Decathlon uses Chronos-2 on AWS to improve demand forecasting accuracy by 11–15 percentage points

Decathlon, a sporting goods retailer with more than 100,000 teammates and 400 million users, relies on accurate demand forecasting to stock the right products in each store at the right time. The company’s scale—tens of thousands of products across more than 80 sports and multiple continents—introduces highly seasonal demand patterns that complicate forecasting.

After evaluating multiple time-series foundation models, Decathlon selected Chronos-2 as a core component of its forecasting stack. The retailer reports that Chronos-2 improved forecast accuracy by 11–15 percentage points compared to prior methods while reducing operational complexity.

Decathlon runs weekly inference for its global forecast pipeline on CPU-only AWS instances, with each execution costing approximately $0.03. The company describes the architecture it uses to deploy Chronos-2 at scale on AWS and shares practical lessons for other enterprises adopting foundation models for forecasting.

The post is co-authored by Decathlon engineers and AWS contributors, and it appears on the AWS Machine Learning Blog under the Amazon SageMaker category.