The Power of Reactive Upgrades in Dynamic Resource Allocation with General Upgrading

信息管理与工程学院102室
上海财经大学（第三教学楼西侧）

上海市杨浦区武东路100号

Cui Zheng is   currently an Assistant Professor in the Department of Data Science and   Management Engineering at the School of Management, Zhejiang University. His   research focuses on data-driven decision-making, supply chain and inventory   management, robust optimization, and goal-driven risk management. His work   has been published in leading academic journals such as Operations Research.

Product upgrading, which involves offering customers   higher-quality products or services, has become an increasingly popular   strategy for firms dealing with product proliferation. A common approach is   to offer upgrades reactively—providing a substitute only when the initially requested product   is unavailable. Despite its widespread use, the efficiency of this reactive   heuristic, as well as the optimal upgrading policies within a general network   and upgrade structure, remains poorly understood. This paper addresses these   gaps by studying dynamic network revenue management with multiple vertically   differentiated products and general upgrade options. We establish necessary   and sufficient conditions for fulfillment decisions that determine when the   reactive heuristic achieves optimality, focusing on the interaction between   upgrading flexibility and the consumption network structure. For acceptance   decisions involving sequentially arriving customers, we first analyze a   simplified network with three products and demonstrate the optimality of   state-dependent protection levels that exhibit monotonicity and limited   sensitivity. Extending this to a general network, we leverage the structural   properties of the optimal policy to develop computationally efficient   heuristics using fluid approximation, which we show are asymptotically   optimal under high demand. Through extensive numerical simulations, we   further validate the effectiveness of these heuristics under scenarios of   lower demand and analyze how key model parameters affect the value of   upgrading flexibility. Our results offer practical insights for designing and   implementing upgrading strategies in complex product networks.