Service systems often involve multiple classes of customers with different and specialized service needs. Each customer class has a primary pool of servers where agents are trained with the required skill set. To handle large fluctuations in demand from time to time, some servers are cross-trained to help non-primary customer classes during high congestion. However, overflowing customers to non-primary pools can come with certain costs such as lower service quality, a loss of efficiency, and a heavier workload for the staff due to multi-tasking, and is thus referred to as partial flexibility. In this work, we study how to optimally utilize this partial flexibility under demand surges. We derive a two-stage index-based policy that takes the costs of waiting and overflow, the service requirements, and the (potentially time-varying) demands into account. The policy is easy to implement and achieves near-optimal performance for several classic network models. Our analysis also demonstrates the value of future arrival rate information when planning for demand surges. This is joint work with Jinsheng Chen and Pengyi Shi.
Bio: Jing Dong is the Regina Pitaro Associate Professor of Business in the Decision, Risk, and Operations Division at Columbia Business School. She obtained her Ph.D. in Operations Research from Columbia University. Before joining Columbia Business School, she was an assistant professor at Northwestern University. Her research interests are at the interface of applied probability and service operations management. Her current research focuses on developing data-driven stochastic modeling to improve patient flow in healthcare delivery systems. She received an NSF CAREER Award in 2020.