Data-rich companies, such as insurers, banks or retailers, have the opportunity to reduce the underwriting process for their healthy customers with the help of a predictive model.
There was an interesting article published on Forbes.com earlier this year on how Target in the US predicted a teen’s pregnancy long before her father found out. Target identified around 25 products, such as scent-free soap and cotton balls, that when analysed together, allowed them to assign each shopper with a “pregnancy prediction” score. They also found they were able to estimate an expectant mother's due date to within a small window and use this information to send coupons timed to specific stages of her pregnancy.
This rather interesting story illustrates how supermarkets have been using intelligent data for some time.
Other examples of the use of intelligent data are YouTube and iTunes that recommend videos or songs based on an individual's download history. An online dating site also uses an algorithm to suggest possible "matches" based on preferences and behaviour on the site.
What does all of this have to do with insurance? General insurers have demonstrated how to use intelligent data in their offerings for some time, for example offering motor insurance based on credit scores. However, Life Insurance is under-developed in this area.
So what if we could predict mortality based on everyday information?
Paul Hately, Global Head Accelerated Underwriting at Swiss Re, recently visited Australia to talk about predictive underwriting, which he calls a "game-changer" for the Life Insurance industry.
"Predictive underwriting is the use of intelligent data on consumers to reach a view as to their health status", says Paul. "The premise is to find out if there are correlations between lifestyle factors and mortality, and then build a protective model, which results in changing the underwriting process." For example, one insurance partner of Swiss Re managed to reduce its underwriting process from more than 30 health-related questions down to a single in/out statement for the "good prospects", as scored up by the predictive model.
In order to achieve this, two comparable depersonalised data sources are required – data that describes the health (risk) of the individual, such as final underwritten decisions, and independent descriptive data, such as bank account or loyalty card information. These are then analysed to highlight the correlations. Any factor found to be correlated with health is referred to as a "predicto" and by combining the predictive variables, an algorithm is built, which ranks each customer in terms of their likelihood of being offered standard rates when making an application for life insurance.
"Traditional underwriting is about identifying the unhealthy minority among the applicants, whereas predictive underwriting enables us to approach the healthy majority of the population who haven't applied for protection", says Paul.
Any information held on a customer could be predictive of their health status. “The data does the talking” says Paul, and some not too obvious lifestyle factors, like the frequency of someone using an ATM, could be predictive.
Table 1 shows that a bancassurer in the UK found that of its customer base, people aged over 40 who use an ATM less than 15 times a month have a higher likelihood of being rated or declined when applying for life insurance, compared to those people under 40 who use an ATM more than 15 times.
Table 1: ATM cash withdrawals in last month (UK Bancassurer)
"By lowering the main barriers to life insurance sales i.e. long underwriting process, costs and time to complete an application form, the predictive underwriting model can be a new way to offer life insurance and improve customer satisfaction with the process, without having to increase the price compared to a fully underwritten product," says Paul.
Table 2 shows how the predictive underwriting approach could be compared to some of the existing fully underwritten products.
Table 2: Comparison between traditional and predictive underwriting application
In the past 12 months, Swiss Re has developed and launched a predictive underwriting product with two bancassurers in the UK and has a proven track record in this space. Banks use credit scoring techniques to grant pre-approvals for loans and so this extension into life insurance is a natural progression.
Paul also notes that, “If you believe that underwriting is a barrier to sales and you have sufficiently rigorous underwriting, and hold data of sufficient quantity and quality for a predictive model build, this model might work for you.”
Published May 2012
In our ongoing efforts to improve the quality and relevance of our publications, we would like to know more about you.