Lessons from 10 years of predictive analytics in life insurance
Article information and share options
There's a lot to learn from over 10 years around the life insurance industry. And when it comes to predictive analytics, that can be distilled into 5 lessons.
Lesson 1: If underwriting is a barrier to sales and you have rich data on prospective customers, you can speed up the process for some
An effective predictive underwriting model is built on rich data sources. This could come in the form of credit scores, demographics, or credit card use – it could come from less traditional sources, such as visits to a particular department store. Around 30 data points can form good correlations. A typical good source of data is banking data.
A good predictive model can provide preapproved underwriting for the top 40% segment. They are only required to sign one health declaration. The underwriting time for this segment fell from an hour to 15 minutes.
Lesson 2:…but don't expect that an easier onboarding process will in itself increase demand for your product
So making the underwriting process easier should increase sales, right? Actually, that fact was not supported by the data. Analysis suggested that predictive underwriting did not make any material change to demand.
Lesson 3: If anything, the main benefit is for the sales agent rather than the end customer
Preapproved underwriting means that sales agents don't need to take time asking sometimes difficult questions. Their conversations with clients become much easier. In itself, this is no bad thing for the insurer. Their sales team has more time to convert leads.
Lesson 4: It doesn't have to be banking data; you can find predictors of health in other data-sources too (but beware the limitations)
Some of the best predictive models have come through bank assurance, as banks have good data sources on their customers. If bank data is not at hand, many new sources are – such as credit scores, social media, or wearables data. The challenge is knowing how much of this data can be used and how predictive this data actually is.
Lesson 5: Sometimes less is (genuinely) more
Data points should be used intelligently to inform the underwriting process. A customer taking out a business loan, for example, is likely to be in good health. That provides a potential point of sale for a life insurance policy without requiring extensive underwriting. Equally, mortgages provide an opportunity to cross sell with minimal underwriting.
The ultimate learning: However much or little you have, use data intelligently.
About the speakers
Claire Nolan is Head Underwriting Origination UK&Ireland at Swiss Re; William Trump is Customer Behaviour Consultant at Swiss Re. They spoke at the "Next Generation Insurance Customer: Bringing it to Life" which took place on 22-23 June 2016 at the Swiss Re Centre for Global Dialogue.