FMCG Demand Forecasting at SKU and Depot Level
Supply Chain & Logistics — Forecasts kept in spreadsheets failed to anticipate festival spikes and regional variation, so some depots ran out of stock while…
Client: FMCG Distribution Company. Sector: Supply Chain & Logistics.
The challenge: Forecasts kept in spreadsheets failed to anticipate festival spikes and regional variation, so some depots ran out of stock while others held weeks of surplus.
What we built: We delivered a hierarchical forecasting system on Databricks at SKU × depot × week level, using promotion, festival and weather signals, with a lightweight web app for planner overrides and automatic replenishment recommendations.
Tech used: Databricks, LightGBM, Prophet, MLflow, Delta Lake.
- Forecast Accuracy: +22%
- Out-of-Stocks: -35%
- Surplus Inventory: -18%
- SKUs Covered: 8,000+