Companies in many industries are extolling the benefits of data analytics to generate new business opportunities and better serve customers. For insurers, there's an additional benefit: early detection of fraudulent activity. Insurers are leveraging a growing number of data sources - including social media - and applying high-performance analytics to help detect patterns of fraud as early as possible.

Insurance fraud affects not only every insurance company, "but virtually every consumer and taxpayer worldwide, and it shows no sign of easing," according to a 2013 report by research and advisory firm Aite Group. The firm estimates that claims fraud in the U.S. property/casualty industry alone cost carriers $64 billion in 2012 and will reach $80 billion by 2015.

Claims fraud detection is an area where insurance companies will "drastically" increase investments in the next three to five years, says Nicolas Michellod, senior analyst with the insurance group at research and advisory firm Celent, and author of the firm's 2013 report "The Market Dynamics of Claims Fraud Detection."

The pace of insurers' investments in claims fraud detection technology will depend on their understanding of the importance of analyzing data using modern technologies, Michellod says. A survey of 178 insurers conducted by Celent in 2013 shows that just under 40 percent have already invested in tools supporting fraud detection, and more than 40 percent have invested in predictive analytics systems.

"Growing numbers of insurers are adopting high-performance analytics such as predictive analysis," says James Quiggle, director of communications at the Coalition Against Insurance Fraud, a Washington-based anti-fraud alliance.

"Higher-power programs compact the time-space continuum for claims," Quiggle says. "Insurers can spot claims far earlier in the cycle, in close to real time. Investigations can begin faster, thus reducing payouts for bogus claims."

Nationwide Mutual Insurance Co. in Columbus, Ohio, "believes in the intelligent combination of data and analytics to detect and deter fraud from the point of sale to claim," says Rick Wahls, claims director, special investigations unit (SIU) at Nationwide.

"For example, early detection of masked or hijacked IP [Internet Protocol] addresses along with IP addresses originating from foreign countries can help identify fraud rings at the point of sale," Wahls says. "Fraudsters have behavioral patterns that once learned can be potentially identified as early as during the quote process."

Being able to see the entire view of interactions between customers and Nationwide enables the company to use predictive models to identify patterns within a claim that could only be generated by computer analytics, Wahls says. "Viewing as part of an entire ecosystem of customer interactions and activities makes predictability of fraud more of a science than a gut feel or just following one-off red flags," he says.

Nationwide has used a multi-layered approach to fighting fraud since the SIU's inception in 1985. "We do not feel that a one-tool technology solution fits our needs," Wahls says. The firm has a diverse product line and leverages independent and exclusive agents, brokers, affinity relationships and the Internet to sell insurance.

Due to these distribution channels and offerings, Nationwide uses process, technology and people to create an integrated system to identify and detect and mitigate fraud, Wahls says.

"At the heart of our fraud fighting system is the belief in a zero tolerance to fraud," Wahls adds. "This belief system enlists, motivates and engages associates from all parts of the company in our efforts to make Nationwide as fraud free as possible."

Sources of data to detect fraud can be from anywhere that makes sense and is permissible, Wahls says. "Insurance is a heavily regulated industry and compliance is a priority," he says. "Some of the emerging sources include mounted camera data, satellite photometry, facial recognition, voice recognition, cyber resources, text and vocal mining, third party data. And better leveraging our own data can help in the detection of fraud."

Nationwide has invested significantly in its ability to obtain real-time data to identify patterns and behaviors and integrate the data into a system that can help the company make sense of the output.

Commercial P&C insurer CNA began using Fraud Framework from SAS, a predictive analytics program for fraud detection, in 2011. SAS built predictive models for CNA's main four lines of business: workers' compensation, general liability, commercial auto and commercial property, says Tim Wolfe, AVP of the SIU at CNA.

"SAS built models based on our own historical data, such as medical bill and financial data, so they could see who we paid and how much," Wolfe says. Also added to the models was data from CNA's central claims platform, including information from completed fraud investigations, as well as data from the company's SIU case management system.

"The models are built on factors common to known fraud claims," Wolfe says. The SAS solution has two main facets: scoring individual claims for fraud potential, and scoring provider networks. SIU managers are alerted via a user interface of suspicious claims or networks following the analytical process.

CNA feeds data to the models on a weekly basis. "When we come in to work on Monday morning we go through the new alerts to review both the claims and the networks that score high for fraud potential," Wolfe says. "We determine which ones are viable for investigation, then assign them to investigators."

The key driver for adopting the SAS technology was to gain more granular information to more effectively detect fraud. "We knew we were seeing a certain amount of fraud referred by our claims adjustors, but we didn't know how much we were missing," Wolfe says.

Industry research indicates 10 percent of all claims contain an element of fraud, Wolfe says, and as of the end of 2010 CNA was seeing just 3.7 percent of its claims referred as potential fraud. "That's considerably below 10 percent, and we wanted to find out how much fraud we were missing that wasn't identified by adjustors," he says. Data analytics "gave us another tool to detect fraud that was flying under the radar."


With the help of analytics technology, CNA has seen the ratio of cases flagged for potential fraud rise to 7.6 percent as of October 2013. With the analytics models, "we can sort out legitimate claims more quickly, and we save money by not paying for fraudulent claims," Wolfe says. He says the analytics system has saved CNA $4.25 million in the 21 months since it was fully deployed, including money not paid for fraudulent claims across all lines of business.

In another anti-fraud effort, CNA is participating in a data aggregation program organized by the National Insurance Crime Bureau (NICB), a not-for-profit organization that partners with insurers and law enforcement to facilitate detection and prosecution of insurance criminals, and Insurance Services Office, which offers information about property/casualty insurance risk.

Under the program, medical claims data from 22 participating carriers is subjected to a sophisticated analytical process to detect potential fraud. The organizations collect data, such as the identity and specialty of the medical provider, the nature and length of treatments and the amounts paid.

"There might be questionable activity or fraud which might not be apparent in our data sets, but when you [combine data] from other carriers, patterns start to emerge," Wolfe says. He says in the past insurers have been reluctant to share such information because of legal risks. But NICB addressed that concern by "depersonalizing" all patient information when submitting it through the analytics process.

Another insurer, MetLife Inc., integrates external public records and other available industry aggregated data into an "analytical toolbox" for potential fraud evaluation, and to help adjusters in the claims evaluation and mitigation process, says John Sargent, director of MetLife's SIU department.

"This information and analysis is integrated into our claims-handling system at first notice of loss and throughout the claim process, to support the adjuster for the appropriate claims-handling strategy and investigative resource assignments," Sargent says.

MetLife uses several proprietary applications along with tools in the Microsoft Office Suite and social media search engines to gain a comprehensive evaluation of all available data, Sargent says.

"Our ability to quickly identify suspicious claims for a specific evaluative strategy while providing prompt, expeditious claim service is critical," Sargent says. "We owe it to our policyholders to keep the impact of claims fraud from adversely affecting their premiums, while satisfying customer service expectations."

The analytical approach improves the company's ability to identify questionable claims more quickly, Sargent says, and use the available data to strengthen initial suspicions or mitigate fraud.

"The integration of critical and meaningful data from all available sources across all transactional systems, in a real-time process, has great potential to identify potential fraud activity sooner and further expedite the claims customer service process," Sargent says.

Social media is playing a bigger role in the detection of fraud. "Social networking analytics can capture and synthesize seemingly random and often hidden unstructured data," Quiggle says. "Sifting through social media for clues has gone beyond the state of a gee-whiz new tool. It's now becoming a routine, must-have detection device for many insurers."

In the coming months, more vendors will be offering data aggregation tools or services to make it easier for insurers to tap into social networks to find information that might help support the fraud mitigation process, Celent's Michellod says.

"The industry has begun to recognize the significant value of social media communications, event data recorder information and telematics as critical and timely to thoroughly evaluate suspected claim fraud," MetLife's Sargent says. "This data can be used to help confirm loss description, injury severity and also show possible relationships indicating the collaboration of individuals in organized fraud activity."

Bob Violino is a business editor and writer who covers a variety of technology and business topics.

Register or login for access to this item and much more

All Digital Insurance content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access