Tailored recommendations on a matching platform

Matching platforms not only mediate matches but also work as information gatekeepers. When users with private tastes use such a platform to find a partner, the platform asks them to provide match-relevant information; subsequently, it aggregates and distributes the collected data back to each user to facilitate the effective coordination of matches. This study aims to examine how such a platform can design its information flow to make users form matches in a way that is desirable for the platform. I characterize a form of two-way communication that employs both verifiable and non-verifiable messages; then, I delineate the conditions under which a platform can (cannot) achieve its ex-post optimal matching outcome. On a platform that achieves such an outcome, users would fully reveal their private tastes, but the platform would return personalized and only filtered information back to each user in the form of a “recommendation.” I identify three key factors that enable such communication, namely (1) the distance between the distribution of tastes of each side; (2) the uncertainty measure of each distribution; and (3) the population size. As applications, I first study the markets with costly verifiable information and propose a sufficient condition that achieves the optimal matching outcome. Then, I study a two-way communication protocol with non-verifiable messages and demonstrate that communication strictly improves efficiency under any circumstances.