Selling with Recommender Systems and Price Dynamics

Abstract

A long-lived seller sells a product of unknown value through a recommender system that offers prices and recommends products to short-lived consumers in continuous time. The seller receives feedback about the product at a rate that increases with the instantaneous sales volume. The optimal selling mechanism is characterized by information cutoffs and features episodes of price discounts and overrecommendation. The discount periods begin or end when the seller's belief about the product value hits the information cutoffs. During each discount period, the seller lowers the price and selectively sends out unwarranted recommendations to consumers. In this market, restrictions on either prices or recommendation strategies lead to Pareto-inferior outcomes.

Keywords: Dynamic pricing, information design, social learning.

JEL Classification: D82, D83, L12