Photocreo Bednarek/AdobestockHow many delivery vehicles does a company really need, and which vehicle configurations make sense throughout the year? This question poses major challenges for corporate logistics departments. Christian Truden (University of Klagenfurt) and Mike Hewitt (Loyola University Chicago) have now developed a new data-driven approach that shows how vehicle fleets can be planned efficiently even under strongly fluctuating demand and pronounced seasonality.

The research focuses on deployment scenarios in which demand, customer requests and operational conditions change regularly, as is the case, for example, in grocery home delivery. Here, vehicles with multiple temperature zones must be deployed, while at the same time seasonal fluctuations and short-term orders must be taken into account. “Let us think, for example, of Christmas: at this time, shopping takes place less frequently, but in much larger quantities at once. In summer, by contrast, many people are on holiday, and there are more, but smaller, purchases,” Christian Truden cites as examples of seasonality.
In the journal Transportation Research Part E, Christian Truden, postdoctoral researcher at the Department of Economics, Analytics and Operations Research at the University of Klagenfurt, and Mike Hewitt, Professor of Supply Chain Management at the Quinlan School of Business at Loyola University of Chicago, present a method for fleet sizing in dynamic and uncertain environments. The approach combines methods from operations research with statistical models and machine learning, and makes it possible to realistically predict the performance of different fleet configurations.
Using the new method, thousands of real operational situations can be simulated across different seasons of the year. The approach uses these data to develop forecasting models for key performance indicators, including service level, capacity utilisation and costs. “With this, we present a method at a very general level that can be applied to many different problems. Nevertheless, we can only calculate proposals for fleet planning that best meet the objectives; the final decision is ultimately made by people,” says Christian Truden.
For future research, the authors of the study see further potential for increasing complexity: “Approaches such as ours can be extended to include additional dynamic factors such as uncertain staff availability. For fleet management, the integration of new delivery vehicles such as autonomous pavement robots and electric bicycles is also promising in order to expand and optimise delivery methods. Here, too, we will need models that support the identification of the optimal mix of traditional and new vehicle types for an efficient fleet,” Christian Truden outlines.
Christian Truden & Mike Hewitt (2026). Multi-objective, multi-attribute fleet sizing in a dynamic and stochastic environment: A data-driven approach. Transportation Research Part E: Logistics and Transportation Review, Volume 207, https://doi.org/10.1016/j.tre.2025.104585.
Der Beitrag Data-driven fleet planning: How many vehicles does a company need to deploy? erschien zuerst auf University of Klagenfurt.