Comment & Analysis

What goes into finding the next big data insight

By Eleanor Brodie, Manager for Data Science at LexisNexis Risk Solutions

If you have had a prior claim, it’s fair to assume you are a higher insurance risk. Or is it? The factors that go into understanding insurance risk are growing thanks to the emergence of new data insights, providing a much clearer, three dimensional-like view of each risk to help price fairly and appropriately.

Compared to just five or six years ago, insurance providers today have access to data attributes about an individual, an asset, a business or property that offer a much deeper understanding of risk, allow enhanced personalisation, expedite the quote process and improve pricing accuracy. This is down to a combination of factors, not least the growth of contributory databases where insurance providers share policy history and soon historical claims data.

In private motor, for example, insurance providers can understand at point of quote No Claims Discount (NCD) entitlement, cancellation risk, predicted claims cost relative to past policy behaviour, potential named driver fraud plus the Advanced Driver Assistance System features (ADAS) fitted to that individual vehicle – both standard and those chosen as optional extras.

The common factor in these data insights is that they have solved a pain point for the market. Cancellations cost money; named driver fraud leaves innocent motorists exposed; paper-based evidence of NCD was time consuming for the customer and insurance provider; and ADAS is described in a myriad of ways by car makers making it impossible to use in insurance rating until a classification system was created for the insurance market.

So when I am asked what goes into finding the next big data insight, it starts by understanding the pain point then working back to understand how and when data could be used as a solution.
We begin the process of validating an idea for a new data attribute by building an analytical prototype with the appropriate data sources and outcomes. If the concept works and the market opportunity exists, we create the final specs for technology to implement.

While this development work is happening, a testing strategy is developed for insurance providers to trial the new data attribute. This may involve creating actionable insight studies to benchmark performance or performing retro validation tests as part of a batch process. The goal of this work is to demonstrate the value of the new data attribute on the insurance provider’s own data.

As product launch draws close, we look at any required regulatory documents on the solution inputs, outputs and overall performance. Once the product is in production, the process of validation continues for new insurance providers interested in testing the data solution for themselves.
Finally, the data attributes need to be monitored on an ongoing basis to help ensure they continue to perform as expected. At some point, the product will need to be redeveloped which means rebuilding the solution and starting the cycle all over again.

The biggest barrier to creating new data insights is time. It takes time to build a contributory database. It takes time to achieve the volume of data needed to derive true insights. And it takes time to make that new finding available and useable in the insurance market. There always needs to be a balance between availability and accuracy to ensure the product produces the expected outcomes.

While contributory databases have created some of the biggest insights coming into the UK insurance market in recent years, there is a constant process of evaluating new potential data sources for product development as well as to enrich existing solutions. The decision to incorporate any data into solutions is heavily reviewed.

The insurance market is in the midst of great change driven by shifting consumer needs and behaviours, industry regulation, and increasing competition from innovators and market disruptors. The pressure is on to better understand risk than the next insurance provider. It’s not just about who is deemed a better risk or who is worse, it’s understanding the wider context to ensure the fairest treatment possible.

The good news is that with the growth in data sources and new data attributes, the better insurance providers are able to understand, segment and price their customers, helping to keep loyal customers loyal and achieving optimal customer lifetime value.


By Eleanor Brodie, Manager for Data Science at LexisNexis Risk Solutions

Show More
Back to top button