Today, more and more companies are providing web-based product configuration systems in order to better meet individual customer preferences. In many cases, pre-defined product specifications are additionally offered to facilitate the corresponding choice decisions. Against this background, we present a Poisson regression approach for analyzing customer preferences and a genetic algorithm for determining preference-based default products. The basis for this is transaction data, as they are automatically generated when configuring a product online. The potentials of the suggested methodology are demonstrated by means of two case studies referring to different product categories.
Key words: Customization, Genetic Algorithm, Poisson Regression, Preference Analysis, Product Configuration, Web Usage Data
(ProQuest: ... denotes formulae omitted.)
The diversity of products offered to meet customer needs has rapidly increased in recent years. Even simple devices with few basic functions, such as mobile music players or phones, nowadays provide a multitude of additional functions like USB interface, games, Bluetooth, etc. This development is accelerated by several factors, particularly the increasing globalization of competition, the availability of new technologies, and more demanding customers . In addition, the Internet allows consumers an easy access to a wide range of information on product varieties, prices, conditions of supply, etc., which facilitates comprehensive overviews of alternative purchase options . The Internet not only offers the opportunity to seek information but also provides product recommendations, which significantly improve the possibilities of identifying those products that adequately satisfy consumers' individual needs. At the same time, many industries are using one-to-one marketing techniques to better keep pace with the growing market dynamics and the increasing differentiation of individual preferences .
Nevertheless, the more products a manufacturer or retailer offers, the harder it is for the consumers to identify those which best meet their individual preferences. Studies have shown that consumers can even become overwhelmed if too many choices are offered and/or if those choices are poorly organized . Particularly two options for supporting customers' purchase decisions are worth a closer examination in online business, namely the provision of passive search tools and that of active search tools . Passive search tools, such as most recommender systems, use the consumers' revealed preferences from past purchases to suggest products that might be interesting in the current decision context. Active search tools, such as product configuration systems, allow customers to successively specify the products that presumptively best meet their individual preferences. The focus of this paper is on the latter.
According to , interactive environments like the World Wide Web are predestined for product and service customization. The net payoff of interactivity to consumers is usually assumed to be positive. Existing web-based configuration systems enable the individual customization of package tours (see, e.g., Site 1), laptops (see, e.g., Site 2), and even bags (see, e.g., Site 3) or sports shoes (see, e.g., Site 4). Car manufacturers, for instance, mostly offer large numbers of vehicle colors, equipment components, financing options, etc. However, retailers who are going to implement a high-variety strategy need to ensure that the customers are not confused by the complexity inherent in a wide range of options .
Recommendations provided by retailers or manufacturers are a confirmed means to concretize unclear customer preferences in the purchase decision process. Senecal and Nantel  p. 159, for example, showed by experimental research that consumers "who consulted product recommendations selected recommended products twice as often as subjects who did not consult recommendations". …