the importance of each criterion in evaluating regenerative technologies.Lee et al. 5 presented an integrated model for evaluating various technologies for NPD by taking into account the benefits, opportunities, costs and risks BOCR aspects of different technologies, and adopting interpretive structural modeling ISM , and fuzzy analytic network process FANP .
To summarize, the technology selection in new product development NPD problem is a fuzzy multiple-criteria and group decision-making problem which involves the consideration of fuzzy assessments and the opinions of multiple experts.Although technology selection is not a new subject in the field of management, no research, in the authors' understanding, has considered the interrelationship of the criteria in the decision making process by incorporating interpretive structural modeling ISM and fuzzy analytic network process FANP .The rest of this paper is organized as follows.Section 2 reviews the related methodologies, and Section 3 develops an integrated model for selecting technologies in NPD.Section 4 examines the model on a flat panel manufacturer in Taiwan for NPD.Some concluding remarks are made in Section 5.

Interpretive Structural Modeling (ISM)
Interpretive structural modeling ISM was first proposed by Warfield to understand complex situations and to put together a course of action for solving a problem 6, 7 .It is a process to develop a map of the complex relationships among elements by calculating a binary matrix, called relation matrix 8 .The relation matrix is obtained through individual or group mental models to represent the relations of the elements 8 .A question such as "Does criterion x i affect criterion x j ?" is asked.If the answer is yes, then π ij 1; otherwise, π ij 0. A reachability matrix is obtained to consider transitivity, and a final reachability matrix can be calculated under the operators of the Boolean multiplication and addition.The final reachability matrix can reflect the convergence of the relationship among the elements.Since its introduction, the ISM has been applied in various fields, including assisting government bodies to prioritize activities, facilitating companies to select projects, and aiding researchers in relevant works.Some recent works that applied ISM are briefly reviewed here.

Fuzzy Analytic Network Process (FANP)
The AHP is a popular MCDM methodology which has been applied vastly in various fields, for example, in Punniyamoorthy and Ragavan 2 , Hsu et al. 4 , and Chai and Sun 12 .The analytic network process ANP approach is a generalization of the AHP 13 .The ANP approach replaces hierarchies with networks, in which the relationships between levels are not easily represented as higher or lower, dominated, or being dominated, directly or indirectly 14 .Figure 1 shows the structural difference between a hierarchy and a network, in which a node represents a component or cluster with elements inside it, and an arc denotes the interaction between two components.The direction of an arc indicates dependence, a twoway arrow represents the interdependencies between two components, and a loop signifies the inner dependence of elements within a component 15 .
The fuzzy set theory can be incorporated into the ANP to tackle the uncertain and imprecise pairwise comparison in the conventional ANP, and it is called FANP.The procedure for the FANP can be summarized as follows 10, 16 .
1 Decompose the problem into a network.
2 Prepare a questionnaire based on the constructed network and ask experts to fill out the questionnaire.
3 Transform the scores of pairwise comparison into linguistic variables and aggregate the results of the experts' questionnaires.
4 Obtain crisp numbers by defuzzying the synthetic triangular fuzzy numbers.
5 Calculate the maximum eigenvalues and eigenvectors.
6 Check the consistency property of the matrix. . . .8 Form a weighted supermatrix.
9 Obtain the limit supermatrix and the priority weights of the alternatives.

A Model for Evaluating Technologies
In this study, we construct a technology evaluation model that incorporates FANP and ISM.The proposed steps are as follows.
1 Form a committee of experts to define the technology evaluation problem in the flat panel industry.Through literature review and interview with experts, a network for evaluation is constructed.The network should comprise the goal, criteria, subcriteria and alternatives.
2 Use ISM to determine the inter-relationship among the subcriteria, with the same upper-level criterion: 2.1 prepare the relation matrix to construct the relationships of the subcriteria with the same upper-level criterion.Let the subcriteria under criterion c as x c i , i 1, 2, 3, . . ., n; x c i is the ith subcriterion, x c j is the jth subcriterion, and π c ij is the relation between ith and jth subcriteria.If otherwise, π c ji 0. In the case that there are several experts, a questionnaire is prepared to ask the contextual relationship between any two subcriteria, and the associated direction of the relation.A relation matrix which shows the contextual relationship among the subcriteria is established for each expert.The geometric mean of experts' opinions on the relationship between a pair of subcriteria is calculated.A threshold value is used to determine whether the two subcriteria are dependent or not 9, 16 .If the geometric mean value between two subcriteria, that is, π c ij , in the mean relation matrix is higher than the threshold value, x c j is deemed reachable from x c i , and we let π ij 1 9, 16 .Establish relation matrix which shows the contextual relationship among the factors.The relation matrix D c is presented as follows: i is related to x c j and x c j is related to x c p , then x c i is necessarily related to x c p .Under the operators of the Boolean multiplication and addition i.e., 1 × 0 0 × 1 0, 1 0 0 1 1 , a convergence can be met where π c * ij denotes the relation between the ith sub-criterion and the jth subcriterion.
Plot the network structures for the subcriteria with the same upper-level criterion.
Employ questionnaire and collect experts' opinions.Experts are asked to pairwise compare the elements in a questionnaire.The scores of pairwise comparison of each part of the questionnaire from each expert are transformed into linguistic variables by the transformation concept listed in Table 1.The fuzzy positive reciprocal matrix can be defined as 17, 18 : where A k is the positive reciprocal matrix of expert k, a ij is the relative importance between decision elements i and j, and a ij 1, for all i j and a ij 1/ a ji , for all i, j 1, 2, . . ., n.
Build aggregated pairwise comparison matrices.Apply geometric average approach to aggregate experts' responses, and prepare fuzzy aggregated pairwise comparison matrices.Let there be k experts, every pairwise comparison between two criteria has k positive reciprocal triangular fuzzy numbers.Employ geometric average approach to aggregate multiple experts' responses, and the aggregate fuzzy positive reciprocal matrix is Form aggregated pairwise comparison matrices by defuzzifying fuzzy aggregated pairwise comparison matrices.Synthetic triangular fuzzy numbers a * ij x ij , y ij , z ij can be transformed into crisp numbers using a defuzzification method, such as the Centroid method: , ∀i, j 1, 2, . . ., n.

3.7
The consistency test 19 is performed by calculating the consistency index CI and consistency ratio CR : where n is the number of items being compared in the matrix, and RI is random index 19 .If CR is less than 0.1, the experts' judgments are consistent.If the consistency test is not passed, the part of the questionnaire must be done again.
8 Form an un-weighted supermatrix.The priority vectors are entered in the appropriate columns of a matrix, called an un-weighted supermatrix, as shown in Figure 2.
9 Form a weighted supermatrix.A weighted super-matirx is prepared so that each column in the supermatrix sums to unity.When there is interdependence among clusters in a network, some columns of a supermatrix sum to more than one.The supermatrix must be transformed first to make it stochastic.The procedure can be found in Saaty 13 and Lee et al. 16 .
10 Calculate the limit supermatrix.The limit supermatrix is obtained by raising the weighted supermatrix to the power of 2p 1 so a stable supermatrix is reached.
11 Rank the alternatives.The priority weights of alternatives are shown in the alternative-to-goal block in the limit supermatrix.

A Case Study
With an increasing global demand of information technology, flat panels with low weight, slender profile, low power consumption, high resolution, high brightness and low radiance  are demanded by end-users.While manufacturers enter and expand their production capacity, the production value of the flat panel industry increases tremendously, and an extremely competitive and cost-cutting war is foreseeable.Therefore, manufacturers, in order to survive in this global competitive market, need to develop products with advanced technology.
Through literature review and interview with five experts in the company, the most important criteria for selecting a technology for producing panels are R&D, cost and quality.Under each criterion, there are subcriteria, which are interrelated.The technologies, which are the most probable to be adopted and which are considered currently in making smallto-medium-sized panels, are organic light-emitting diode OLED , thin film transistor TFT , amorphous silicon a-Si , low temperature poly silicon LTPS , and color supertwist nematic CSTN .The network for the problem is constructed in Figure 3.
ISM is applied next to determine the interrelationship among subcriteria under each upper-level criterion.The five experts are asked to fill out a questionnaire to determine the contextual relationship between any two subcriteria, and the associated direction of the relation.The experts' opinions on the relationship between each pair of subcriteria are synthesized using the geometric mean method.A threshold value of 0.5 is used to determine whether the two subcriteria are dependent or not 9, 16 .For example, the integrated relation matrix among subcriteria under the criterion R&D is as shown in Table 2.

4.1
The final reachability matrix M * 1 for subcriteria under R&D is Based on M * 1 , the inter-relationship among the six subcriteria under the criterion R&D can be depicted as in Figure 4.The same procedure can be carried out for determining the inter-relationship among the subcriteria under the criterion cost and the criterion quality.

4.3
The Centroid method is applied next to prepare a defuzzified comparison matrix.For example, with the synthetic triangular fuzzy number for the comparison between R&D and cost of 0.572,1.125,1.81, the defuzzified comparison between R&D and cost is 1.169.The defuzzified aggregated pairwise comparison matrix is:

4.6
Since CR is less than 0.1, the experts' judgment is consistent.If the consistency test fails, the experts are required to fill out the specific part of the questionnaire again until a consensus is met.
The relative importance of the R&D, cost, and quality is 0.32853, 0.38336, and 0.28812, respectively.Based on the questionnaires, a similar procedure is done to calculate the priority vector for the importance of subcriteria under each criterion, the inter-relationship among subcriteria under each criterion, and the performance of alternatives with respect to each subcriterion.These priorities are entered into the designated places in a supermatrix, as shown in Table 3.To make the matrix stochastic, a weighted supermatrix is formed, as shown in Table 4. Finally, by taking the weighted supermatrix to a large power, a limit supermatrix is obtained, as shown in Table 5.The priorities of the alternatives can be seen from the alternative-togoal block of the limit supermatrix.The ranking and the priorities of the alternatives are TFT 0.3176 , LTPS 0.1959 , a-Si 0.1790 , OLED 0.1672 , and CSTN 0.1403 .To summarize, the company, with the consideration of various importance of criteria and subcriteria, should select TFT as the most suitable technology for its new product.
A detailed evaluation of the expected performance of the five technologies under different criteria and subcriteria is shown in Table 6.For example, under criterion R&D C 1 , TFT A 2 ranks the first, with a priority of 0.33027, followed by LTPS A 4 , a-Si A 3 , OLED A 1 , CSTN A 5 with priorities of 0.19837, 0.17506, 0.17198 and 0.12432, respectively.Under sub-criterion environmental SC 11 , TFT A 2 also has the highest score, 0.28242, while the scores for OLED A 1 , a-Si A 3 , LTPS A 4 , and CSTN A 5 are 0.23462, 0.16523, 0.17641, and 0.14131, respectively.In fact, TFT A 2 performs the best under all criteria and all subcriteria.However, note that TFT A 2 is not always the best alternative under every sub-criterion initially.This can be seen from the alternative-to-sub-criteria block of the unweighted supermatrix in Table 3.It ranks the second, second, third, second, and third under the subcriteria environmental SC 11 , technology transfer SC 14 , technology risk SC 15 , inventory cost SC 23 , and return on investment SC 24 , respectively.Its ranking becomes the best under all subcriteria after the inter-relationship of the subcriteria is considered, that is, when a convergence is reached in the limit supermatrix.Since TFT A 2 has an integrated largest priority of 0.31764 and it performs the best under all criteria and subcriteria, it has a better overall expected performance than other technologies.Thus, TFT A 2 should be selected under the current circumstances.
A company should understand its underlying reasons in selecting a new technology.As calculated before and also shown in Table 7, the priorities of the three criteria, R&D, cost and quality are 0.32853, 0.38336 and 0.28812, respectively.Criterion cost ranks first, followed by R&D and quality.This means that cost-related issues are the primary concerns for the company.However, R&D factors are also very important since the company is in a high-tech industry and NPD is essential for the long-term survival of the company.The priorities of         the subcriteria are also listed in Table 7.It is not appropriate to study the priorities of the subcriteria with the consideration of interdependence due to the fact that there are transient subcriteria and sink subcriteria in the network 13 .Therefore, the priorities of the subcriteria, with the assumption of independence among the elements, are examined here.Under the criterion R&D, R&D capability SC 12 , with a priority of 0.23477, is the most important sub-criterion, followed by core technology SC 13 with a priority of 0.23211 and technology maturity SC 16 with a priority of 0.20988.This means that under the aspect of R&D the Mathematical Problems in Engineering company concerns the most about whether it has adequate R&D capability to develop the panel with the alternative technology, whether the technology is a core technology of the company, and whether the technology is a matured technology in the industry.Under the criterion cost, on-time delivery SC 25 is the most important with a priority of 0.26516.Inventory cost SC 23 and production cost SC 22 ranked the second and the third with priorities of 0.20772 and 0.19694, respectively.Because the company's operation is basically make-to-order, on-time delivery is very important to meet the fundamental demand of customers.With the obsoleteness issue of high-tech products and relatively high production cost, the company must emphasize on cost reduction.Under the criterion quality, reliability SC 32 is the most important sub-criterion, with a priority of 0.52578.This implies that the selected technology alternative should be able to provide reliable products to customers.
In conclusion, the company should base on the priorities of the criteria and subcriteria in evaluating and selecting the technology for NPD.

Conclusions
Companies often need to develop new products in order to survive in the market.Since different technologies are appropriate in different setting under different time frame and for specific purposes, understanding the environment in selecting the most appropriate technology can be a complex decision with many variables to consider.In this paper, an integrated model for selecting the most appropriate technology is developed.The ISM is applied to understand the inter-relationships among the factors, and the ANP is used to calculate the priorities of the technologies.A case study is used to verify the practicality of the proposed model.The model also provides a good evaluation of the factors on technology evaluation, and these factors can be a reference for the company or other companies in the industry to make performance evaluation of technologies that can be adopted in designing a new product.The proposed model can be tailored and applied for a manufacturer which needs to make the technology selection decision.

Figure 1 :
Figure 1: Differences between a hierarchy and a network.

1 , 2 , 1 2. 2 2 2. 3
. . ., n; j 1, 2, . . ., n, 3.develop reachability matrix and check for transitivity.The initial reachability matrix M c is calculated by adding D c with the unit matrix I: M c D c I, 3.develop final reachability matrix M * c .The transitivity of the contextual relation means that if subcriterion x c

Figure 3 :
Figure 3: The network for selecting the most suitable technology.

1 3
and λ max of the defuzzified aggregated pairwise comparison matrix for criteria are calculated: is performed by calculating the consistency index CI and consistency ratio CR : CI λ max − n n −

Goal
0.17506 0.17560 0.18791 0.16523 0.17161 0.17080 0.17192 0.19356 0.17727 0.16312 0.17954 0.17677 0.19425 0.16594 0.20246 0.18664 0.0.124320.14879 0.14732 0.14131 0.10392 0.11337 0.11885 0.11823 0.16171 0.17077 0.14652 0.12239 0.12817 0.17302 0.13925 0.15184 0.14754 0 0 0 0 1 Yang et al. 9 applied the ISM to study the relationships among the subcriteria and used integrated fuzzy MCDM techniques to study the vendor selection problem. Lee et al. 10 employed the ISM to determine the interrelationship among the critical factors for technology transfer of new equipment in high technology industry and applied the FANP to evaluate the technology transfer performance of equipment suppliers.Lee and Lin 11 constructed a systematic framework that incorporates fuzzy Delphi method FDM , fuzzy interpretive structural modeling FISM and FANP into quality function deployment QFD for new product development.

Table 1 :
Membership functions of triangular fuzzy numbers.Calculate priority vectors and examine the consistency of the aggregated pairwise comparison matrices.Calculate the maximum eigenvalue, λ max , and the eigenvector, w, for the matrix 19 :

Table 6 :
Expected performance of the five technologies.

Table 7 :
Priorities of factors.