Case study - Mail Service Logistics

Introduction

S2DS Project Partner Royal Mail Group delivers approximately 15 billion letters and parcels a year over the entire UK. In particular, parcel delivery is becoming increasingly important in their business due to digitalisation and e-commerce. The logistics system consists of 38 large mail centres and 1,300 delivery offices. Royal Mail are focussing on a more data driven approach to the distribution of parcels and letters from delivery offices and mail centres.

Problems to solve

For the team of four Pivigo data scientists, the first goal of the project was to predict the parcel traffic reaching a delivery office at a given day, with an accuracy of 85% over 90% of the days in a one-year test data set, for each delivery office. The second goal was profiling the delivery offices based on characteristics of the area they serve (number of rural/town/commercial addresses, etc.) and develop a model solely based on these profiles. This allows for future assessment of the impact of any planned changes in the logistics network.

Results

The team were provided with historical parcels traffic data for three years. In order to meet operational constraints for deploying such a model, the team needed to solve a regression, and not a time series, forecasting problem. The data scientists created a model, with an accuracy of 85% for most of the data for each delivery office. Further, they were able to reach a similar level of accuracy when generalising the model after profiling the delivery offices and then creating a general delivery office profile.

Conclusion

Through further development of the team’s model it is possible that it will increase cost efficiency, by approximately £1.5 million. This depends on the accuracy of the available data, allowing for an effective organisation of the logistic network.

“The team of Pivigo Data Scientists did a great job at picking up quite a challenging problem quickly and collaborating well with my internal team. I was particularly impressed at how quickly they embraced the tools and processes we introduced them to, which helped them deliver some very useful results in a very short time.”

Ben Dias, Head of Data Science, Royal Mail