Big data versus good data: Using ML to plug data gaps in insurance – Alicia Montoya & Antonio Savona, Swiss Re


Significant founding of enterprise machine intelligence for the past decade has not always resulted in a measurable return of investment in the insurance landscape, as the profitability of large and dynamic datasets is often hampered by insufficient comprehensiveness, lack of explainability and increased information management complexity.  We look at these gaps, and we explain how the latest advancement in ML can help insurers close them implementing ML-driven data augmentation, building intelligible ontologies and allowing accurate risk modelling based on accessible, curated datasets. 

Key Takeaways

Insurers are using technology to access relevant, granular data to steer risk assessments and shape better products and services. Key approaches include:

  • FIXING DATA: Out-of-sample predictions of missing values with machine learning algorithms can help close data gaps in a wide range of insurance specific dataset
  • MAKING DATA ACCESSIBLE: ML-driven data curation algorithms can transform non-intelligible data-streams into accessible and explainable ontologies that underwriters can leverage to take informed decisions 
  • MAKING DATA USEFUL: ML-curated data allows insurers to move from classic Generalised Linear Models to more accurate, data driven, risk assessment, thus implementing automated, data-led underwriting.

Add comment