Insurance is changing. The opportunities of a data-lead strategy running throughout the entire insurance lifecycle are obvious, but not so easy to achieve. By focussing on achievable goals and the appropriate use of non-perfect data much can be done to find valuable pockets of risk that are significantly out of step with the business appetite.

While the SEC will rightfully continue to demand workflows that follow familiar and transparent pathways, alternative valid analyses can be transformative, whether they are based on creative solutions to untidy data, or opaque Machine Learning solutions.

This discussion starts with a view of what ideal data would be, and then moves to how imperfect data can, in practice, be used to move towards the ideal. Ideal data would always have Loss Ratios consistent with explicit appetite, whatever the specific analysis, and whether by historical analysis or projection. While we can aim towards perfect data and perfect analysis, much can be achieved by using imperfect data with well thought out analysis.

Finally a metric (Annual $’s Optimized) for valuing the financial opportunity of unbalanced risk is offered. This metric can be used to measure progress, as well as validating spending on analytic resources.

Data analytics for Specialty Insurance will always be challenged by the dramatically broad range of products. In contrast Personal Lines insurance has a 35 year history of focused and statistically meaningful analysis due to the width, depth and heterogeneity of its data. Value creating analysis for many Specialty lines is now becoming possible with recent technology, such as fast & cheap storage and processing using the new BI tools.

This bewildering variation of Specialty products is compounded by how often Specialty Insurance companies have grown by acquisition leading to data sources that are inherently incompatible. Compounding the differences is the fact that data structures are often tightly connected with established business processes which are both costly, and risky, to change. The result is that data is often far from fully integrated.

The same underlying single source of data can be integrated into a circular insurance workflow:

The flow of insurance data runs from appetite setting, marketing, submissions, binding, pricing, reserving, capital modeling, all the way to claims, and back to the beginning. Ideally the same data flows through all parts of the journey.

The present challenge across the speciality industry is to reform the inherited data silos to enable this flow through the organisation.