However, the seventh installment of the company’s Property & Casualty Claim Officer Survey indicates that use of these technologies is on the rise.
Despite many months of commentary indicating the value of predictive analytics in claims, the number of complete implementations at P&C insurers is low, according to the latest installment of New York-based Towers Watson’s Property & Casualty Claim Officer Survey.
The research, which polled 41 chief claims officers this summer, found that only 17% of respondents had been using predictive analytics technology in their claims organization for a year or more. However, another 7% nearly are ready to roll out the capabilities, and a further 39% have initiated exploration of the technology.
Respondents who had implemented predictive modeling found it contributes “to better fraud recognition and response, improved payout discipline and operational improvements,” according to the survey.
“Predictive modeling applications are growing in claims, with fraud the initial focus, followed by triage. Carriers making the investments are increasingly recognizing benefits,” said Brian Stoll, director of Property & Casualty practice for Towers Watson, in a statement.
More than half of the insurers surveyed have made changes to their claims best practices over the past year, and predictive analytics, however, isn’t the only technology driving this change. For example, 63% of CCOs reported that “claim-handling assignments and workflow processes are most positively influenced by recent system technology improvements,” according to the survey.
“Technology investments in this area help insurers better identify and respond to trends culled from data-driven analytics, and make it easier to implement predictive models and best practice enhancements that carriers are developing to improve claim performance,” Stoll adds.
Towers Watson also found that these types of capabilities, which promote operational effectiveness, are likely to be added through proprietary development, while vendor applications are more likely for predictive analytics.