User Engagement Maximization Through Preference Switching

Abstract

User engagement is essential for the sustainable growth of many online content-based platforms. These platforms depend on high-quality content, which is often limited due to high costs. Additionally, users may only find a subset of the available content interesting, presenting significant operational challenges for maintaining engagement. Platforms can strategically encourage users to experience new products, potentially increasing the time they spend exploring available options. However, this approach carries the risk of short-term churn due to preference switching. We propose a new modeling framework to capture this trade-off, and design a scalable policy that is asymptotically optimal, with provably fast convergence rates, as the supply of each content type scales to infinity. In numerical experiments, we demonstrate that our policy performs well even with a moderate content inventory. Our policy features a threshold structure to guide the timing of preference switching and a persuasion sequence to determine which types of products to show to users at any given time. Our findings indicate that preference switching can significantly enhance both user welfare and the platform's revenue. Furthermore, our near-optimal policy reveals that the platform should initiate preference switching later for users having higher willingness to explore or having higher churn risk associated with trying new products.

Keywords: customer disengagement, recommendation, engagement maximization, preference switching