Data-Driven Dynamic Assortment in Online Platforms: Learning about Two Sides

2026-06-09

AI Summary

This study addresses the dynamic assortment selection problem for online two-sided platforms under complete uncertainty regarding both customer and seller preferences. Within a discrete-time, heterogeneous-agent framework with incomplete information, the platform periodically displays sets of sellers, and matches are formed according to a multinomial logit model on both sides. We propose the first online algorithm that jointly learns preferences from both sides while simultaneously optimizing long-term revenue, thereby tackling the challenging setting where preference parameters on both sides are unknown. Theoretical analysis shows that the algorithm achieves worst-case regret growing at a polylogarithmic rate, which matches the established lower bound and thus attains the optimal theoretical rate.

Abstract

We study a dynamic assortment problem on a two-sided service platform with incomplete information and heterogeneous customers in a discrete-time setting. In each period, a customer arrives seeking service, and the platform chooses an assortment of sellers to display. The customer then proposes a transaction to at most one seller in the assortment according to a multinomial logit choice model. After a fixed number of periods, sellers review the proposals they have received and each chooses at most one customer according to another multinomial logit choice model, after which the cycle repeats. A key challenge is that the platform does not know the choice-model parameters of either customers or sellers in advance. To our knowledge, this is the first study of a dynamic assortment problem in which both sides' choice parameters are unknown. We develop a data-driven algorithm that learns these parameters while optimizing the platform's objective over time. We evaluate performance using regret, which measures revenue loss relative to a clairvoyant benchmark that knows all parameters and customer arrivals in advance. We show that the algorithm's worst-case regret grows polylogarithmically over time, and we derive a matching lower bound, establishing its rate optimality.

Problem

• dynamic assortment
• two-sided platform
• incomplete information
• multinomial logit choice model
• regret minimization

Innovation

• dynamic assortment
• two-sided platform
• multinomial logit choice model
• online learning
• regret minimization