Insurance Claims Forecasting on a Databricks Lakehouse
Insurance — Claims data lived in legacy warehouses, so actuarial and finance teams waited weeks for reserving figures and claims-volume forecasts. Results:…
Client: General Insurer. Sector: Insurance.
The challenge: Claims data lived in legacy warehouses, so actuarial and finance teams waited weeks for reserving figures and claims-volume forecasts.
What we built: Moved claims, policy and payment data onto a Databricks lakehouse governed by Unity Catalog, then built forecasts for claim frequency and severity using MLflow-managed models retrained every month.
Tech used: Databricks, Delta Lake, Unity Catalog, MLflow, Spark.
- Data Refresh Frequency: Weekly → daily
- Time to Forecast: -75%
- Records Moved: 1B+
- Platform Spend: -30%