Fraud investigators have been using technology to turn the tables on scammersand con artists. From claims to underwriting, heres whats working.BY MICHAEL P. VOELKER

Weve made some quantum leaps both in the development of [shared industry] databases and the funding of insurance fraud bureaus within all the state jurisdictions, adds Thomas Mulvey, national director of the special investigation unit for Prudential Property and Casualty Insurance Company, Holmdel, N.J. Ten years ago, if we identified a fraudulent claim, wed pat ourselves on the back and move on. Today, that identification is just the beginning of an investigation, and our ability to detect organized fraud has gone from ground zero to where we are today.

Insurers are doing quite well despite the economics of the industry, particularly toward organized groups, Bryant says. Also, theyre using technology better not only to make sure to identify fraud, but also to pay meritorious claims. Technology has helped insurers focus shoe-leather time on meaningful issues, rather than investigating hundreds of claims individually.
The particular technologies that work for insurers depend in part on what types of fraud theyre most trying to focus on, whether the emphasis is on the application side (prevention), or post-claims (detection and prosecution), and what industry an insurer is in.

Life Issues
Today, the biggest fraud-related blip on life insurers radar is the need to create anti-money-laundering (AML) programs, a requirement of the USA PATRIOT Act (which does not apply to property/casualty carriers). Though final rules and dates are pending, prudent life insurers have been working in anticipation of compliance.

Unlike other types of fraud, money laundering doesnt necessarily cause direct financial loss to an insurer; for example, an annuity payment is the same whether the premium paid is with dirty or clean money. However, not only would a carrier in good conscience not want to be an unwilling participant in a money-laundering scheme, but also the operational risks to insurers are very real.

The degree to which the Treasury will monitor this is unknown, but when a fine gets applied to a firm for having money laundering occurring through its channels, or when it is found to be holding assets for an individual or entity on a terrorist watch list, the reputational impact is huge, says Neil Katkov, analyst at Celent Communications Tokyo office. As a result, many vendors have jumped into the AML space, and Celent recently ranked them in a report Katkov authored, identifying leaders in both transaction-monitoring and watch-list-filtering technologies.
One component of AML programs is matching names against a list of Specially Designated Nationals and Blocked Persons, maintained by the Office of Foreign Assets Control (OFAC), as well as national crime databases to search for activity that might indicate money laundering. Information associations serving the life and health industries, such as MIB Group, have added OFAC name data to the searches they perform for member companies. An insurer also could obtain name data from OFAC in ASCII text form and look for matches using existing data mining technologiesor even obtain the list as a PDF file and do so manually.

But while name matching is in theory rather straightforward, in practice, It is very difficult to find someone whos actually on the OFAC list, Katkov explains. Financial institutions are quite frustrated their system will produce a lot of hits, but very few will represent a person on the list. Theres a heavy work burden on [companies] to look into identification and case history details, and almost every time, it turns out to be someone whos not actually on the list.

Firms also are frustrated OFAC list matching might not get to the core business problem. You dont have Osama Bin Laden walking into an insurance agents office trying to buy term insurance or being named as a beneficiary, says David Olson, visioneering director at MIB Group.

Identifying the real money-laundering activity therefore is a more complicated technological matter, requiring analyzing transactions and looking for either individual red flagssuch as premium payments coming via wire transfer from countries and banks known to be problem spotsor, by pattern matching, a series of transactions or an unexpected connection of data across a book of business that raises suspicion.

The need to connect and analyze data across a variety of internal and external sources poses the problems of integration and data quality with which insurers are all too familiar. AML vendors talk about how they can accept any data from any legacy system, and that is true, but it takes them a long time to do so. So these large-scale implementations can take up to a year and a half to implement, much of which relates to data cleansing, says Katkov. The implementation costs of one of these big AML systems can be more than double the cost of the system itself.

Insurers may be able to reduce costs in an AML program by leveraging existing technologies. For example, vendor CSC, which offers its @First fraud tool, also provides PATRIOT Protector and Compliance Assessment. NetMap Analytics, a data mining and link analysis solution used by many property/casualty carriers, also is used by some U.S. government agencies in their own AML efforts. Additionally, an insurer could itself devise AML rules for existing business intelligence tools already in place.

However, because insurers have less familiarity with money-laundering activity patterns than they do with other types of fraud, they may find they have no choice but to turn to an AML vendor or consultant at some point. A solid understanding of money laundering is essential to create the appropriate rules in an AML technology implementation, Katkov says. Banks have had to comply with these rules for years, and AML vendors therefore have developed some expertise, so its at least instructive for insurers to look at what the newer-generation vendors are offering, such as Searchspace, Mantas, and Net Economy.

More News from the Fraud Front
Money laundering is a type of hard, or criminal, fraud that primarily affects the life insurance industry. In health and property/casualty, incidents of hard fraud often involve organized crime rings that engage in such activities as faked injuries and accidents.

Equally troublingand much more difficult to quantifyis soft fraud: inflating repair bills to cover deductibles, claiming past damage to homes or cars on new claims, or taking more time off work for injuries than is needed. Also perhaps because it lacks the big-bang impact of breaking an organized fraud ring, fighting soft fraud has seen arguably less of a focus among insurers. However, it lends itself particularly well to the application of various business intelligence technologies by virtue of being primarily confined to analyzing internal claims data.

Tracking individual policyholders and comparing claim details against expectations to uncover patterns of abuse actually is more straightforward than uncovering [organized fraud] that involves hundreds of different individuals, says Andrew Bartels, vice president at Forrester Research. An employee who has injured his left shoulder four times, for example, is easier to uncover than four individuals who have had similar shoulder injuries and who are connected only by the same medical provider.

But whether uncovering hard or soft fraud, data mining techniques still are the primary means of doing fraud detection, says Bartels, with recent improvements in technologies such as predictive modeling, similarity searching, and visualization being evolutionary rather than revolutionary. Naturally, industry databases, such as those maintained by MIB and ISO, get bigger every year. MIB adds roughly 50,000 names per week to its dictionary, including language, spelling, and phonetic variations. ISOs ClaimSearch database has grown to include more than 345 million claimsmore than double the amount in 1998and handles nearly 10 million inquiries annually.
Matching services provided by these industry databases also have become more sophisticated. We scrub addresses through the address that the post office delivers to and find matches based on that, as well as the addresses that are reported to us. We also know Sixth Avenue in New York is the same as the Avenue of the Americas, or in another town, Route 9 also might be Main Street, says Vincent M. Cialdella, vice president of ClaimSearch.

We will use NetMap [analysis and visualization software] against our internal databases and then go out to ISO ClaimSearch and determine what a person or group of people are doing in the industry, says Prudentials Mulvey. If, for example, a VIN has been in 10 accidents with 10 insurance companies and was owned by 10 different people, that can indicate a fraud ring is active. Without a technology application to sift through millions of records, you wouldnt be able to make that detection.
Finally, a continued trend toward outsourcing business processes has had an impact on fraud fighting as well, according to Bartels. Part of the benefit of outsourcing claims management is cost reduction in claims, but a secondary benefit is that because [outsourcing firms] are handling tens of thousands of claims daily, they can run analytics on behalf of multiple clients and provide more detailed and sophisticated analyses than a carrier could do if it were analyzing only its own claims data, he explains.

Success Stories
Combating fraud, whether in the application or claims processes, is a careful balancing act between detecting, investigating, and acting upon fraudulent claims while still maintaining acceptable levels of service. The true purpose of our [fraud] project was to ensure we had a superior claims system, MetLifes Sargent explains. While there are a fair amount of suspicious claims that need investigation, the vast majority are legitimate and warrant prompt payment.

MetLife Auto & Home uses CSCs Fraud Investigator, a data mining tool that identifies suspicious relationships between parties involved in claims. MetLife also teamed with CSC to develop @First, an early-detection system that combines predictive data modeling with company claims data and third-party data to flag suspicious claims at the first notice of loss.
We had been looking for several years for technology to better detect suspicious claims, says Sargent. I could find companies that could do modeling or that had a business rules system, but no product that pulled it all together.

Fraud Investigator was deployed in early 2002 after a six-month pilot, relatively quick for a data mining application. Weve had great success with Fraud Investigator, he adds. In the past, where we had missed [fraudulent] cases, today were finding most of them. It also allows us to maximize the resources of our field investigators to allow them to focus on major cases rather than routine data analysis.

The insurer rolled out @First in May 2003, and it has required some post-installation tweaking. For the first three months we worked through some false positives, and there were a lot of volume issues in the learning phase. As a result, we changed our business rules and also dedicated additional analysts to the system. For example, @First initially analyzed all open claims, resulting in many referrals on claims in subrogation where fraud investigative work would be of little value; therefore, MetLife changed the rule set to tighten the focus on actionable claims.
However, increased identification of potential fraud has increased the overall workload in MetLife Auto & Homes SIU department, which was anticipated. We did add staff, but we also are reengineering the way the SIU works, and we now have the ability to handle more cases with inside staff. The increase in referrals certainly has exceeded the increase in staff.
MetLife Auto & Home has yet to measure the benefits of the CSC systems by traditional ROI. It has enabled us to provide better customer service while at the same time being a safety net to ensure suspicious claims dont slip through the cracks. Its been a very good investment, Sargent says.

Zurich Life, Schaumburg, Ill., has had success in combating fraud in the application stage of its life insurance business. All applications undergo MIB checks for medical conditions that were reported on other applications, for application activity using MIBs Insurance Application Index (IAI), and for matching to the OFAC names list. A direct interface between Zurich Lifes new business system, called Z-App, and MIB automatically sends applicant data to MIB, which returns a report to Zurich Life that is presented to the underwriter as part of the application package.
The IAI database has proven effective in detecting a particularly vexing type of fraud, according to Lynn Patterson, chief underwriter and senior vice president at Zurich Life. [Applicants] who wouldnt meet the requirements for a $1 million policy might instead go to 10 companies for $100,000 policies, assuming they wouldnt be underwritten as stringently, he says. Yet at the end of the day, they have $1 million coverage in the industry. Identifying these individuals allows Zurich Life to underwrite these applications more fully to ensure they qualify.

MIB provides a follow-up service to Zurich, relaying information reported by other carriers on inquiries Zurich already submitted. This gives Zurich a second look at policies before they become uncontestable after two years of being in-force.

Zurich also uses the system to look for instances where agents will submit applications to receive higher first-year commission and then let policies lapse after 12 months. When these situations are discovered, we work closely with our marketing department to take the appropriate corrective action, Patterson says.

At an average cost of $2 per inquiry, Patterson reports the MIB system is well worth the investment. If we have 100,000 applications, it pays for itself in just one claim that doesnt occur from a policy we dont write he says. More important, the protective value of the system is so huge I cant imagine operating without it.

At the Maryland Automobile Insurance Fund (MAIF), as you might imagine, we end up with drivers who have a lot of different types of problems and who are exposed to a fairly large amount of fraud, says MAIF SIU manager Joe Asplen. Right now, about 68 percent of our claims get a return from the ISO ClaimSearch database.

MAIF also faces an additional challenge to fighting fraud because, by law, it must insure nearly every driver who applies for coverage, with only nonpayment and nonresidency allowed as reasons for declination. As a result, MAIF relies not only on motor vehicle records to verify addresses, but also databases that are less commonly used in insurance, such as LexisNexis and state property records. Fraud in the application process allows us to void a policy back to inception, even after a claim, so those are important tools, Asplen says.

On the claims front, MAIF chose Bladeworks from Infoglide Software Corporation, which helps it detect fraud not only in individual policies, but also in such fraud as staged accidents, attorney/provider fraud, and crime rings. One of the biggest benefits to Bladeworks is the ability to search on similar terms and receive very accurate results, Asplen says. It also provides the visualization capabilities to help us connect addresses, phone numbers, and other data.

MAIF has seen the impact of the system since its installation. Weve often used it in rate evasion situations. For instance, we recently uncovered a common phone number used across a number of losses, and we realized it actually was a commercial entity that kept changing names to avoid loss hits on a CLUE report, Asplen says.

In years past, it would take some of my senior investigators almost 30 days to pull together all the paper files they needed and to create charts to try to connect the information they had interest in, he continues. Now, we can take those pieces of information we want to look at, analyze them the same day, and quickly determine the cases we want to spend our time on.

The Human Component
Even though the antifraud technologies in place at these insurers can do more analyses faster than any number of human investigators, Its trained investigators who ultimately make the difference in fighting and prosecuting fraud, according to Richard P. Boehning, ISO senior vice president.

And despite the continued evolution of these technologies, they likely will never replace the human component of investigation as long as unpredictable human beings are perpetuating fraud. Sometimes, youre not looking for what is there, but what isnt there, and theres no system right now that will do that, Asplen maintains. There isnt one solution thats the end-all, and I havent seen one yet that can mimic the suspicion and intuition of a seasoned investigator.