Predictive insurance plus machine learning is no dream

Why can’t the insurance business be more like Netflix? Or Tripadvisor or Booking.com?

In my opinion, the transformation of insurance services and products by digitalisation will be nothing short of mind-blowing. Nonetheless, there is a great deal that insurers can learn from the machine learning technology behind the success of some top 21st century digital service brands.

These brands are businesses that thrive by constantly absorbing detail about their customers’ likes, dislikes, tastes and even locations and travel patterns. This informs how TripAdvisor or Netflix tailor their suggestions for places to stay or eat or content that you might like to watch. What’s more, what they learn about you allows them to refine suggestions for people who are like you.

In very simple terms, this summarises how predictive analytics works on a huge and exciting scale. Something similar could be transforming how insurers operate and interact with their own customers.

As these high profile digital service brands are data-driven businesses, so similarly is insurance. For many years, the industry has recognised data analysis as its lifeblood; indeed, actuaries could be described as pioneer data analysts, and they have been around since the late 1700s. Big data has therefore been a natural topic of debate in the industry since the term first appeared several years ago.

Today, as they become true digital businesses, interacting with customers over various technology platforms in real time, the pressure for insurers is to understand what could happen next and what the best decision should be. With so many complex, fast-moving variables, predictive analytics can help insurers identify which values in data sources (such as telematics, social media etc…) are relevant, and which are irrelevant.

What drives the best predictive analytics tools is superior machine learning. Such tools are constantly searching through data interactions, creating and testing model after model to produce the most reliable predictive models.

The applications of machine learning for insurers are similar to how Netflix knows to suggest Orange is the New Black rather than Pretty Little Liars. Machine learning algorithms are applied to data about policies and claims to create models of client retention and loss ratios. These models can be used to pick out rich market segments that the competition can’t see, be first out with smarter pricing and underwriting, as well as find the best customers, and manage claims more cost effectively.

Machine learning does what computers are best at doing: checking out almost countless possible combinations. What does this mean for insurance professionals? Digital transformation is already changing the nature of work inside insurance companies. Machine learning accelerates this process, but the changes aren’t necessarily negative; it can reduce the tedium of repetitive tasks, and help with real-time scoring in assessing risk more accurately, triaging claims and detecting fraud.

Insurers have tended to adopt technology at a gentler pace than other sectors. But predictive analytics offers them tools that are going to be critical in how they transform their businesses. There is some truth to the argument that big data has been a distraction, a background noise. Predictive analytics bypasses this by learning to identify quickly the signal from that noise.

This makes me believe that this will be a technology the industry will adopt faster than others. As legacy software is replaced by modern core systems that support digital transformation, the adoption of predictive analytics is going to take off dramatically, especially as new products are developed and market segments are targeted. Forget the gentle pace; insurance will be an enthusiastic proponent.