
“Modeling and Analysis of On-demand Service Systems “
by Ke Sun
McGill University
Place: MA-330
Abstract
On-demand services, ranging from ride-hailing to mobile ordering, rely on real-time matching between customers and service providers. These systems increasingly allow customers to engage remotely before arriving at the service facility, creating new operational challenges around congestion, information, and incentives.In this talk, we focus on the order-ahead setting in quick-service restaurants, where remote customers place orders before traveling and their orders begin advancing in the preparation queue. Although order-ahead is widely believed to reduce delay and increase throughput, we show that adopting the order-ahead technology can sometimes make the system unintentionally yield lower throughput than an onsite-only system, even when the provider optimally chooses whether to share queue-length information. We then examine several mitigation strategies. Allowing remote customers to cancel upon arrival can restore much of the throughput benefit, while a threshold-based rejection policy for remote orders can further improve system performance when optimized.
Bio
Ke Sun is currently a postdoctoral fellow at Desautels Faculty of Management, McGill University. Her research interests include queueing theory, stochastic modeling, Markov decision processes, and their applications to on-demand service systems. She received her Ph.D. in Statistics in Beijing Jiaotong University.


