September 18, 2018
3 Ways to Optimize Predictive Analytics
by Kirstin Marr
In 2012-17, P&C industry loss ratios improved by 18 points, while 20 carriers that used predictive analytics well gained 35 points.
A few years ago, simply applying predictive analytics to insurers’ underwriting practice was enough to gain a competitive edge against the large portion of the market that was still operating with traditional methods. That ship has sailed with increased adoption of analytics, raising the stakes for companies that once enjoyed a first mover advantage. Currently, 60% of insurers have welcomed predictive analytics into decision-making and underwriting processes, and research continues to show correlation between predictive analytics integration in the property & casualty industry and improvement to top and bottom lines. Companies that view analytics as a necessary commodity for modern underwriting instead of the centerpiece to their decision making will find themselves falling short of their competition. The biggest differences between the winners and losers in analytics today is equal parts ideological and technical.
In its recently published ROI study, Valen Analytics observed 20 insurance companies, representing $1.8 billion in premium, and compared their loss ratios and premium growth against the industry. The study showed that data-driven insurers consistently outperformed the market on both metrics.
- Between 2012 and 2017, the industry saw its loss ratios improve by 18 points, whereas these 20 carriers averaged improvements that were nearly twice that (loss ratios improved by 35 points).
- Between 2012 and 2017, industry-wide premium grew 18% on average, while the carriers studied grew by 53%.
For the first time since its inception, the ROI study isolated the impact of applied analytics on insurers with concerning loss ratios: those whose loss ratio were greater than 60%. This group of insurers saw loss ratios improve to market average within 12 months, and then outperform the market with each subsequent year. These results underscore the value of predictive analytics in insurance.
See also: 3-Step Approach to Big Data Analytics
Below are three best practices that the insurers studied have implemented to draw the most value from their predictive analytics programs:
The considerably positive findings of Valen’s study do not imply that predictive analytics should replace traditional underwriters. Instead, research suggests that predictive analytics tools should aid traditional insurance writers. This year’s study found that underwriter performance improves 3x when they combine predictive analytics with expertise. A well-implemented analytics solution helps underwriters leverage powerful data that they wouldn’t be able to otherwise, and underwriters provide the expertise to make the final decision. In other words, an insurance underwriter’s wealth of knowledge and contextual expertise is a largely irreplaceable asset. Underwriters know the critical variances between the price suggested by the analytics model and the historical habits of a policyholder and can incorporate this information into their decisions. Thus, predictive analytics usage augments an underwriter’s decision-making process rather than supplements it.
Streamline the workflow
Predictive analytics enable insurers to accurately align price to risk exposure, helping underwriters price policies within the context of an insurer’s risk appetite, and oftentimes allowing insurers to implement straight-through-processing (STP) for specific types of risk. In doing so, insurers can eliminate the need for underwriters to be heavily involved in certain decisions and allow them to focus on the decisions that will have the greatest impact to a book of business. This, again, leverages the expertise of an underwriter.
Incorporate the right data
Insurers that have incorporated a consortium of anonymized data into their model-building initiatives tend to be better-positioned for growth. This additional information can be crucial to initiatives like expansion across states or business classes, often by identifying risks that might fall in a blind spot of institutional knowledge. In other cases, the incorporation of consortium data will eliminate sample bias in an existing book of business. For instance, an insurer that’s relied heavily on its expertise in knowing how to underwrite low-risk construction accounts in one state to build a data set that determines good risks in a new state will risk overfitting the model, essentially giving it too high a standard. This will leave an insurer vulnerable to underpricing risky accounts without third party data to balance the scales. Consortium data increases the predictive power of models and helped the group in our ROI study of analytically inclined insurers grow premium last year, even as the market declined.
See also: Global Trend Map No. 5: Analytics and AI
For the third consecutive year, Valen’s ROI study has identified just how much value applied analytics can add to insurers. The carriers that have leveraged analytics and consortium data and empowered their underwriters have realized significant advantages over competitors to improve both profitability and growth.