Queueing analysis and waiting management in ride-sourcing markets

Abstract

The rapid development of ride-sourcing services has profoundly impacted people’s lives and reshaped their travel behavior. Attracted by the convenience of these services, increasing demand has surged into the market, presenting a double-edged sword for platforms. While this influx of demand can accelerate growth and boost revenue, it also introduces significant challenges, the most critical being how to manage the resulting supply-demand imbalances. A key consequence of these imbalances is a decline in user experience, which may lead to passenger attrition. This, in turn, can negatively affect platform revenue and threaten the long-term sustainability of the business. To address this problem, we first conduct an empirical analysis of passengers’ waiting behavior using real data from four metropolitan areas in China. In the ride-sourcing market, passenger behavior during the waiting process is influenced by constantly updated delay announcements and the effect of sunk waiting time. However, due to the limited predictive capabilities of the platform or unexpected traffic conditions, the remaining wait time displayed in these updates may not decrease as expected. This expectation disconfirmation may ultimately lead customers to abandon. Therefore, we explore the interaction between the expectation disconfirmation effect and the sunk waiting time effect under different scenarios. We introduce the willingness to wait to further quantify this joint effect. After analyzing the passengers’ waiting behavior, we consider how to reduce passenger cancellations due to long waiting times. We propose an incentive-based queue management strategy—a discount method—to retain passengers. We model passengers’ reneging behaviors by characterizing the impact of waiting discounts and updated delay announcements on their travel utilities. Based on such a behavior model, we can analyze the effect of the discount strategy on the queueing process, which allows us to maximize the platform’s profit growth by tailoring the discount strategy. In addition to addressing this issue from an economic perspective, we further explore how to solve it by improving the platform’s matching efficiency. We propose an adaptive two-stage broadcasting matching mechanism that dynamically adjusts the search radius. In the first round, orders are broadcast within a smaller radius to reduce pickup times, and if no match is found, a second broadcast with a larger radius is initiated to increase the likelihood of a match. This mechanism improves both driver flexibility and passenger experience, and we aim to optimize the two-round matching radius in different market conditions. We further consider a data-driven method to improve the matching efficiency. We propose a generalized on-demand adaptable matching range technique based on reinforcement learning algorithms, aiming to achieve optimal decisions from a long-term perspective while comprehensively considering future information. Our method is suitable for different kinds of matching modes.

Date of Award	2024
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Hai YANG (Supervisor) & Jin QI (Supervisor)