Analytics keep the insurance fees affordable

Ever wondered why your insurance policy gets more expensive every year? There are various factors contributing to this unpleasant trend but insurance fraud might well be the most important one.

According to statistics from the Dutch league of insurance companies, insurance fraud has increased by 25% over a five year period, which in turn has led to a €150 increase of the average insurance policy fee. This is not only worrying the customer but the insurance companies as well. They are looking hard to find better methods to defend themselves against fraudsters.

An important challenge when fighting fraud is the enormous diversity: there are many different types of fraud, each with their own method and typical behavior, ranging from the opportunists that exaggerate their claim hoping for a better refund to members of criminal gangs that stage fake accidents in order to cash insurance money.

This makes it hard to fight fraud with one single encompassing system. Even more so because the relevant data to detect fraud is often spread over different data silos which are far from integrated. This usually makes the available data incomplete and unreliable.

Another problem that insurance companies face is the large volume of false positives when using their traditional fraud fighting methods. This would lead to a significant rise in operational costs, with even higher insurance policy fees as a result. Additionally, you don’t want to treat innocent customers like criminals: insurance claim handling is one of the rare moments when insurance companies truly get in touch with their customers.

Go for hybrid

So what do insurance companies need to fight fraud more effectively? More sophisticated detection algorithms, of course. But even the best algorithm in the world needs good data if it is to produce good results. Good data means: cleansed data and integrated data. All data need to be collected, de-duplicated, combined and analyzed, in order to get a holistic view of the organization and of the activity in and around the organization.

Next, we would advise a hybrid approach, combining various analytic methods and approaches, from business rules detection models for deviations to text and social network analytics to identify suspicious behavior.

This analytics-based approach has led to very positive results, judging from a SAS survey from last year. 57% of the surveyed companies that use business analytics have been able to increase the number of fraud detections by at least 4 percent. In contrast, only 16% of the companies that don’t use automated solutions have seen a similar increase.

More practical advise and results from this survey can be found here.