Trading Flexibility for Adoption: Dynamic Versus Static Walking in Ridesharing

Abstract

On-demand ridesharing aims to fulfill riders' transportation needs whenever and wherever they want. Although this service level is appealing for riders, overall system efficiency can improve substantially if riders are willing to be flexible. Here, we explore riders' flexibility in space via walking to more accessible pickup locations. Ridesharing platforms have traditionally implemented dynamic walking, which jointly optimizes rider-driver assignment with rider pickup locations. We propose the new paradigm of static walking, which presents a predetermined pickup location to the rider, and then optimizes rider-driver assignment. Conventional wisdom would make dynamic walking appear to be the gold standard, but we find that static walking (i) is surprisingly competitive with dynamic walking in congested urban environments, and (ii) can promote adoption from riders, thereby surpassing the dynamic approach. To demonstrate (i), we introduce a static walking optimization algorithm and evaluate it with simulations and a large-scale empirical study of Lyft rides in Manhattan. We show that static walking achieves up to 96% of the value of dynamic walking, and we explain why and when this is the case. To support (ii), we share a large-scale observational analysis of Lyft riders' walking behavior and analyze a randomized control trial on the Lyft user interface, which suggests that riders are sensitive to the design of the walking product and that certainty in pickup location can drive walking adoption.

Keywords: ridesharing, ride-hailing, walking, matching, optimization, experiment