Optimize airplane cargo loading and improve shipment efficiency by leveraging advanced analytics

Background

Our client is an American, multinational courier delivery services company, with +600,000 employees and shipping over 18 million packages per business day. The company is known for its overnight shipping service with +680 aircrafts and pioneering a package tracking system with real-time updates on package locations.

Challenge

A poor prediction of the volumes (instead of the weight of transported packages), in combination with the target destinations, resulted in aircrafts not being loaded optimally. This increased transport costs significantly, leading to lower margins and profit. Being able to predict the volume of transported packages is a key aspect in optimizing airplane cargo loading.

Approach

• Organize a pre-screening of the data relevant for the use case to check upfront if it can deliver value or not • Build a predictive model to estimate the volume of packages using dimension information gathered from the sorting centers (IOT data) • Improve predictive model by using text mining techniques based on customer descriptions on packages • Use bootstrapping techniques to evaluate prediction error intervals at container level

Key results

The accuracy of every model heavily depends on the availability and quality of the needed data.

Being able to incorporate unstructured data (i.e. text information) into the model.

Industrialization and operationalization of the loading process.

Benefits

The implemented solution resulted in a higher reliability of the volume estimates, thus optimizing transport costs by reducing the number of airplanes required. The model was constructed in a flexible way and its modularity allows the client to use more elaborated machine learning algorithms.