How Decathlon runs demand forecasting at scale with Chronos-2
Decathlon replaced weekly-retrained models with Chronos-2 foundation model for global demand forecasting
“would these models work on Decathlon's specific retail datasets?”
Decathlon, one of the world's largest sporting goods retailers, replaced a hybrid DeepAR/Holt-Winters forecasting stack with Chronos-2, a time series foundation model, to eliminate weekly retraining overhead and scale across global supply zones covering up to 25,000 products each. The move reflects a broader industry pattern of enterprises adopting pre-trained TSFMs to reduce MLOps complexity without sacrificing accuracy. This is a notable production-scale validation of foundation models in retail forecasting, though it is a vendor case study rather than a novel research or strategic signal.