Life Insurers Warm to Predictive Modeling Tools
Gen Re has followed up on its 2010 survey to find out how the use of predictive analytics is developing in the life insurance industry. Our latest research shows that predictive modeling (PM) is taking off in the U.S. individual life product market, as more vendors offering a host of different PM products enter the space.
By harnessing Big Data technologies, PM allows insurers to analyze patterns in past events and model future outcomes or behaviors. In 2013 we found eight companies utilizing some type of predictive analytics, compared to five in 2010 - a 60% increase in the past three years, though starting from a low base.
We found that PM in life insurance takes two forms. One involves using conventional underwriting evidence in new ways to improve risk classification and mortality projection. In the other, unconventional underwriting evidence is employed to classify applications quickly and/or cheaply while replicating the decisions reached with conventional evidence.
As in our prior survey, nearly all respondents recognize the latter form, with 93% telling us that PM uses public records. Fewer respondents (70%) extend this concept to credit records and even fewer (41%) see the potential to replace conventional underwriting criteria. Biomedical evidence is the mainstay of conventional underwriting and 68% of respondents observe that PM uses it.
Respondents suggested a range of potential uses of predictive modeling tools in categories ranging from marketing and risk classification, to fraud detection and differentiation. Interest in target marketing was more heavily favored this time around.
Seven of the companies surveyed said they have a predictive model currently in use. One said that it is too soon to judge performance but all of the rest stated that their model meets or exceeds expectations.
Among the four carriers using a homegrown model, three had examined vendor offerings. The other carrier reviewed vendors sufficiently well to identify expense as a prohibitive barrier. Encouragingly, the seven active carriers that have implemented PM share levels of confidence in their performance that vary from “somewhat” to “extremely” confident.
Read the summary report for more findings in Predictive Modeling.