Optimal Load Acceptance Policies for Intermodal Freight Transportation

Order promising is central to the profitability of intermodal marketing companies. We study their order promising processes, where limited inventoriable containers are dynamically allocated to customers with different profit levels and delivery leadtime requirements. The container supply in the short-term decision horizon is exogenously determined by scheduled arrivals of repositioned containers and accepted loads from other locations in previous cycles. The acceptance decision involves the tradeoff between accepting a tendered load now with leadtime capacity and rejecting it to reserve capacity for potential future high profit loads, in light of real-time demand forecast. We formulate the problem as a stochastic dynamic programming model with integrated strategic targets and network effect in the model. We characterize the optimal policy by a sequence of state-independent thresholds and develop an efficient algorithm suitable for industrial-size applications. The optimal policy is driven by market segmentation, fulfillment flexibility, and intertemporal rationing over different time and capacity ranges. Through stochastic comparison we characterize the impact of dynamic demand forecast and the central role of intertemporal effect in order promising. Our experiments on real-world data demonstrate the significant profit advantage and the robustness of the optimal policy, providing insights on the value of dynamic forecast and its accuracy. Our model can be used for short-term local order promising execution in conjunction with a central tactical network capacity planning in freight transportation. A major intermodalcompany has implemented our model and seen substantial performance improvement