Maximizing Revisiting and Purchasing: A Clickstream-Based Approach to Enhance Individual-Level Customer Conversion

Abstract

Online retailers are increasingly mindful of maintaining a long-term relationship with customers, rather than simply focusing on one-time purchases. It is thus critical to ensure that customers continue revisiting a retailer’s website to explore its products and subsequently make purchases. The objective of our paper is, therefore, to help retailers maximize the likelihoods of customers’ revisiting and purchasing, which we achieve at the individual customer level using clickstream data. First, we build a hurdle model that better predicts repeat customer visits when compared to existing methods. This model incorporates heterogeneity in individual customer search behavior, distilling her search history into four theory-driven sets of antecedents: (a) choice, (b) information, (c) pricing, and the (d) search environment. Second, we use a stochastic model to predict customer revisits by incorporating customer heterogeneity in search effort exerted during prior visit sessions. Third, and most importantly, using computationally efficient simulation-based prescriptive analytics, we adapt our modeling approach to propose intervention strategies that maximize the individual-level joint likelihoods of revisiting and purchasing. We find that, for some customers, the antecedents of search operate in opposing directions in their role in revisiting and purchasing, thus necessitating a balancing act between these two key objectives. The intervention strategies render retailers’ insights on personalizing the customer experience based on each individual customer’s own search history and offering each customer relevant personalized content to maximize her revisiting and purchasing.

Keywords: click-stream data, conversion, probabilistic modeling, prescriptive analytics, consideration set, online reviews