【报告人简介】

曹平，中国科学院数学与系统科学研究院运筹学博士，中国科学技术大学管理学院特任教授，主持多项国家自然科学基金委基金项目，研究成果发表在Operations Research，Management Science，Production and Operations Management 等期刊，并担任Operations Research, Management Science, European Journal of Operational Research 和《管理科学学报》《运筹与管理》等杂志审稿人。

【摘要】

Motivated by omnichannel service systems in which online customers can place orders remotely while walk-in customers must queue on site, we study a two-stage tandem queue with a cashier queue followed by an order-processing queue. Walk-in customers enter the first queue before proceeding to the second, whereas online customers bypass the cashier queue and join the second queue directly. Customers strategically decide whether to join upon arrival and, after joining, whether to abandon while waiting. Because online customers can overtake walk-in customers waiting in the cashier queue, the model generates overtaking-induced abandonment incentives that are absent from standard tandem-queue settings. In the base model, online customers do not observe the order-processing queue, and we compare two information structures depending on whether walk-in customers observe that queue through an in-store dashboard. In the dashboard scenario, we show that walk-in customers follow a two-dimensional threshold strategy and may abandon while waiting in the first queue. In the no-dashboard scenario, we prove that walk-in customers never abandon after joining and that their equilibrium joining behavior has a threshold structure. For online customers, equilibrium existence in the dashboard scenario is nonstandard because their expected utility may be discontinuous in the joining probability; we establish existence using a quasi-increasing fixed-point argument. Building on these equilibrium characterizations, we study how online adoption affects system performance. We show that both throughput and social welfare can decrease as online adoption level increases, implying that service providers should not blindly promote online ordering. We further compare the two information structures and show that a revenue-maximizing provider should conceal the order-processing queue when the online adoption level is low and disclose it otherwise, whereas a social planner prefers concealment only for an intermediate range of online shares. Notably, there exists an intermediate region in which disclosure improves both throughput and social welfare. %, and the outcomes of both walk-in and online customers. We also show that these insights remain robust when online customers observe the order-processing queue.

撰稿：王戈 审核：许建军 单位：现代供应链管理研究院