SwitzerlandRetail and e-commerce2023 – ongoingCoop (Switzerland)
ML demand forecasting for fresh produce at distribution centers
Coop Switzerland tested cloud AutoML for fresh-food demand forecasts and reported a 43% performance improvement over its in-house models in a proof of concept
Problem
Coop needed accurate forecasts of fresh produce demand at its distribution centers to keep shelves stocked and cut food waste. Its on-premises workstation struggled to scale.
Approach
Coop's ML team moved data from SAP and its pipelines into BigQuery. It compared an XGBoost model, a Temporal Fusion Transformer and Google's Vertex AI Forecast on a proof-of-concept test set, and chose Vertex AI Forecast. A small production pilot for one distribution center was being set up, with possible expansion to all Swiss distribution centers.
Results
Takeaway
The 43% figure comes from a proof of concept against Coop's own earlier models. Coop did not report a measured reduction in food waste.
Categories
- Size
- Large enterprise
- Industry
- Retail and e-commerce
- Business function
- Logistics and supply chain
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor's case study
Sources
Source summaries
Google Cloud and a Coop data scientist reported a WAPE of 14.5 for Vertex AI Forecast and a 43% improvement over Coop's in-house models in a forecasting proof of concept. Coop was preparing a production pilot at one site.
Full details are in the original sources.
