Incorporating traffic patterns to improve delivery performance

Abstract

Traffic, construction and other road hazards impact the on-time performance of companies that operate delivery fleets. This study examines how incorporating traffic patterns in vehicle route development compares with standard, deterministic methods. We seek to understand how using historical data improves both planning and overall delivery efficiency. Our analysis contrasts manifests that were developed by an industry standard routing software tool with projections that use traffic data by benchmarking them against actual routes run by drivers. In addition to evaluating the differences between route planning tools, we explore why those differences exist, including how uncertainty is incorporated. Evidence suggests that incorporating traffic patterns into vehicle routing does produce improved solutions. Needless to say, the delivery process needs to be evaluated holistically. Our recommendations involve the various steps for creating and executing a route. Operational considerations, the potential for improving customer service, and areas for further exploration are discussed. This thesis is being conducted with sponsorship from a leading consumer products company and in coordination with the CarTel mobile sensing data project at Massachusetts Institute of Technology (MIT).