Decathlon's New AI Predicts What You'll Buy
Fewer sold-out gear and less waste at your local store.
Decathlon is one of the world's largest sports retailers, with 100,000 employees and 400 million customers. Their huge challenge is predicting how many ski gloves, surfboards, and running shoes each store will sell. If they guess wrong, shelves go empty or warehouses fill with unsold gear.
To fix this, they tested several AI forecasting models on two years of their own sales data. They picked Chronos-2, a pre-trained AI that spots patterns in sales history without needing to be rebuilt every week. Their old system had to be re-trained weekly and struggled to expand to new regions. Chronos-2 works in "zero-shot" mode right out of the box, matching their old system's accuracy. With a light tune-up every six months, it performed even better.
This matters for shoppers: forecasts feed the entire supply chain over a 12-week horizon for reordering and a 52-week horizon for long-term planning. The system now runs across regions covering up to 25,000 products each, including Europe and India, with Middle East and Africa coming soon. For Decathlon, it means less wasted inventory, lower costs, and more reliable stock. For you, it means the bike pump or yoga mat you want is actually on the shelf when you need it.
- Decathlon switched to an AI that predicts product demand weeks to a year in advance.
- Tests on 2 years of real sales data showed the new AI beats their old weekly retrained system.
- The model runs across regions with up to 25,000 products each, helping avoid both stockouts and overstock.
Why It Matters
Better AI forecasting means fewer empty shelves and less wasted inventory at a store you probably shop at.