Abstract
We study delay information disclosure policies in on-demand platforms, modeled as two-sided queues, with two user classes --- customers and providers --- who seek matches to each other using the platform. The primary objective is maximizing the rate at which these matches occur by adopting one of three information regimes: the occupancy regime (disclosing the current system occupancy to both user categories) or two distinct asymmetric regimes (sharing no information with one user class while sharing occupancy information with the other). The users of each class are strategic and decide whether to join based on the available delay information. We use continuous-time Markov chains to model the system as a two-sided queue and employ equilibrium analysis to characterize the users' joining behavior and the platform's match rate under each information regime. The optimal policy reveals a complex dependence on system parameters and is strongly influenced by the users' patience profiles. We demonstrate analytically that it is strategically advantageous to withhold delay information from one user class, especially when it consists of a substantial number of relatively delay-insensitive users. The optimality of the asymmetric information-sharing regimes becomes more prevalent as the discrepancy in the patience profiles or the market sizes of the two user classes increases. However, our extensive numerical analyses find that the occupancy regime proves to be optimal in many other settings. In cases where it falls short of optimality, the sub-optimality gap is usually minimal (on average, about 5%). Our findings hold crucial implications for platform managers, indicating that in such two-sided systems, the occupancy regime is a safe choice unless the two user classes exhibit a large patience profile discrepancy or a large market size imbalance. In such cases, opting for the occupancy information regime could adversely impact the platform's match rate. In such situations, carefully evaluating the chosen information-sharing strategy becomes imperative.

This publication has 26 references indexed in Scilit: