Demand Forecasting Logistics

Challenge

Understanding the demand structure of one's customers offers numerous opportunities to optimize internal and external processes and workflows. Our client, an international logistics company, faced the challenge of needing to identify transport demands for their customers several days in advance. This foresight allows for the timely allocation of transport resources, preventing bottlenecks in the client's supply chain.

Approach

To forecast the anticipated transport demands, we first conducted a detailed analysis of the data and classified the corresponding time series. As is often the case in the B2B sector, not all customers exhibited stable, seasonal patterns; instead, their behaviors varied significantly, making a single model specification insufficient for satisfactory forecasting performance. Therefore, we developed a model factory consisting of various model types that automatically select the most suitable model for each customer's forecast daily. The selection is based on backtests of the models and bootstrap resamples of the available time series data.

Result

The results of the fully automated forecasts are fed daily into the client's internal reporting and dashboard systems, where they are successfully used by the respective account managers as decision support for disposition.