LEADER IN LUXURY PERFUME AND COSMETICS
Enhanced sales forecasting
Our client, a global leader in perfume and cosmetics, wants to increase the accuracy of its sell-in forecasting. The objective is to establish an automated baseline forecast on catalogue products that outperforms traditional statistical models, in order to focus Demand Planners effort on novelties and value-added forecasts adjustments on top -sellers
Scope : 20 countries – 2500 SKUs – 4 years of historical data (25M+ records)
OUR ROLE
- Develop a Machine Learning algorithm to predict future sales at a SKU x Country level, on an 18 months horizon
- Assess contribution of additional data to the overall model accuracy, such as : yearly budgets, trade marketing plans, Instagram & Twitter activity
- Deploy Machine Learning model to Production, with interfaces to a scheduler and appropriate monitoring dashboards
IMPACT
- Improved Forecast Accuracy by +10pts on a 3-months lag
- Reduction in over-forecasting bias, leading to a 2/3 reduction in local inventory obsolescence
- 11M€ reduction in safety stock inventory
- Saved 3 to 5 working days for Demand Planners each month, to focus on value-added adjustments