FranceRetail and e-commerce2026 – ongoingDecathlon
Chronos-2 weekly demand forecasting
Decathlon reduced forecast error with Chronos-2 in 2 supply regions
Problem
Decathlon's forecasting models needed weekly retraining. Expansion into new regions required additional engineering.
Approach
Decathlon fine-tuned Chronos-2 with AutoGluon and ran weekly forecasts on AWS. The model entered production in Southeast Asia and Latin America. Wider deployment remained planned.
Results
28%12-week weighted absolute percentage error in Southeast Asia, previously 39%, per AWS
38%12-week weighted absolute percentage error in Latin America, previously 53%, per AWS
Takeaway
Decathlon reported results from deployed regions separately from its planned wider rollout.
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 case study
Sources
Source summaries
[1]Amazon Web Services28 Aug 2026
AWS and Decathlon staff described the production forecasting system. They reported lower forecast errors in Southeast Asia and Latin America, with expansion to other supply regions still underway.
Full details are in the original sources.
Added 1 Nov 2026 · Updated 1 Nov 2026
