The ability to accurately define risk and determine policy premiums has been at the heart of the insurance business since the late 17th century. Over the years, the science of rate-making across all lines of business has become ever more sophisticated and never more so than now with the explosion of Big Data. Insurers have access to more data than ever before and the ability to make strategic use of this information throughout the life cycle of customer policies is pivotal to engineering the right conditions for business success.
Modern risk profiling since the 1970s adopted predictive models based on traditional statistical techniques to arrive at a risk ‘score’ for a customer at all stages of the life of a policy (Point of Sale, Point of Claim and Point of Quote). Statisticians translate business problems into statistical and mathematical formulation, applying statistical modelling techniques to help solve the original business problem: generalised linear model, logistic regression, cluster analysis, decision trees and expert rules are the most frequent methodologies used for this purpose. These risk-scoring approaches have undoubtedly given insurers a keener insight into risk and propensity to claim in the last 20 years.
Today, with the availability of more data (unstructured data, telematics, social network, and behavioural data) these traditional methodologies are used jointly with innovative risk assessment methodologies; as used independently they are unable to draw from these new, vast and more complex data sets.
In collaboration with our insurer clients, CRIF has been investing in research and development to design advanced techniques in risk profiling and develop our DNA of Analytics. We are drawing on information about human behaviour, lifestyles and habits and computing large quantities of both structured and unstructured data available via social media platforms and beyond, dishonest behaviour, and discrepancy between what is declared and what is found on CACHE/CUE databases, to generate a more powerful and accurate predictive risk profile.
The new methodologies refer to machine learning and Genetic Algorithms to solve problems and find indicators or Evolutionary Neural Networks to optimise the relationship between information. The application of knowledge discovery and data mining through the use of Link Analysis and Neural Networks enables identification, analysis and visualisation of patterns in data; embracing intelligent technology to auto learn and inform processes. Proof of concepts are underway and providing exciting results.
In a recent study conducted by CRIF which involved reviewing an insurer client’s internal data, the application of telematics data improved the accuracy of risk scoring significantly.
So, as the industry innovates to adapt to the changing face of risk and consumer expectations; how can insurers expect to benefit from the move to embrace enriched intelligence from these new data sources?
The top 5 anticipated benefits can be categorised as:
- More accurate risk evaluation.
- Reduced false positives improving the quality of alerts.
- Increased ability to detect and prevent fraud.
- Time savings with reduced requirement for human intervention.
- Improved profitability.
Today, risk profiling intelligence is invaluable at all stages of a policy life cycle, to include mid-term adjustments and claims, alongside point of quote and sale. The ability to integrate and apply this intelligence as necessary will differentiate insurers within their chosen markets going forward, supporting their drive to build profitable books of business and deliver tailored quotes and service to their customers.
As we move forward with our DNA of Analytics we will continue to apply a bespoke approach, building rules specific to each insurer client’s individual underwriting and claims handling strategies to deliver ongoing competitive advantage.
