Study Before Play: Pairing Educational and Gamified Content to Align Usage and Purchase Decisions 

with Paulo Albuquerque, under review at Marketing Science


The purchase and usage of products and services are frequently not driven by the same motives, in part because in many categories the user is not the buyer. In this paper, we study usage and purchase decisions in the context of an online educational platform for children, where the completion of math lessons is rewarded with access to gamified activities. To explain usage, we develop a multiple discrete-continuous time allocation model that accounts for the existence of a conditional activity, i.e., when the completion of one activity – a lesson – allows access to another activity – a game. The child’s usage decisions in turn influence their parent’s subscription decision. We estimate the model on data from two online field experiments involving more than 21,000 pairs of children and parents. In counterfactual simulations, we show that alternative customized product configurations, such as increased time of core math content and limiting access to gamified content, can reduce the misalignment between parent and child preferences, leading to increases in subscriptions of around 10%, without a significant decrease in child participation

Do Basket Recommendations Lead Consumers to Save Time, Buy More, and/or Buy Better Products? 

with Paulo Albuquerque, Andres Elberg and Raluca Ursu


Because shopping for groceries can be time consuming, online grocery retailers recommend consumers a basket of products. The conventional approach to designing these recommendations is predicting the set of products a consumer would be likely to buy on a given visit  based on their past purchase history. However, we argue that recommendations themselves have a causal effect on consumer search and purchase decisions. Specifically, by reducing the effort and time to search in recommended categories, recommendations free up the time that consumers can then spend either leaving the retailer web-site early, or searching for non-recommended products. We test our predictions using the data on consumer search and purchase decisions from a large online grocery retailer and we find patterns in consumers’ shopping behaviour consistent with our hypothesis: customers who make smaller orders spend time saved by recommendations searching for non-recommended categories, while consumers making larger orders just finish their shopping sooner. To quantify the effect of recommendations, we develop a structural model of consumer search over multiple categories in which the time constraint is captured through the assumption on search costs increasing over time. The specification that we propose has a simple closed-form expression for the likelihood function,  facilitating application in online settings through fast recovery of the underlying preference and search cost parameters. 


Less Haste, More Speed: How to Improve Learning by Decreasing the Usage Intensity

with Paulo Albuquerque

Leveraging Social Interactions to Provide Relative Performance Feedback: Application to Online Chess

Adaptive Task Recommendation System for a Multi-Subject Educational Platform

with Daria Dzyabura