Modeling User Choice to Compose Offer Sets
by
Sumit Sarkar
Charles and Nancy Davidson Chair
Professor of Information Systems
Director, PhD Programs
Naveen Jindal School of Management
The University of Texas at Dallas
Friday, Nov 12
10 – 11 am | Zoom
Abstact:
Firms are increasingly using clickstream and transactional data to tailor product offerings to visitors (users) at their site. The sites have the opportunity, at each interaction (i.e., whenever a user clicks on a link), to offer multiple items (referred to as an offer set) that might be of interest to the user. By displaying links to multiple items, the site hopes to increase the chance that the user will find at least one of those items to be interesting, thereby increasing the probability the user will make a purchase. We examine the problem of composing an offer set that maximizes the firm’s expected payoff over a user’s session, which would typically include multiple interactions. An offer set provided to a user has important implications: the composition of the offer set impacts the immediate choice of the user (item externality), and the user’s choice in one interaction may influence the subsequent actions the user will take during that session. Hence, in order to determine the full impact of an offer set, it is necessary to consider not only the likelihood of the user purchasing one of the offered items, but to also evaluate how it may impact the choices of a user at subsequent interactions. By providing the right offer set a site can guide a user to paths that lead to items with higher likelihood of conversion while simultaneously learning what other items the user may examine (view) before making a purchase. We show that identifying the optimal offer set is a difficult problem in general, and develop an efficient heuristic that can be used in real time. Simulated experiments based on real data show that the joint consideration of item externalities and of looking ahead lead to both higher conversion rates and longer user sessions on average, and consequently to increased total sales.