We study optimal experimentation by a monopolistic platform in a two-sided market. The platform provider is uncertain about the strength of the externality each side is exerting on the other. Setting participation fees on both sides, it gradually learns about these externalities by observing actual participation levels. This provides an informational rationale for introductory pricing, with the platform provider charging a fee below the myopically optimal level on at least one side of the market. If the externality that the other side exerts is sufficiently well known and weaker than the externality it experiences, the platform provider extracts surplus from that side by charging it a fee above the myopically optimal level. This interplay between learning and surplus extraction is crucial to the market outcome and its dynamics.
two-sided market network effects monopoly experimentation Bayesian learning optimal control