Int. J. Electronic Business, Vol. 2, No. 5, 2004
Web-based knowledge management in product concept development: the DELI approach Martin Natter, Andreas Mild, Alfred Taudes* and Christian Geberth Abteilung für Produktionsmanagement, Wirtschaftsuniversität Wien, Pappenheimg. 35/3/5, A-1210 Vienna, Austria E-mail:
[email protected] E-mail:
[email protected] E-mail:
[email protected] E-mail:
[email protected] *Corresponding author Abstract: In this paper, we describe DELI (‘Development Interactive’), a method and software system that supports the product concept development phase by multi-functional product development teams using Marketing Engineering methods. After motivating our research, we present core concepts underlying DELI: a data model encompassing customer and product attributes and preferences as well as a questionnaire generating tool based thereon, a novel integrated clustering and attribute/product positioning algorithm for market map generation and a conjoint-based market simulation component. All concepts are explained using a real-world mobile phone case study. Keywords: distributed product development; knowledge networks; data mining; marketing engineering; e-business potential. Reference to this paper should be made as follows: Natter, M., Mild, A., Taudes, A. and Geberth, C. (2004) ‘Web-based knowledge management in product concept development: the DELI approach’, Int. J. Electronic Business, Vol. 2, No. 5, pp.471–479. Biographical notes: Dr. M. Natter studied MIS at the University of Vienna, Austria. From 1990 to 1993, he was Project Assistant at the Institute for Advanced Studies in Vienna, since then he has been working at the WU’s Department for Production Management, at first as Assistant Professor and since 1999 as Associate Professor. Dr. A. Mild studied business administration at the WU. Since 1998, he has been working at the WU’s Department for Production Management, at first as Assistant Professor and since 2003 as Associate Professor. Dr. A. Taudes studied business administration and management information systems (MIS) in Vienna (Doctorate 1984), in 1991 he received his PhD from the Vienna University of Economics and Business Administration (WU). He was Assistant Professor at the WU (1986–1991) and Professor for MIS at the German Universities of Augsburg (1991), Münster (1991/92) and Essen (1992/93). Since 1993, he has been Professor for MIS at the WU and Head of the Department for Production Management. Since 2000, Taudes has been Speaker for the Special Research Area SFB #010 (‘Adaptive Information Systems and Modelling in Economics and Management Science’). Copyright © 2004 Inderscience Enterprises Ltd.
471
472
M. Natter, A. Mild, A. Taudes and C. Geberth Ch. Geberth, MSc, studied economics and business administration at the WU. From 1998 to 2002, he worked as a Process Consultant. Since 2002, he is the CEO of the Prolytic GmbH, a spin-off of the WU’s Department for Production Management, focusing on consulting in new product development and marketing engineering. Since 2003, he is Project Manager for the xCom Group at EC3 and Research Assistant within the Special Research Area SFB #010.
1
New product development and web-based marketing engineering
A product concept is a description of a product in accordance with attributes perceived by the target customers. It is developed in the early stages of a new product development (NPD) process and serves as basis for a technical specification (product engineering) using methods such as, e.g., the ‘house of quality’ [1]. If conceptual marketing is used for concept development, then a number of product ideas will be evaluated on the basis of tacit knowledge gained only through market involvement. This approach has a number of drawbacks. Just to name a few: dependence on few market experts, inter-functional conflicts due to lack of understanding of other departments’ motivations and constraints as well as expensive trial and error product launches. Nevertheless, this approach is economical with respect to data collection, processing and communication (see [2]). Now, the World Wide Web has dramatically increased the efficiency and effectiveness of obtaining and sharing customer preference data: the web provides fast and cheap ways of conducting internet surveys and tracks customer behaviour via log files, past purchases over several channels, trouble ticket systems, customer community sites, etc. Also, preference information can be readily shared between product development team members from different functional areas and/or other stakeholders such as suppliers on an intra/extranet basis. Thus, the potential to use data-driven, engineering methods to develop product concepts has increased. ‘Marketing engineering’ is the process of collecting structured market data via questionnaires, segmentation and positioning and revenue simulation via data mining algorithms (see [3]). If this were done by traditional statistical methodology, an additional NPD interface: between models, parameters, test results and market knowledge, and vice versa, due to the need for direct market experience when developing questionnaires, interpreting results and selecting models would be necessary. Thus, DELI has been designed as a generic tool that automates statistical decision-making and graphically represents market structure (so-called market maps). Such tools can be used by the NPD team directly, i.e. it can serve as basis of joint discussions, which merge tacit and data-driven knowledge and represent the basis of interactive experiments with different product concepts. Supported by blackboard and chat systems, the search and evaluation process can be executed in a distributed web-based environment and the results of this joint learning and evaluation process can then remain a part of the organisational memory. Therefore, the DELI project naturally fits into ec3’s xCommerce and knowledge management projects: DELI is currently being used for the market model component, when evaluating xCommerce models. In addition, the market maps and simulation results generated can be integrated with other sources of knowledge on the basis of results described in [4,5] to build a market knowledge management system mentioned above.
Web-based knowledge management in product concept development
2
473
Existing methods and tools
Various statistical techniques are available to support the marketing engineering process: in the first step, customers are segmented, then a MDS solution is computed based on distances between product and attribute ratings/rankings and finally, the consumers are positioned in the map according to their individual preferences for all assessed products and attributes. Different clustering methods for market segmentation and multi-dimensional scaling are available in standard statistical packages like SPSS or Clementine and various software packages supporting conjoint analysis are on the market (see, e.g. [6,7]). However, for our purposes, the availability of statistical algorithms is not sufficient: model selection, integration of results and graphical representation of market structure remain to be done manually – tasks that can be automated and must be automated, if marketing engineering is to be done by the NPD team itself. Consider, for instance, the consecutive application of clustering and MDS: since the MDS solution is not conditional on the segment solution derived in the first step, customer positions and segment memberships do not match. Therefore, a joint graphical representation produced by overlying the outputs of the two steps is not useful, but rather a simultaneous approach is necessary. Various simultaneous approaches have been described in the literature: Hruschka [8], for instance, proposes to use fuzzy clustering methods, Wedel and Steenkamp [9] suggesting a clusterwise regression method. Furthermore, methods based on ‘latent class multidimensional scaling’ (LCMDS) techniques [10–12], have attracted considerable attention in both marketing research and practice.
3
The DELI approach
3.1 Overview DELI [13] employs the following philosophy to fulfil the promise of automatic model selection, graphical representation of market structure and interactive concept evaluation: •
Preference data gained through an internet questionnaire is used for market segmentation and product/attribute positioning, which is done via an integrated and iterative algorithm that aims at optimising both, segment membership and preference representation in combination with multi-dimensional scaling (MDS). There, statistical decisions such as the determination of an appropriate number of segments are automated, based on built-in statistical expertise.
•
The resulting graphical ‘market maps’ are further explored through comparisons of histograms regarding socio-demographic, lifestyle and media variables and attitudes toward products, attributes, benefits – supplements in addition to the product and disturbing factors – matters that annoy customers – between different groups (segments, target market vs. all customers, individual customer).
•
Various conjoint models are estimated and automatically selected to determine the success of the product concepts to be explored for a particular target segment selected.
474
M. Natter, A. Mild, A. Taudes and C. Geberth
3.2 The DELI data model DELI is based on a generic (application-independent) data model of data collected through a market survey. It covers products and their attributes, customers characterised by purchase behaviour, socio-demographic (age, gender etc.), lifestyle (e.g., preferred activities such as shopping or clubbing) and media variables, and customers’ preferences for products and attributes. Additionally, attitudes towards benefits and disturbing factors are also covered. This data model provides the semantics necessary for automatic application of mining methods and it is also used to generate a questionnaire for internet, laptop or paper data collection. In addition, DELI automatically generates conjoint designs in which the results of pairwise choices presented to the customers are used to automatically compute conjoint models as basis for evaluating success of product concepts not covered by the survey. In order to allow for processing of data collected by means other than DELI, it also provides an EXCEL interface. In such cases, the data semantics is defined through a data preparation step following data import. Another data processing function allows the exclusion of respondents from further analysis via filters. By showing sample histograms, this function also allows for a simple check of the sample’s representativeness.
3.3 The DELI market map DELI uses the following approach to obtain an integrated graphic market map: First, the k-means clustering algorithm is applied, where the optimal number of segments is determined by using the Davies–Bouldin index [14]. The resulting segment memberships influence the attribute and product positions on the map, which, in turn, determine customer positions. In the next step, an MDS solution is computed using an extended distance matrix including the segment membership vectors. Thereafter, a customer’s position is calculated as a weighted average over all product and attribute positions by using the importance customers place on them as weighting factors. Finally, a new segment solution is derived using an extended segmentation base including the customer positions as additional segmentation criteria. These steps are iterated until convergence is achieved (see also [15,16]). Figure 1 depicts the main market map generated this way for a real-life mobile phone case. In this project, a sample of 738 respondents was collected to refine and check several product ideas for entering the market, as the basis of product engineering and business planning. One particularly interesting product idea was a mobile phone with a competitive SMS price and functionality to better support message handling. The attributes of a cellular phone product (tariff) are the provider, base fee, per-minute prices for calling within the own network, calling into another one and the conventional network, prices for SMS and WAP, etc. In the market map depicted in Figure 1, products are coloured in orange, attributes in magenta and each consumer according to the segment to which he belongs. Only two simple principles are necessary for the interpretation of the graphic market structure thus generated automatically:
Web-based knowledge management in product concept development
475
•
the larger an entry, the more important (attribute)/attractive (product) it is to the entire group
•
the nearer an entry is to a segment, product or attribute, the stronger is the relationship, indicating strength of competition in the case of products and attribute bundles in the case of attributes.
Figure 1
DELI market map for cellular phone case
Figure 1 thus shows that in the present case, there are three segments of approximately equal size in terms of both number of customers and revenues. The most favoured tariff is B, followed by C. Tariff A is relatively similar to B and C to E. The two most important product attributes are the base fee and the price for calls made in one’s own network, A and B having a low base fee and C and the hidden product D a low per-minute fee within one’s own network. B is well received by customers in segments 1 and 2, C and D in Segment 3. So, at first glance, the market appears relatively segmented, and probably there is little chance of penetrating it. On the other hand, when looking at the central part of the map, there are customers to whom the price of an SMS message is very important and who are not satisfied with generally strong products. This supports the product idea of an SMS mobile phone. This hypothesis can be further explored by looking at other DELI market maps where customers are coloured by their favourite product/most important attribute and by inspecting the preferences of individual (‘prototype’) customers.
476
M. Natter, A. Mild, A. Taudes and C. Geberth
Obviously, the next interesting point is how to interpret the dimensions of the main market map. By looking at the factors that contribute most to the explanation of customer’s position calculated via algorithm described above, one notes the following: the more one moves to the right, the higher the importance of per-minute fee elements, the lower the importance of the SMS price and the higher the proportion of women and income. When moving upward, the proportion of women decreases and the income increases, i.e. one may assume that in the target segment of an SMS mobile there are younger men. This assumption is enforced when considering the segments one by one by comparing average preferences for products and attributes and relative frequencies of socio-demographic, media and lifestyle variables per segment with respective overall statistics. Due to the intuitive presentation of the market knowledge contained in the data collected, all these insights can be jointly created in an interactive way and shared between the NPD team members. Data-driven hypotheses can be instantly matched with direct market experiences. The DELI maps can also be part of a blackboard architecture for remote discussions via chat systems and annotated with comments, drawings etc. Based on the created objective and common knowledge-base, the implications for targeting the above-defined segment for product engineering, costing, communication policy and strategy (potential competitors’ reactions) can be formulated by the respective specialists and integrated into the product concept definition. Hence, the success-lowering potential of departmental information hiding and intriguing can be diminished (see [17]).
3.4 Target market definition, concept generation and revenue simulation On the basis of thorough market understanding, a target group can be selected and explored by using the above-described segment-specific analysis for the comparison between target market and the entire market. Based on thorough understanding of the target segment’s preferences resulting from the analysis above, design engineers can develop the product idea into several prototype product designs. This forms the basis of a rough manufacturing plan and of target costing by production engineering and controlling. On this basis, allowed price bands for product concepts can be established so that, in the next DELI step, a synthetic product concept and its potential market success can be defined and analysed. In addition to the existing tariffs A–H, a new tariff is established and for each attribute of this tariff an appropriate value is set. Naturally, to be successful, the SMS mobile will have to score points with SMS message price. Another important criterion of this target group is the possibility to make low-priced calls to other networks and that the cellular phone is convenient in every way. Also, two benefits (a free cinema ticket and a convenient change of the provider) are offered. DELI automatically selects an appropriate estimation model for product choice. Considering linear regression, ridge regression and a logistic regression model, the best model is chosen on the basis of the maximum hit rate achieved in sample. Based on the selected model, one finds that the product concept might reach 14% of the target segment and 12% of the overall market. In addition, Figure 2 shows a competitive comparison of the attributes of the product concept candidate and all other products as compared with preferences in the target segment (red line). It turns out that, as far as the SMS message price is concerned, the position is optimal. As planned, the two most important properties perceived on the total market hold potential for improvement. However, it remains to be
Web-based knowledge management in product concept development
477
seen, if the cost of lowering these tariff components would be feasible! Clearly, this picture is just one round in the best response game, i.e. it is computed on the premises that the positions of other products remain unchanged. Nevertheless, based on knowledge about competitors’ strategies, potential moves can be predicted and simulated by modifying the respective positions too, so that an interactive market simulation can be performed by the NPD team. Figure 2
4
Attributed performance analysis for cellular phone case
Evaluation and further research
Recent empirical studies clearly demonstrate the importance of knowledge and its mobilisation and integration through appropriate organisational mechanisms for NPD success (see, e.g. [18,19]). The web can contribute to this aspect by enlarging the data collecting and communication abilities of NPD teams. However, to ensure a proper transformation in a common knowledge base, appropriate data mining methods are necessary. DELI has been developed to fill this gap by providing easy-to-use and interactive ways of exploring market structure and performing market simulation in order to define and evaluate product concepts. So far, DELI has been applied to energy, retail, banking and car markets. In all cases, a statistically significant and intuitively correct representation of the market has been found. Further developments of DELI will focus on more sophisticated modelling of switching behaviour and the integration of other sources of knowledge mentioned in the introduction.
478
M. Natter, A. Mild, A. Taudes and C. Geberth
Acknowledgement This piece of research was supported by the Austrian Science Foundation (FWF) under grant SFB #010 (‘Adaptive Information Systems and Modelling in Economics and Management Science’).
References 1 2
3 4
5
6 7 8 9
10
11
12
13 14 15
Hauser, J.R. and Clausing, D. (1988) ‘The house of quality’, Harvard Business Review, May–June, pp.63–73. Natter, M., Mild, A., Feurstein, M., Dorffner, G. and Taudes, A. (2001) ‘The effect of incentive schemes and organizational arrangements on the new product development process’, Management Science, Vol. 47, No. 8, pp.1029–1045. Lilien, G. and Rangaswamy, A. (1997) Marketing Engineering: Computer-Assisted Marketing Analysis and Planning, Edison Wesley, New York. Denk, M., Froeschl, K.A., Hrastnik, P., Reinsperger, T. and Urro, R. (2003) ‘KnoWeb – an operational knowledge organization framework’, International Journal of e-Business, this issue. Glawar, F.P., Küng, J., Luckeneder, T., Steiner, K., Tjoa, A.M., Wagner, R.R. and Wöß, W. (2003) ‘Applications of topic maps in knowledge management systems’, International Journal of e-Business, this issue. ACA (2002) Sawtooth Software, Technical Paper Series, ACA 5.0 Technical Paper, Sawtooth Software Inc. CBC (1999) Sawtooth Software, Technical Paper Series, Choice-based Conjoint (CBC) Technical Paper, Sawtooth Software, Inc. Hruschka, H. (1986) ‘Market definition and segmentation using fuzzy clustering methods’, International Journal of Research in Marketing, February, pp.117–134. Wedel, M. and Steenkamp, J-B.E.M. (1991) ‘A clusterwise regression method for simultaneous fuzzy market structuring and benefit segmentation’, Journal of Marketing Research, Vol. 28, pp.385–396. Böckenholt, I. and Gaul, W. (1989) ‘Generalized latent class analysis: a new methodology for market structure analysis’, in Oppitz, O. (Ed.): Conceptual and Numerical Analysis of Data, Springer, Berlin, pp.367–376. DeSarbo, W.S., Howard, D.J. and Jedidi, K. (1991) ‘MULTICLUS: a new method for simultaneously performing multidimensional scaling and cluster analysis’, Psychometrika, Vol. 56, pp.121–136. DeSarbo, W.S., Manrai, A.K. and Manrai, L.A. (1994) ‘Latent class multidimensional scaling: a review of recent developments in the marketing and psychometric literature’, in Bagozzi, R.P. (Ed.): Advanced Method of Marketing Research, Blackwell, Cambridge, pp.190–222. Copyright prolytic – Marketing Engineering, Consulting & Software GmbH (see http://www.prolytic.com). Davies, D.L. and Bouldin, D.W. (1979) ‘A cluster separation measure’, IEEE Trans. Pattern Anal. Machine Intell., Vol. 1, pp.224–227. Natter, M. and Mild, A. (in print) ‘DELI: an interactive new product development tool for the analysis and evaluation of market research data’, Journal of Targeting, Measurement and Analysis for Marketing.
Web-based knowledge management in product concept development 16
479
Reutterer, T. and Natter, M. (2000) ‘Segmentation based competitive analysis with MULTICLUS and topology representing networks’, Computers and Operations Research, Vol. 27, pp.1227–1247. 17 Feurstein, M., Natter, M., Mild, A. and Taudes, A. (2001) ‘Incentives to share knowledge’, Proceedings of the Hawaii International Conference on System Sciences (HICSS 2000), Hawaii. 18 Duta, S., Narasimhan, N. and Rajuv, S. (1999) ‘Success in high-technology markets: is marketing capability critical?’, Marketing Science, Vol. 18, pp.547–568. 19 Verona, G.A. (1999) ‘A resource-based view of product development’, Academy of Management Review, Vol. 24, pp.132–142.
