An Empirical Two-Stage Choice Model with Varying Decision Rules Applied to Internet Clickstream Data
Published 11/1/2006
Author: Wendy W. Moe
View this content
Executive Summary
In this article, the author presents an empirical two-stage choice model and applies it to observed choice data for both stages. This model possesses several qualities that have not previously been addressed by other two-stage models. First, the model presented in this article incorporates interdependencies between Stage 1 choices through attribute dispersion measures similar to those used in assortment choice models. Second, a set of parameters is introduced that captures the use of decision rules in each stage. These parameters are referred to as “attribute criterion probabilities.” These parameter estimates indicate whether a particular attribute is considered in a given stage of the decision process. The results show that allowing for decision rules and heterogeneity in decision rules provides a better model than simply allowing for preference heterogeneity, suggesting that much of the variance in choice behavior that is observed can be attributed to decision rules and decision rule heterogeneity rather than preference heterogeneity.
The model results provide empirical evidence that is consistent with the idea that consumers use simplified decision rules in early stages of the decision process. The author finds that, on average, fewer attributes are used as criterion attributes in the Stage 1 than in Stage 2 of the decision process. There are also some notable results for specific attributes. For example, price and size tend to be used only in Stage 1 of the decision process. Conversely, ingredient attributes play a role in both stages. The flavor attribute, which is widely considered a variety-seeking attribute, is not considered in the early stage of the decision process but rather plays a role in the purchase stage, suggesting that flavor is not a strict constraint the way size is.
In addition, the empirical two-stage model provides attribute-specific information that lends itself to notable managerial implications. As with other choice models, the model presented in this article can summarize attribute preferences to allow for better promotional design and targeting. From the attribute criterion probabilities in each stage, it is also possible to identify attributes as screening attributes, purchase attributes, or both. This type of information can further enhance the targeting effort by identifying the decision stage most easily influenced by specific attribute information. In addition, the dispersion coefficients also provide managerially actionable information. An obvious implication is for assortment displays. Because the dispersion coefficient indicates consumers’ assortment preferences in the search stage, retail sites can cater their assortment display to customer preferences. However, there is also an argument to be made that dispersion preferences in the search stage indicates some degree of attribute preference uncertainty and learning. If this is the case, retail sites can provide other informational content for attributes identified by the model as having significant preference uncertainty.
Biography
Wendy Moe is Assistant Professor of Marketing in the Robert H. Smith School of Business at the University of Maryland. She has her PhD from the Wharton School of Business at the University of Pennsylvania. Her research interests lie in modeling online consumer shopping behavior and early sales forecasting. Some of her recent work focuses on developing statistical methods and models for Internet clickstream data. Professor Moe has also developed several early forecasting models that can predict the sales of entertainment products early in their lifecycles and, in some cases, even before the actual launch of the product. Her research has appeared in Marketing Science, Journal of Marketing Research, Management Science, Journal of Interactive Marketing, Journal of Consumer Psychology, and Journal of Public Policy & Marketing. Her research has won several awards, including Best Paper in the Journal of Interactive Marketing, the Marketing Science Institute Alden G. Clayton Dissertation Proposal Competition, and honorable mention for the American Marketing Association Howard Award. Professor Moe has also been recognized the by Marketing Science Institute in 2005 as one of the leading young scholars in the field of marketing. Professor Moe has extensive industry experience. She is currently the research advisor for Netconversions, an online marketing research firm (now part of Atlas DMT) that she helped establish. She has consulted for several firms, developing state-of-the-art statistical models to forecast sales. Before her academic career, Professor Moe had extensive experience in the consumer packaged goods industry, working at ACNielsen on pricing and promotional response models.
J Marketing Research, Volume 43, Number 4, November 2006
View Table of Contents.