GLOBAL TIRE MANUFACTURER
Improve sell-in forecast processes and accuracy using AI models
Our client wants to better forecast its long tail products, which display a very erratic and seasonal sales behavior. This complexity was priorly managed by forecasting sales at macro levels using traditional statistical methods, which high levels of errors were observed after disaggregation at the SKU level.
Scope : 2000 SKUs, Europe
OUR ROLE
- Engineer relevant features based on products characteristics to feed into the forecasting model.
- Evaluate the positive impact of including other data sources in addition to sales history: pricing, vehicles and market data, competitor offers…
- Develop a Machine Learning forecasting algorithm to predict future sales at the SKU level
IMPACT
- +15pts Forecast Accuracy improvement, measured at the SKU level (Lag 0 to 6 months)
- -10pts Bias improvement
- Machine Learning model fully implemented after 6 months of double run with no loss of performance.
- Strong support and engagement from the operational forecasting teams, who had been fully engaged in the project from the beginning.