Order dispatching strategy and pricing scheme in ride-sourcing markets with consideration of service cancellation

Abstract

In a ride-sourcing system, dispatching order requests to available drivers entails a comprehensive consideration of factors such as pickup proximity, order rewards, driver rating, safety behavior, passenger preferences, real-time road conditions, and other relevant variables. Inefficient dispatch processes often result in service cancellation by either the customer or the driver. This paper represents a pioneering effort to examine order dispatching strategy and pricing scheme while taking service cancellation behaviors into account. By assuming the platform has limited knowledge of the valuation of service of each customer and the reservation earning rate of each driver, we develop a two-period model that captures the dynamic decision-making processes of multiple stakeholders (customers, drivers, and platform) and formulate the platform’s order-dispatching problem as a stochastic programming model. Within a greedy approximation framework, our analysis reveals the significant implications of pricing scheme for critical performance metrics while considering service cancellation. These include the matching probability (probability of customer-driver acceptance for platform’s match results), the platform’s rewards, and the effects on the platform’s order-dispatching decisions. Specifically, within the realm of linear pricing, the matching probability demonstrates a positive correlation with trip distance, and thereby establishes a consistent dispatching order compared with one that does not consider service cancellation. Conversely, with nonlinear pricing (whether sublinear or superlinear), extended trip distance is generally associated with a reduced matching probability when it exceeds a threshold; this results in prioritizing orders with intermediate trip distances in order-dispatching decisions. Moreover, numerical experiments support that an integration of sublinear, superlinear, and linear pricing is conducive to optimizing rewards across short-, intermediate, and long-distance trips. Finally, scenarios of unimodal distributions of customer’s valuation of service and driver’s reservation earning rate consistently yield the highest rewards, through sublinear, linear, and superlinear pricing schemes.