Unified Data Platform for a National Retailer

A national retailer with 300+ stores ran inventory, POS, and e-commerce data through 14 disconnected systems. Store managers made restocking decisions on data that was 24-48 hours stale, driving both stockouts and overstock. Finance and merchandising teams routinely produced conflicting reports because each pulled from a different source of truth.

We consolidated ingestion through Databricks Lakeflow into a medallion architecture (Bronze/Silver/Gold), governed centrally through Unity Catalog. Real-time POS and inventory events stream into Delta Lake tables, with Gold-layer aggregates powering both BI dashboards and a demand-forecasting model served through Mosaic AI. Unity Catalog enforces row-level security so regional managers only see their own store data.

Store managers now see inventory positions updated every few minutes instead of once a day, and a forecasting model recommends replenishment quantities automatically. Finance and merchandising teams query the same governed Gold tables, eliminating conflicting reports. A lightweight internal app surfaces recommended actions directly to store managers.

The retailer cut reporting latency by 62% and reduced carrying costs by consolidating overstock, saving an estimated $2.1M annually in the first year. Store-level stockout rates dropped as replenishment recommendations became data-driven rather than intuition-based.