According to the Coalition Against Insurance Fraud, companies attribute from 5% to as much as 20% of their claims cost to fraud. In a year’s time that adds up to a whopping total of $80 billion. As if that weren’t bad enough those numbers are expected to keep growing.

To contain the problem, many insurers have entire departments and very large budgets dedicated to risk mitigation and fraud. Of course, no company wants to pay a claim only to find out it was bogus.  Once the payment is made it becomes all that much more expensive and difficult to recover the loss. That’s why so much time and energy is spent to detect fraud before paying a claim.


Traditionally, that means having humans painstakingly go through an overwhelming number of claims one at a time keeping an eye open for suspicious activity. The process is a huge drain on time, resources and revenue.  To increase efficiency and productivity while lowering costs, much of that backend research is now being done with data analytics. Even so, a lot of data that might prove meaningful never gets analyzed and often, highly skilled analysts are still needed to interpret results.

But what if …

What if instead of wading through masses of worksheets you could cut right to the chase and “see” possible fraud?  Well, now you can with advanced analytics and visualization techniques. Let’s take a simplified walk through a business case we call Fraud Invaders.

What if instead of wading through masses of worksheets you could cut right to the chase and “see” possible fraud?

In this case, we set out to answer an insurer’s crucial business question:  Can you find a new way to bring focus to a tighter subset of cases to make fraud investigations more effective?

We started by collecting claims documents that had been filled out and submitted by the company’s customers. Some of those were known to be fraudulent.  Those were flagged and put through text mining to extract anything that was a distinctive identifier such as a bank account, email address, phone number, or car registration details. Once that was done, analysis was run on to find correlations between claims.

Using that output, we built a network graph or what’s commonly called a data visualization. The resulting image that I’ve included with this blog consists of dots which represent individual claims and lines which draw a connection between data between two claim documents such as matching names or bank accounts.

Sure, it’s a pretty picture.  But it’s so much more.  See those clusters of dots?  Those are the “fraud invaders.” With little more than a glance, you can visually gauge the potential for fraud based on the size of the dots and number of connections. The bigger and more connected the cluster, the greater the suspicion of fraud.

Using this graph as a foundation the claims team can pull out the likely suspects and concentrate the weight of their investigations on that group.  Some will not prove to be fraudulent but for those that are, much less time, resources and cost will have been exerted to achieve those “gotcha” moments. Plus, additional incidents may be uncovered that would have otherwise slipped through.

In the end, we delivered what the insurer needed and we did it in a compelling and easy-to-understand format.

Not in the insurance business or not out to nab fraudsters?  Maybe your “gotcha” is reeling in new customers or retaining current ones.  Advanced analytics and visualization can help there too by revealing networks of people and heavy influencers who can help you attract new customers or cause you to lose them.  There are so many opportunities to explore.

No matter what your goal is, Fraud Invaders offers a good lesson in how to achieve the desired business outcome when you start with a solution—rather than just a problem—in mind.

Christopher Hillman

Chris Hillman is a Principal Data Scientist in the International Advanced Analytics team at Teradata. Based in London, he has over 20 years experience working with analytics across many industries including Retail, Finance, Telecoms and Manufacturing. Chris is involved in the pre-sale and start-up activities of Analytics projects helping customers to gain value from and understand Advanced Analytics and Machine Learning. He has spoken on Data Science and analytics at Teradata events such as Universe and Partners and also industry events such as Strata, Hadoop World, Flink Forward and IEEE Big data conferences. One of his achievements is as a founder member of the Art of Analytics, a programme of striking visualization production that draws people in, while retaining a solid business-led foundation.

View all posts by Christopher

Related Posts