Abstract: 

"Online media platforms rely on long-term user engagement to generate revenue. The primary operational lever they control is content recommendation, which determines what content should be suggested to each user. However, due to the constantly evolving supply of content and the heterogeneity in user behavior, achieving optimal recommendations is challenging. It necessitates a delicate balance between experimentation (to gauge the effectiveness of new content) and exploitation (recommending high-quality existing content).

Motivated by a real-world dataset, we present a comprehensive model for the platform recommendation problem with the aim of maximizing long-term user engagement. Our model captures two key features of this problem: (1) supply-side experimentation, where the platform tests new content of uncertain quality, and (2) demand-side heterogeneous churning, where different users churn at different rates as a function of their engagement history (state). We use our model to study the interplay between experimentation and churn management to maximize long-term user engagement. Our analysis highlights the importance of state-specific experimentation. As such, we propose a simple myopic policy we term "churn minimization" (CM), and study its optimality (analytically and numerically) on a wide range of models."

Short Bio: 

Michael Hamilton is an Assistant Professor of Business Analytics and Operations at the University of Pittsburgh, Katz Graduate School of Business.  He is broadly interested in problems related to pricing, prescriptive analytics, and market design. He received his Ph.D. in Operations Research from Columbia IEOR in 2019. For more information and papers please see: https://mhamilton-pitt.github.io/