Ask any insurance company, and they’ll tell you that industry is built on the back of data.
Ask any insurance company, and they’ll tell you that industry is built on the back of data. For years, insurance companies have judiciously collected information on people – where they live, statistics about claims – and the more information they have, the better they are at pricing insurance products. Actuarial science is legitimate and has worked for decades. Until now.
There’s a new player in the game that has the potential to blow historical data sets out of the water. Enter machine learning. ML techniques are opening up whole new ways in which companies can use data to become better at measuring risk and pricing insurance products. For an industry that has a proven track record of building their business out of historical data sets (information about the past), it’s understandable that their first inclination would be to use machine learning techniques to examine the data they already have. And they’re not entirely wrong. It’ll provide marginal improvement in their existing data and analytics.
Beyond applying machine learning to historical data sets, insurance companies need to break the mold in the way that they think about data. Think of Amazon. The eCommerce giant has acquired hordes of information on customers’ spending habits. They know where their customers live and how much they earn. The same goes for Google. By examining your search patterns, the behemoth platform can pinpoint not only exactly where you are in the world but also precisely which mechanical toothbrush you’re currently coveting. Those are some deep insights and these companies are infinitely better equipped at artificial intelligence and machine learning compared to insurance companies.
Today’s rapidly changing digital landscape has afforded a host of data like never before. It’s the application of these seemingly uncorrelated data sets that have the potential to generate superior pricing capabilities. There’s a reason why 95 percent of insurers have hastily invested in a life raft of machine learning tools. They’re on the right track. But what they should be doing is using these new tools to both procure and analyze data that hasn’t yet been obtained.[…]