(De)Pooling Eases Spatial Mismatch

Abstract

Spatial mismatch, i.e., misalignment between where supply and demand arise, is a common source of inefficiency in matching applications such as ride-hailing, bike/scooter sharing, and e-commerce fulfillment. In ride-hailing, this misalignment leads to the "wild goose chase," where drivers travel long distances to reach riders. This paper investigates how spatial mismatch can be mitigated through pooling (i.e., aggregating demand spatially or temporally) and depooling (i.e., restricting matches to be more local). We consider a stylized model of a 2D disk-shaped service region with spatially random arrivals of riders and drivers. Drivers are short-lived and leave if unmatched, while riders can wait in a virtual queue. We compare five operational policies: (1) FCFS ride-hailing (baseline), (2) spatial pooling, (3) temporal pooling, (4) zoning, and (5) localization.  We evaluate each policy based on the rider's expected total time in the system and the driver's matching rate. Our analysis reveals that pooling strategies are more effective under high system load, while depooling strategies perform better when the system is lightly loaded. Spatial pooling improves the rider's total turnaround time when riders can walk fast enough, or not fast but the system is highly loaded. Temporal pooling outperforms ride-hailing when it waits for better matches without excessive delay, which means it may turn away an available driver but still achieves an average shorter pick-up distance, demonstrating a little temporal flexibility goes a long way. Zoning and localization offer improvements in low-load settings by reducing pick-up distances but must be carefully tuned to avoid excessive fragmentation or loss of supply by optimizing the number of zones or the proximity of localization. We extend our model to one- and three-dimensional settings and find that the effectiveness of pooling and depooling strategies in mitigating spatial mismatch varies with dimensionality, offering new insights into spatial management in diverse applications, from linear cities and urban corridors (1D) to automated warehouses and aerial fulfillment systems (3D). Finally, we discuss how the insights we obtain from the ride-hailing setting can be applicable to mitigate the spatial mismatch in other settings, such as free-floating vs. dock-based bike sharing. 

Keywords: spatial pooling, temporal pooling, ride-hailing, wild goose chase, zoning

JEL Classification: C78, D47