About Decathlon
Decathlon is the French sporting goods retailer founded in 1976 and now operating in dozens of countries. It’s known for great-value products across a huge range of sports — from technical equipment to everyday activewear — and for its commitment to making sport accessible to everyone.
The challenge
Decathlon needed more accurate sales forecasts — particularly for new products with less than six months of sales history. With new products making up more than a third of the range, traditional time-series methods struggled to capture how these items would sell.
Most of the range is also highly seasonal. Classic forecasting models often fail to adapt to these cycles, especially for seasonal products launching in the near future.
The goal was clear: fewer overstocks, fewer stockouts, better inventory management and happier customers — all by improving forecast accuracy.
The solution
Convect deployed its Algoflow forecasting engine. Rather than relying on one method, Algoflow uses AI to evaluate dozens of machine-learning and statistical algorithms against each product’s sales history and attributes, then automatically selects — or blends — the approach that fits best.
That adaptability matters most in exactly Decathlon’s situation:
- New products borrow patterns from similar items and product attributes, instead of waiting for history to accumulate.
- Seasonal products are modeled with methods that recognize and adapt to cyclical patterns.
- Every model is backtested on historical data before it’s trusted, so only proven approaches go live.
Results
- Forecast accuracy improved from about 55% to more than 85%.
- Less excess inventory and fewer stockouts.
- Higher customer satisfaction, and more confident inventory and merchandising decisions.
At 55% accuracy, every downstream decision — replenishment, allocation, promotions — carried significant risk. At 85%+, Decathlon can plan each product’s trajectory with far more precision.
From forecast to action
A better forecast is only valuable if it changes decisions. In Convect’s platform, forecasts flow straight into inventory and production optimization as distributions — and a demand sensing agent watches actual sell-through, triggering a re-forecast and replan when reality drifts from the plan.
