Implementing Real-Time Optimization Models: A Case Application From The Motor Carrier Industry

Abstract

Online models for real-time operations planning face a host of implementation issues that do not arise in more strategic arenas. We use the seemingly simple problem of assigning drivers to loads in the truckload motor carrier industry as an instance to study the issues that arise in the process of implementing a real-time dispatch system. Although the project was moderately successful, our focus is not on documenting the benefits, but rather on summarizing the challenges that arose. The most significant theme running through the implementation hurdles we encountered was the lack of information available to the model. Computers are very good at processing vast quantities of information; humans are very good at challenging the information that is in the computer and augmenting computer-provided data with head knowledge. Our study includes a careful comparison of actual decisions with model recommendations, using a six-month database of actual transactions. This comparison is the first we have seen of its kind and provides the most rigorous evaluation of an online dispatch model that we have seen. Although the model was well used, the results demonstrate that significant improvements could have been obtained if the level of model utilization had been even higher.