Case Study: Nationwide achieves more accurate insurance pricing and faster ML at scale with Databricks

Accurate insurance pricing with data and ML

Nationwide, a major insurance provider, faced inaccurate pricing and slow model development because their legacy batch processes and limited infrastructure had to analyze hundreds of millions of volatile insurance records. Infrequent, unpredictable claims plus cross-team silos made it hard to extract actionable insights and produce reliable actuarial models at scale.

By adopting the Databricks Unified Analytics Platform, Nationwide centralized data ingestion, enabled collaborative notebooks, and used managed MLflow to train hierarchical deep learning models in production. The change delivered big performance gains—data pipelines sped up 9x (34 hours to <4), featurization 15x faster (5 hours to ~20 minutes), training time cut by 50%, and scoring improved ~60x (3 hours to <5 minutes)—producing more accurate pricing predictions and a material positive impact on revenue.

Nationwide

Bryn Clark

Data Scientist