Preference elicitation in B2C e-commerce

Making preference information in consumer markets explicit can benefit both the individual customer and the business side. The costumer can benefit by faster finding a product that fits his needs whereas the e-commerce site gets important information about segments and needs in the market. Preference information can be obtained in an e-commerce environment either from historical transaction data or actively from webbased questionnaires. Here we address the active assessment approach. We compare two methods to compute consumer’s preferences from stated choice data. The data are obtained from choice experiments, where each respondent is confronted with a small set of products (or descriptions of products) from which he has to choose the one he is most likely to buy. We compare a method to analyze stated choice data that we developed ourselves with the industrial ‘gold standard’, namely, the hierarchical Bayesian method for conjoint analysis from Sawtooth Software, Inc. The two main differences between the two methods are: First, we do not exploit / need a conjoint structure underlying the data. Second, in our method we first segment the respondents into classes and afterwards estimate the preferences for each class whereas the standard way is to use the computed individual preferences to segment the respondents.