In the past, insurers could write off fraud expenses with investment income and capital reserves. But those days are over. Reduced investment income and reserves have forced insurers to face such operational bugaboos as fraud and subrogation head on. Fortunately, insurers ahead of the curve have identified technology-based strategies to get to the heart of the matter.
At Novato, Calif.-based Fireman's Fund Insurance Co., claims adjusters are expected to reconcile claims rapidly. Therefore, adjusters might overlook irregularities associated with a claim.
If an adjuster is unable to detect traces of fraud when assigned to an auto or workers' compensation claim, Fireman's Fund's special investigation unit (SIU) experts can't adequately do their jobs. And even though Fireman's Fund historically has experienced a low preponderance of fraud, the carrier adheres to a no-tolerance attitude where even marginal instances of fraud can be enough to undermine its bottom line.
Three years ago, Fireman's made a commitment to leverage its data to enhance the tactical aspects of its business through the mastery of business intelligence. "In the highly competitive property and casualty insurance market, it is critical to have a clear understanding of all profitability factors," declares Marty Ellingsworth, director of operations research at Fireman's Fund. "We use predictive models to address loss from fraud and subrogation."
With business intelligence strategies in place, claims fraud at Fireman's Fund-particularly for commercial and personal auto coverage as well as workers' compensation-is becoming less of an operational sore spot each year.
Clean data
Becoming better practitioners of fraud reduction and subrogation enhancement didn't occur overnight. Similar to many insurers, Fireman's Fund once faced a challenge of operating from multiple and disparate legacy systems. This made enterprisewide data mining efforts an exercise in futility.
"Property/casualty insurers have relied on traditional ways to fight fraud, such as using ISO's ClaimsSearch database, but they can't optimize this effort when operating with antiquated systems," says Ritu Jain, global industry strategist for insurance for SAS, Cary, N.C.
"An SIU team can't do its job working under these circumstances. It all adds up to an effort that can't properly ferret-out claims fraud."
The numbers speak for themselves: In examing claims payouts that many P&C insurers face, they relinquish on average about 10% to 15% of claims on undetected fraud or failing to recoup subrogation revenue, Jain states.
"Data mining is essential to discovering anomalies along with unusual associations. If insurers could reduce fraud even 2% a year it would make a large difference in their operations," Jain explains. "The last couple of years, declining investment income has hurt insurers' ability to subsidize their fraud losses," so insurers must devise others means to turn things around.
Data warehousing is the linchpin to best-practices data mining. "Insurers have to implement a plan that will produce clean data," says Jain. "A carrier might have multiple and disparate legacy systems that all house data that an SIU team would need to tie together, but is unable to. On our end, we can develop access engines that link these systems, but I recommend that insurers start with clean data first. That helps maximize the power of our solutions."
In a recent independent survey of its European user group, SAS revealed that 66% of organizations indicated that "dirty data" impacts company profitability, and that 74% of companies in the financial services and telecommunication industries are acting or planning to act this year to improve data quality.
To further reinforce the relevance of quality data, a survey of 2,000 attendees at SAS' recent European Users Confer-ence in Vienna, Austria, revealed that 99% of respondents viewed accurate data as essential or important in the ability of business intelligence applications to deliver ROI.
Fireman's Fund, a $4.3 billion (written premiums) business and personal insurance provider and subsidiary of Allianz AG of Munich, put itself in position over the past three years to implement effective fraud and subrogation efforts through internal systems integration-unifying as many as seven data systems that are now interlinked. This in turn enables the insurer to produce an enterprisewide view of data.
Fireman's Fund licensed a business intelligence product, Enterprise Miner, developed by SAS. The insurer uses SAS solutions for virtually all its financial and operational reporting, and harnesses Enterprise Miner to improve cash flow, and control losses and costs, Ellingsworth says.
Over the years, Fireman's Fund has invested in technology supported by IBM Global Services along with SAS, so as it conducted due diligence on selecting a business intelligence platform, the decision to license Enterprise Miner seemed to be a natural fit. "We did research in 2001 and early 2002 and validated Enterprise Miner in pilot phase. We went live with the program in September 2002," says Ellingsworth.
Fireman's Fund deploys a business intelligence team internally that conducts ongoing analytics to reduce fraud and improve subrogation. Consisting of a team of four individuals, the business intelligence team deploys an analytical toolset that resides internally within the insurer's operations.
"It's a GUI-based drag-and-drop system that provides reports within a batch-delivery mode. Eventually, we hope to have it deployed to provide real-time data exchange," Ellingsworth says.
Miner details
Across all of its many applications, Ellingsworth estimates that Enterprise Miner has added $20 million to $30 million a year to the company's bottom line in cost savings and increased revenue via financial and operational reporting competencies.
Designed with highly efficient neural network capabilities, decision trees, self-organizing maps and algorithms that provide segmentation and cluster analysis to recognize claims fraud, Enterprise Miner functions as a "backstop to uncover activity that a claims adjuster might otherwise miss," adds Ellingsworth.
The solution can be licensed and hosted by the insurer's internal system-the route Fireman's Fund has selected-or it can be deployed as an ASP model where SAS conducts ongoing reports and provides results directly to the insurer.
Indeed, the inroads that Fireman's is making via the one-two punch of fraud reduction and subrogation enhancement can't be denied. It estimates annual fraud-reduction savings of $700,000.
Enterprise Miner is able to detect patterns in huge volumes of claims data that may reveal suspicious claims. The insurer also uses text mining to spot words associated with fraud.
"When these cases are automatically flagged and sent straight to investigators and specialists, both worker productivity and customer satisfaction increase," Ellingsworth says. "Adjusters spend their time on the appropriate cases and do not dwell on cases that require investigators. Investigators can apply their skills where needed instead of randomly looking through files for fraud.
"In addition, customers with low-risk scores for potential fraud receive claims payments quicker, increasing customer satisfaction through timely service."
Detecting likely subrogation
Fireman's Fund has also realized cost savings with SAS predictive models by detecting likely subrogation-the process by which an insurer can recoup claims payouts from a legally liable third party. Thus far, Fireman's has identified about $2 million in subrogation revenue per year-money that would have been lost since most claims had been closed.
Ellingsworth identifies ease of use as another added benefit of Enterprise Miner, estimating that his unit can build predictive models 20% to 30% faster with the help of the intuitive and graphical nature of the solution. This enables better use of staff resources, less duplication and saves time.
And the good news is that even novice programmers can confidently deploy the solution without feeling overwhelmed. "The data mining capabilities of Enterprise Miner help point a programmer in the right direction to make assumptions about fraud," says Jain.
"The system makes it extremely simple and quick for even an inexperienced IT programmer to generate results. It was designed to benefit both IT programmers and business users."
"Overall, the system has really been integral to our success in ferreting out fraud. We've been able to rely on SAS solutions to intelligently drive business decisions. Our aggressive use of these products enables us to protect our marketplace and to improve our brand," Ellingsworth concludes.
