The challenge
An industrial manufacturer was losing €24 M annually to unplanned downtime across 14 plants. Sensor data sat in plant historians with no central aggregation, and maintenance was calendar-based rather than condition-based.
What we delivered
- 01Azure IoT Hub ingestion of 48,000 sensors across 14 plants.
- 02Databricks Delta Live Tables streaming pipelines with bronze / silver / gold.
- 03Predictive maintenance ML models with MLflow tracking, champion-challenger, and drift monitoring.
- 04Maintenance scheduling integration with SAP PM.
- 05OEE dashboards on Power BI with plant-level drill-down.
Outcomes
- −41% Unplanned downtime.
- +6.4 pts Overall equipment effectiveness (OEE).
- €24 M Annual unplanned downtime cost avoided.
- 48,000 Sensors streaming with <5 s latency.
“Condition-based maintenance is now our default, not an aspiration.”
— VP Operations, Industrial Manufacturer