Designing Products Using Quality Function ... - Springer Link

2 downloads 0 Views 2MB Size Report
A Comparison in a Market for Elderly People. Samah Abu-Assab and Daniel Baier. Abstract In this paper, we compare two product design approaches, quality ...
Designing Products Using Quality Function Deployment and Conjoint Analysis: A Comparison in a Market for Elderly People Samah Abu-Assab and Daniel Baier

Abstract In this paper, we compare two product design approaches, quality function deployment (QFD) and conjoint analysis (CA), on the example of mobile phones for elderly people as a target group. Then, we compare between our results and the results from former similar comparisons, e.g., Pullman et al. (J Prod Innov Manage 19(5):354–364, 2002) and Katz (J Innov Manage 21:61–63, 2004). In this work, the same procedures and conditions are taken into consideration as that taken by Pullman et al. in their paper. They viewed the relation between the two methods: QFD and CA as a complementary one in which both should be simultaneously implemented since each provide feedback to the other. They concluded that CA is more efficient in reflecting the end-users’ present preferences for the product attributes, whereas QFD is definitely better in satisfying end-users’ needs from the developers’ point of view. Katz in his response from a practitioner’s point of view agreed with Pullman et al. However, he concluded that the two methods are better used sequentially and that QFD should precede conjoint analysis. We test these results in a market for elderly people. Keywords Conjoint analysis  Product design  Quality function deployment.

1 Introduction Conjoint analysis (CA) is widely accepted by marketing researchers as a tool to measure consumer preferences, whereas quality function deployment (QFD) is often used to translate the customer requirements into appropriate technical requirements for the various stages of product development, e.g., Sullivan (1986), Terninko (1997). Since the 1980s, the traditional CA and QFD approaches have been proposed, disseminated and improved. However, as both traditional approaches are Samah Abu-Assab (B) Chair of Marketing and Innovation Management, Brandenburg University of Technology Cottbus, P.O. Box 101344, 03013 Cottbus, Germany, e-mail: [email protected]

A. Fink et al., (eds.), Advances in Data Analysis, Data Handling and Business Intelligence, Studies in Classification, Data Analysis, and Knowledge Organization, c Springer-Verlag Berlin Heidelberg 2010 DOI 10.1007/978-3-642-01044-6 47, 

515

516

S. Abu-Assab and D. Baier

burdened with a number of practical and theoretical problems, recently, many researchers and practitioners have focused on finding more efficient methods to overcome these obstacles and improve the applicability and the results of those methods, e.g., Baier (1998); Pullman, Moore, and Wardell (2002); Katz (2004); Baier and Brusch (2005); Kazemzadeh, Behzadian, Aghdasi, and Albadvi (2009). So, e.g., Kazemzadeh et al. (2009) pinpointed many of the problems that the traditional QFD approach are fraught with: the difficulty to differentiate between diverse and conflicting customer demands and needs, the hardship to prioritize customer needs and engineering requirements with the rating used, the imprecise way in which the customer needs are translated and correlated among technical requirements as well as the relation between customer requirements and technical requirements, and – finally – the necessity to trade off among the various customer needs. The authors used CA within QFD to measure the importance of different customer needs and applied cluster analysis for benefit segmentation. Baier and Brusch (2005) demonstrated a new approach “conjoint QFD” that combined CA into QFD. The correlation between the customers’ needs and the engineering characteristics as well as the importance of the customer needs are estimated using CA. Then they compared the predictive validity of the new approach “conjoint QFD” with the traditional approach. Their results designated by hook or by crook that the validity of the new approach surpasses the validity of the traditional QFD one. Pullman et al. (2002) compared QFD and CA by applying each at the example of a new all-purpose harness for the beginning/intermediate ability climber which added to an existing harness-product line of a leading manufacturer. In their conclusion, they stressed the fact that the two approaches are not competitive but rather complementary and should be simultaneously implemented in which each supply a feedback to the other. In his reply, Katz (2004) concluded from a practitioner point of view that the two approaches should be considered as supplementary rather than complementary and that QFD should be implemented first in the early product development stages and then be followed by CA. However, further comparisons are needed. Consequently, in this paper we compare QFD and CA for product design in a similar way like Pullman et al. The application field is the mobile phone market for the elderly group 50 plus (50 years old or elder). The paper is structured as follows: In the next section, Pullman et al.’s experiment is described including their results. Then, the own comparison is described. Finally, the results from Pullman et al. and our comparisons are summarized. The paper closes with conclusions and outlook.

2 Product Design in a Climbing Harness Market A quick review of the Pullman et al.’s experiment (description, settings and results) is presented in this section. The design object was an all-purpose climbing harness for the beginning/intermediate ability climber. Figure 1 (from Pullman et al., 2002) shows its key features.

Designing Products Using QFD and CA: A Comparison

517

2.1 Application of Conjoint Analysis For applying CA, firstly, an expert team was formed that collected a lot of features for climbing harnesses from numerous resources (e.g., self-inspection, catalogues, and discussions with others). After a managerial judgement the initial list was decreased to nine attributes with minimum two to maximum five levels each for the conjoint experiment. Then, a questionnaire was developed where these different harness features where described, including 20 harness conjoint profiles and two harness choice sets. Finally, conjoint data were collected from 105 respondents and analyzed applying Hierarchical Bayes logistic regression as analysis method. Figure 1 already shows the resulting average utility weights for the nine selected attributes. As a result, it could be seen that – on average – respondents had higher utilities for brand B, stuffed webbing harnesses, wide waist belts, threaded buckles, a belay loop, four gear loops, a dedicated tie-in loop, adjustable leg loops, and lowest prices.

Feature Levels Brand A (0.265); B (0.132); C (0.065); D (0.069) Harness construction Webbing/Fleece (0.040); Stuffed webbing (0.071); Laminate foam (0.053); Thermo-formed (0.023) Waist belt width Narrow (0.114); Wide (0.114) Buckle Threaded (0.401); Non-threaded (0.401) Belay loop Yes (0.440); No (0.440) Gear loops Two (0.670); Four (0.670) Dedicated tie-in loop Yes (0.241); No (0.241) Leg loops Fixed (0.414); Adjustable (0.414) Price $39 (0.500); $50 (0.167); $61 (0.167); $72 (0.500); $83 (0.833)

Fig. 1 All-purpose climbing harness with average utilities for attribute-levels

518

S. Abu-Assab and D. Baier

2.2 Application of Quality Function Deployment For constructing the so-called house of quality, firstly, customer needs (CNs) had to be collected. Seventeen one-to-one interviews were conducted in a climbing gym. Then, a second group of climbers were asked to rate the importance of these CNs on a six-point-rating scale. A third group of 30 respondents assessed the competitive brands A, B, C, and D. They were asked to rate how well each of them met the different CNs, again on a six-point-rating scale and an additional ten-point-buying intention scale. Secondly, the engineering characteristics (ECs) and target values for them were defined by using workshops with the above described expert team. They identified one or more measurable ECs for each CN and esteemed the correlation between each EC and each CN on a 5 to +5 rating scale. Finally, the part-deployment matrix was constructed with ECs as rows and the so-called design feature (DFs) as columns. The ECs were taken from the house of quality whereas the DFs were specified by the expert team. The correlation between ECs and DFs are determined using a 3 to +3 scale. Table 1 from Pullman et al. (2002) shows the results. It can be seen that the target harness should have a soft inside fabric, web fleece construction, narrow waist belt, adjustable wide range belt, a non-threaded buckle, a belay loop, four gear loops, no dedicated tie-in-loop; adjustable leg loops with the lowest price and come in five different sizes.

3 Product Design in a Mobile Phone Market In this section, a thorough description of the experiment that the authors conducted will be presented and results will be demonstrated. Eventually, a comparison between the two experiments will be conducted.

3.1 Application of Conjoint Analysis The expert team identified quite a number of key attributes for a mobile phone by the group 50 plus by using various resources mostly from articles, discussions, surveys, and so forth. The matrix was then reduced to nine key attributes with three levels each (see Table 2). An adaptive conjoint analysis conjoint experiment was developed using Sawtooth Software (2002). One-to-one pre-test interviews were used to make sure that the questionnaire is apprehensible and complete. The first section of the questionnaire was to rate the levels’ preference on a seven-point-rating scale. Next the attributes’ relative importance was determined. This information is useful to discard the relatively unpopular attributes from further evaluation besides it supplies information upon which the initial estimates of the respondents’ utilities can be based (Sawtooth Software, 2002). Subsequently, the paired comparison trade-off questions followed in which conjoint tradeoffs are collected. Fourteen paired questions were selected with seven pairs of questions with two attributes and seven with three. At last the calibration

Table 1 Part-deployment matrix for all-purpose climbing harnesses

Designing Products Using QFD and CA: A Comparison 519

520

S. Abu-Assab and D. Baier

Table 2 Attributes and levels for mobile phones in the CA experiment Feature Form (8.62) Size (10.08) Display (9.65) Battery time (12.36) Mobile phone price (13.80) Running costs (13.83) Intelligent functions (10.48) Keyboard (11.73) Additional functions (9.44)

Levels Folding (9.29) Big (11.61) Big normal (6.77) 3 days (49.75) 30 euro with contract (7.15) 25 euro/month 5 free numbers (31.66) Emergency call with pos. localization (19.65) Big (12.74) SMS, voice output, voice command (28.06)

Sliding (7.24) Medium (25.66) Medium normal (14.36) 7 days (5.03) 80 euro without contract (37.22) Prepaid card 15 euro (35.25) Program. emergency number (8.65) Medium (27.40) Voice output (10.49)

Standard (2.04) Small (14.05) Small sens. (21.13) 10 days (44.72) 150 euro without contract (30.07) 10 euro/month 9 ct/min to 5 numbers (3.59) Defined emergency number (10.99) Small (40.14) Voice command (17.56)

concepts section followed. Although this is optional, yet it is very important in scaling the utilities from rating scales to buying intentions. Fifty-four completed questionnaires from elderly people could be used for evaluation. Table 2 shows the results: On average, respondents have higher utilities (i.e., largest value in each row) for folding mobile phone, medium size, medium and normal display, 10 days battery standby time (longest standby time), 80 euro without contract (lowest price when including the running cost), prepaid card 15 euro, emergency call with position localization, medium keyboard, with SMS, voice message and voice command.

3.2 Application of Quality Function Deployment For constructing the so-called house of quality, again, customer needs (CNs) had to be collected. They were generated from seventeen one-to-one interviews (which is according to Griffin & Hauser, 1993 a sufficient number to draw out the majority of relevant product needs). The respondents were randomly selected people over 50 in Cottbus area. During the interviews, respondents were asked to talk about their relation with their mobile phones. So, e.g., they were asked how frequently they use them, whether they send SMS or not, what they feel the advantages of some mobile phones are and how much they have paid for their mobile phones. Three expert team members (again, the expert team was the same for the application of CA and QFD) independently read and analysed the interviews’ transcripts and grouped the statements into CNs from the point of view of respondents. Six primary CNs with one to three secondary CNs were deduced. Then, 30 respondents rated the importance of the secondary and the primary CNs on a six-point-rating scale. After that the secondary CNs were rescaled so that the sum of all secondary CNs was equal to its primary CNs importance (Table 3).

Table 3 Mobile phones for elderly people: the house of quality

Designing Products Using QFD and CA: A Comparison 521

522

S. Abu-Assab and D. Baier

Next, three competitive mobiles phones were rated from the perception of the group 50 plus. The three products: Nokia 6300 (standard form), Nokia E65 (sliding form), and Motorola V8 (folding form) were selected as good examples of the three forms of mobiles that can be used by this target group. Thirty respondents rated these products on a six-point-rating scale, evaluating all mobiles on the same secondary need before moving to the next. Eventually, they rated the likelihood of purchasing each mobile on a ten-point-buying intention scale. The results of the comparison showed that Nokia 6300 lies comfortably in the hand and with a very long battery duration time. However, it was perceived to be the least preferred mobile phone. Regarding Nokia E65, it is considered “cheap” and “comfortable in the hand” with longest battery duration in this experiment. Yet it is considered to be the least “easy to use”, least “easy to call”, and least “easy to read”; whereas, Motorola V8 was clearly perceived to be the most “easy to use”, most “easy to call”, most easy “to hang a call”, and most “easy to read keyboard”. Therefore Motorola V8 is best rated in the survey by the group 50 plus. Then, as in Pullman et al.’s experiment, the expert team identified one to three ECs (engineering characteristics) for each CN and assessed the correlation between the CNs and the ECs on a 5 to C5 scale as well as the change tendencies for each determined ECs. Regarding the strength of the relationship between the ECs pair (i.e., roof of the house of quality), it was considered to have a small number of interdependencies and for this reason it was not demonstrated. The impact of preferences for the ECs was calculated by multiplying each EC correlation value with its CN’s importance and summing over all the CNs. The results show that the features with the greatest impact on the customer preferences are: battery capacity (27.4), power consumption (26.1), cost (22.9), display brightness (18.0), robust of sending signal (17.4) and number of menu layer (15.27). The analysis indicates that the preferred mobile phone for the target group 50 plus should have long battery duration, good display brightness with no reflexion, minimum cost besides a robust sending signal and a minimum number of layer menus as possible to make it easier and more comprehensive to use. Finally, the parts deployment matrix was constructed (see Table 4). Using again the ECs as rows, the design features (DFs) make now the columns of the matrix. These features were specified and rated on a 3 to C3 scale by the expert team which were related to each EC and eight features from CA. The correlation between ECs and DFs were compared in the same way as in constructing the house of quality between the CNs and ECs. The features with the highest impact on meeting customer requirements are mobile phone without SMS/MMS (92.04), medium sensitive display (92.02), 10-days (long time battery) battery standby (90.46), high robust of sending signal (87.71), not very bright (dark) display (81.59), 80 euro without contract mobile (66.10), with few number of menu layers (46.84), medium keyboard (32.89), with defined emergency number (27.44) and small volume mobile (1.29). Part deployment analysis shows that the target mobile phone should have no SMS function, with a medium sensitive display, long standby battery, with high sending signal and a not very bright display, low cost mobile (with low running cost), simple

Table 4 Mobile phones for elderly people: parts deployment matrix

Designing Products Using QFD and CA: A Comparison 523

524

S. Abu-Assab and D. Baier

menu layers, medium keyboard with a defined emergency number and a small size mobile.

3.3 Comparing the CA and QFD Results In both approaches the optimal mobile phone had the longest battery standby duration (10 days), economical price (80 euros in this case) and a medium keyboard. Whereas, QFD and CA differed somehow in respect to the three optimal levels: display, intelligent functions (i.e., emergency), additional functions (i.e., SMS) and mobile size. Cost (price) ranked first in importance in CA whereas by QDF cost (price) was not considered so important yet in both approaches the same economical price was yielded. In the CA approach, the attribute “additional functions” was not considered so important but “SMS” was the most important level. On the contrary, QFD’s most significant design feature was the absence of the SMS function which could be accounted for the strength of the negative interdependency of SMS function with many other important ECs. The attribute “mobile’s size” yielded a big difference in both approaches. With QFD, it was estimated to be the least important design feature whereas with CA it was weighted to be rather significant. These divergent results may have occurred because of the different basic conception behind each approach: QFD is optimizing DFs for production based mainly on the perception of the expert team whereas CA is optimizing the attributes based on the perception of customers’ needs. These customers’ perceptions sometimes contradict with design features (e.g., customers want mobile to be relatively small with relatively large keyboard or/and with a large display), thus creating a big challenge for the design and production team.

3.4 Comparing the Results with Pullman et al.’s Experiment One can agree with Pullman et al. that QFD and CA optimize results according to their own criteria and the amount of these differences (see Table 5) implies that the two methods are optimizing rather different functions. The main focus of the expert team was to figure out the estimations of the most important features. Therefore, here, we compare the relative importance of common features for the mobile phones for the target group 50 plus (see Table 5) and we also show the result of the comparison that Pullman’s et al. conducted. In both experiments, QFD importance was measured by the feature’s contribution to overall need satisfaction. Regarding the conjoint analysis, attributes’ importance were calculated only in the traditional way, in which each attribute’s utility was calculated as the average difference in the importance of its most and least preferred levels. Although Pullman et al. measured the importance of CA in three ways, yet in this paper it was only measured and compared in the traditional way because of lack of information when running the experiment.

Designing Products Using QFD and CA: A Comparison

525

Table 5 Comparison of design feature importance in Pullman et al.’s experiment (left table) and the mobile phone experiment (right table) Design feature Harness construction Price ($50–$72) Number of gear loops Belay loop Type of buckle Dedicated tie-in loop Type of leg loops Waist belt width 33.2

QFD 34:6 43 58:9 37:1 100 0:3 3 17

CA 9:3 49:8 100 65:7 59:9 36 61:8

Design feature Volume Display size Battery standby Price Emergency call Keyboard SMS

QFD 1:40 99:98 98:28 71:82 7:89 35:73 100

CA 42:03 37:57 100 71:03 32:43 71:49 48:27

The correlations between the CA and QFD utilities weights from Table 5 are 0.319 for Pullman et al.’s experiment and 0.390 for the mobile phone experiment. Both correlation indicate a fairly weak correlation between CA and QFD importance measures with a slightly better correlation in the new mobile phone experiment.

4 Conclusions and Outlook In their conclusion, Pullman et al. were restricted not to generalize their results out of one study and recommended further research. Therefore, this paper is a contribution in this research field. Obviously, the results of the two approaches share some common recommendations as well as they differentiate in some aspects. Yet these deviations between the two approaches are logically explainable since CA reflects the customer wishes and desires whereas QFD represents the engineering/management view of what the customer needs in addition to the fact that the two considered research products differs, too, in their complexity. The employment of attributes from conjoint study to design features through engineering characteristic fosters innovative solutions and new deployment of the design process (Pullman et al., 2002). At the end, it is clear that implementing the two methods together is recommended to get a more accurate and thoroughly information. In addition, the integration of the two approaches make use of the advantages of each and gaps many of their weaknesses.

References Baier, D. (1998). Conjointanalytische L¨osungsans¨atze zur Parametrisierung des house of quality. In: VDI-GSP (Ed.), QFD – Produkte und Dienstleistungen marktgerecht gestalten. D¨usseldorf: VDI-Verlag. Baier, D., & Brusch, M. (2005). Linking quality function deployment and conjoint analysis for new product design. Studies in Classification, Data Analysis, and Knowledge Organisation, 29, 189–198. Griffin, A., & Hauser, J. R. (1993). The voice of the customer. Marketing Science, 12, 1–27.

526

S. Abu-Assab and D. Baier

Katz, G. M. (2004). Practitioner note: A response to Pullman et al.’s (2002): Comparison of quality function deployment versus conjoint analysis. The Journal of Innovation Management, 21, 61–63. Kazemzadeh, R. B., Behzadian, M., Aghdasi, M., & Albadvi, A. (2009). Integration of marketing research techniques into house of quality and product family design. The International Journal of Advanced Manufacturing Technology, 41(9–10), 1019–1033. Pullman, M. E., Moore, W. L., & Wardell, D. G. (2002). A comparison of quality function deployment and conjoint analysis in new product design. Journal of Product Innovation Management, 19(5), 354–364. Sawtooth Software. (2002). ACA system adaptive conjoint analysis version 2.0.1. Evanston, IL: Author. Sullivan, L. P. (1986). Qualtiy function deployment. Quality Progress, 19(6), 39–50. Terninko, J. (1997). Step-by-step QFD : Customer-driven product design. Boca Raton, FL: St. Lucie Press.