I’ve posted in the past about how Big Data + Machine Learning is disrupting lending, and about how this disruption in financial services often comes from below, from startups targeting the unbanked. The Economist notes that big data + machine learning is changing underwriting at even big insurers:
At least two big American life insurers already waive medical exams for some prospective customers partly because marketing data suggest that they have healthy lifestyles, says Tim Hill of Milliman, a consultancy that advises insurers on data-mining software systems.
The software picks up clues that are unavailable in medical records. Recklessness in one part of someone’s life is a pretty good signal of risk appetite in others, for example. A prospective policyholder with numerous speeding tickets is more likely than a safer driver to end up with a sports injury. The software also detects obscure correlations. People who frequent ATMs so they can make cash payments tend to live longer than those who prefer writing cheques or paying with credit cards, it turns out. People with long commutes tend to die younger. Why this should be is not clear: some speculate that ATM users tend to be more spontaneous types, who like to have cash in their pocket and whose lifestyle may be more active; others hypothesise that sedentary commutes mean less time to do something healthy in the evening.
Interestingly, the advantage in using new sources of data to underwrite appears to lie more in cost reduction and speed to decision than accuracy:
But manual underwriting with medical tests can cost hundreds of dollars and, according to one estimate, drags on for an average of 42 days in America and Europe. That gives potential customers ample time to talk to a competitor or walk away. Automated underwriting can cost a tenth as much and be done once a human reviews the software’s recommendation.
Much of this is still in the anecdotal and experimental stage, but it is exciting to see that even big insurance companies can embrace new ideas.
Source
