Fabric-based supply-chain twin lifted forecast accuracy 18 points
Demand planning ran on a patchwork of spreadsheets and a 12-year-old data warehouse. Forecast accuracy at SKU-region level sat at 61%, costing the group €40M in excess inventory.
A Microsoft Fabric lakehouse unified SAP, MES, and POS feeds; a Dataverse-grounded demand-sensing model was exposed to planners via Copilot, with Purview lineage end-to-end.
- Forecast accuracy up 18 percentage points to 79% at SKU-region level
- €27M working capital released in first 9 months
- 100% of AI inputs governed in Purview with end-to-end lineage