BI’s Article search uses Boolean search capabilities. If you are not familiar with these principles, here are some quick tips.

To search specifically for more than one word, put the search term in quotation marks. For example, “workers compensation”. This will limit your search to that combination of words.

To search for a combination of terms, use quotations and the & symbol. For example, “hurricane” & “loss”.

Login Register Subscribe

PERSPECTIVES: Social media helps insurers detect fraud, but doesn't replace analytics

PERSPECTIVES: Social media helps insurers detect fraud, but doesn't replace analytics

INTRO: As fraudulent claims increase, insurers have new data sources to try and fight the fraud. From traditional but now digitized industry watch lists to mapping of hot spots, Stuart Rose, global insurance marketing manager at the SAS, says data sources are growing rapidly.

The stories of dumb fraudsters, such as the street-racing siblings who recently posted their exploits on YouTube, populate the Web. Because there's so many such examples, it would seem insurance companies should invest in automated technologies or services to Google the names on every claim.

But for every dumb fraudster, there are multiple smart ones who can only be detected by analyzing the information buried in claims data. Does that mean investigators should ignore social media? Absolutely not. As claims fraud grows, investigators need to expand their toolbox.

Questionable insurance claims have increased 27% between 2010 and 2012, according to the National Insurance Crime Bureau. As fraud has grown, insurers have new data sources to try and fight it. From traditional but now digitized industry watch lists to geographic information systems mapping of hot spots, data sources are growing rapidly.

Text analytics has tapped into the rich vein of unstructured data that resides in claims. Insurers can tease out the similar phrases and circumstances that are common to organized fraud. Link analysis helps spot otherwise hidden patterns of relationships between claimants. A recent Coalition Against Insurance Fraud report states that 88% of the insurers it surveyed say they are using anti-fraud technology.


Meanwhile, other units within insurance companies have embraced social media to market products and build customer relationships. So why not harness social media for investigations? Actually, the same Coalition Against Insurance Fraud reports that 36% of companies say they do use social media as part of their investigations.

Is it helpful? If an investigator or automated fraud detection software has already flagged a potential problematic claim, running the claimant name through various search engines to find Facebook pages, Twitter accounts, Instagram sites and YouTube postings would be logical. But paying to have every claimant's name searched is too expensive and too prone to mistaken identity issues.

Rather than turning to social media first, the better choice is for insurers to look for ways to efficiently comb their own data, and deploy social media once after flagging a suspect claim.

The two most commonly used tools for working with their own data, and some outside data sources, are text analytics and link (social network) analysis. Text analytics analyzes unstructured data like claims notes, customer service logs, police reports and medical records. Social network analysis helps uncover organized fraud rings that might otherwise take years to identify. Social network analysis spots addresses, body shops, physicians, phone numbers and employers that multiple claimants have in common. Both types of analysis are designed to use existing information to uncover possible fraud before payment is made.

With text analytics, an insurer can analyze written medical records and find that several rear-end accident “victims” have suspiciously similar diagnosis notes written on their medical records.


Social network analysis might then uncover that they are all seeing the same physician, even though many of them don't live nearby. One major insurer using both techniques avoided paying $2.1 million in fraudulent claims the first year it used these techniques. In addition, it helped their investigators efficiently narrow the pool of suspect claims. By building predictive models against structured and unstructured (adjuster notes) data, the insurer finds about 100 suspicious cases a week. And 20 of those turn out to be worth further review.

While catching the workers compensation fraudster doing the limbo at a wedding despite a “career-ending” back injury makes for a great story, breaking up organized fraud through analytics is probably going to save more money in the long term. Is it ever worth it to look through public postings on social media sites? Sure. Just don't spend money doing it, until you've invested in more proven techniques.

Stuart Rose is global insurance marketing manager at Cary, N.C.-based SAS. He can be reached at or 919-677-8000.