Paper Non-Exclusive Notifications for Ride-Hailing at Lyft II: Simulations and Marketplace Analysis

Uncertain driver acceptance poses a fundamental challenge for ride-hailing platforms. Traditional one-to-one dispatch forces sequential retries when drivers reject or fail to respond, lengthening rider wait times and creating marketplace friction. Non-exclusive dispatch — broadcasting a trip request to multiple drivers simultaneously — addresses this inefficiency but introduces its own design challenges, particularly around how to select which drivers to notify and how to resolve situations where multiple drivers accept the same request. This study, the second in a two-part collaboration with Lyft, develops a theoretically grounded framework for evaluating the long-run performance and marketplace-wide effects of transitioning from exclusive to non-exclusive dispatch. The analysis integrates three complementary approaches: a constrained welfare maximization model, large-scale discrete-event simulations on proprietary Lyft data, and a stylized macroscopic equilibrium model. Across both simulation and equilibrium analysis, non-exclusive dispatch consistently improves key fulfillment outcomes relative to exclusive dispatch — reducing match times, decreasing rider abandonment, and increasing both the volume and average quality of completed trips. The study also quantifies the speed-quality trade-off between two common contention-resolution rules: First Accept maximizes throughput and speed, while Best Accept is necessary to optimize per-match quality. A further finding concerns notification set design: moderately conservative notification strategies — those that avoid notifying too many drivers at once — can improve long-run efficiency by preventing excessive commitment of high-value drivers to any single request, thereby preserving their availability for future matches. Together, these results offer a principled foundation for platform design decisions around non-exclusive dispatch.

Get the Paper