Mondragon Innovation | Humanity at work – Prediction of Home Delivery Activity (Last Mile Logistics)
Mondragon Innovation | Humanity at work
Sector: Commercial sector
Business Case
The number of orders to be delivered to customers’ homes is not homogeneous and fluctuates due to internal and external variables. Tailoring the sizing to the unique requirements of the delivery teams responsible for home delivery enhances both customer satisfaction and operational efficiency.
Objectives
Forecast e-commerce orders scheduled for home delivery, enabling the proper allocation of transport teams and ensuring on-time deliveries as promised.
Use case
Use a predictive model to estimate the daily order volume assigned to each transport team for a 30-day period of home deliveries
Infrastructure
On Premise
Technology
Machine learning and deep learning
Data
Dependent variable time series, seasonality capturing time variables and commercial regressors.
Resources
Staff specialised in data and analytics: data scientists, data engineers, data visualisation engineers. Specific big data architecture including data fabric. Descriptive analytics tool for model output consumption and MLOps.
Difficulties and learning
The key insight from the model was the need to handle the training time series portion that coincided with the COVID lockdown period differently. During this timeframe, there was a rapid and substantial surge in demand for these services.
KPIs (business impact and metrics of the model)
Model predictive capacity monitoring metrics: MAE and WAPE.
Funding
No external financing.
Collaborators, Partners
Specialised consulting.
