Fraud Rings Beware

Fraud has been, and remains, a daunting and costly problem for Canadian property and casualty insurers. This is especially the case when the perpetrators are a moving target, with fraudulent individuals and rings continuously evolving and modifying schemes to execute false claims.

For the most part, insurers have focused their efforts on reducing insurance fraud through analysis of data on the claims themselves. However, a rich source of data has been largely untapped by the industry: personal attributable data on claimants and on other individuals related to the claim or to the claimant.

By incorporating analytic technologies to perform real-time searches across an enterprise’s disparate data, insurers can find, match and link similar entities and uncover hidden relationships between people, places and things. This is the future of fraud detection: focusing on finding patterns that cross multiple claims and claimants to uncover abuse or expose fraud rings.

A GROWING CONCERN

It is undeniable that there is a growing trend of organized fraud, collusion and opportunistic fraud. The insurance industry has been particularly exposed because detection has traditionally been difficult and prosecution can be expensive and challenging.

Often these losses have been perceived as a “cost of doing business.” Yet given the magnitude of losses and industry-wide interest in cost containment, this view is changing – attitudes in the insurance industry are evolving from acceptance that fraud is a “cost of doing business” to a more aggressive stance on proactive prevention and engagement.

Insurance companies across the globe routinely release figures demonstrating how prevalent the issue is. For example, in April 2014, United Kingdom-based provider Aviva reported a 19% rise in fraudulent claims in 2013 compared to the previous year. The company noted it is currently investigating 5,500 suspicious injury claims linked to known fraud rings in the U.K., which represents a 20% increase in the number investigated last year.

REAL OR NOT?

What makes a fraudulent claim successful is that it is often sophisticatedly disguised as a real claim.

Among the common problem areas found to be trending across many insurance lines of business are identity fraud, collusion between victims and service providers, and collusion among parties in accidents across multiple policies and companies for the purposes of driving benefits claims.

While there is proven worth in technology that helps evaluate the likelihood of a claim being fraudulent, identifying suspicious information on individuals related to claims can lead directly to a strong reduction of losses from organized fraud rings as well.

With multiple individuals across multiple entities involved in ongoing fraudulent claims, the “fingerprints” of organized fraud rings can be found across disparate databases in the form of personal attribute information. For example, as members of a ring attempt to defraud various carriers, they use and leave behind fictitious – and shared – personal information, such as addresses, phone numbers and licence numbers.

The bad news is that every major insurance carrier experiences numerous types of organized fraud on a daily basis; the good news is that data on many of the participating fraudsters can be captured. 

BEYOND DETECTION

To leverage the value of data on fraud ring culprits, as well as fraudulent individuals, it is essential for insurers to quickly access as many internal and external data sources as possible, and match the data attributes that serve as the “golden nuggets” of financial crime prevention and investigation.

In terms of fundamental technological power, the key to realizing a meaningful reduction of losses from fraudsters comes from two very distinct abilities:

• the ability to quickly and efficiently access a breadth of data – and the capabilities to conduct thousands of queries per day on hundreds of millions of records across dozens of disparate databases; and

• the ability to understand the identity matches and non-obvious relationships between individuals across dozens of data sources – despite input errors or deliberate attempts to deceive.

As technologies have enhanced and developed the ability to focus on the claimant as well as the claim, insurers can now strengthen their current analysis of claims with an added perspective on risk associated with individuals.

Applied reactively in regular claims processing – or proactively in off-line reviews – the technology uniquely gives insurers a seamless, deeper view into fraud risk.

However, improving detection is just the first step. Insurance companies should consider preparing themselves organizationally as well, such as defining priorities and objectives, providing tools to ensure their investigation teams can handle increased volumes, and developing new policies and procedures.

REMOVING SILOS AND BARRIERS

For the most part, regulators are concerned about privacy, particularly around personalized consumer data.

In addition, companies are often concerned about sharing customer data with competitors. This has traditionally worked in favour of fraudsters who benefit from siloed data to hide their activities across the industry.

In order to work with the regulators, many fraud-detection models are developed using depersonalized data. These can be implemented such that a full cross-review of data can be achieved without the need for replicating or moving data to centralized locations, which can be sensitive both from privacy and from competitive perspectives.

While working closely alongside their technology partners, insurers have developed methodologies and mathematical solutions to analyze the data available to them. Two such methods are cross database identity resolution and link analysis, details of which follow.

Cross data identify identity resolution

One of the first steps when investigating the claimant is to resolve the identity of the claimant. Is this claimant someone who appears, perhaps with slight variations to personal attribute information, on other claims, either as a claimant or as another third-party related to a claim?

If yes, how strong are the matches? Can they be relied on to deem the claimant suspicious, and to stop payment to the claimant? The hope is to determine “who’s really who” by finding, matching and linking similar people across disparate data sources.

Technological advances have led to the creation of algorithms that can overcome data barriers, created by clerical errors, linguistic differences and purposeful misrepresentations that otherwise would make it difficult for typical investigative tools to find similar, but not exact, matches across databases.

The algorithms accurately find matches on people, places and things, even whether or not they are only partially similar.

Link analysis

If an insurer has found reason to suspect a claimant of possible fraud, the next step would be to investigate the claimant’s relationships with individuals listed on the claim.

Through the use of link analysis, information can be gathered by investigating complex webs of evidence and drawing conclusions that are not apparent from any single piece of information. These methods are equally useful for creating variables that can be combined with structured data sources to improve automated decision-making processes.

EXPLORING UNEXPLORED DATA

Data is information that, when accurately examined, can provide insight that drives beneficial action. The insurance industry needs to take additional actions to stop the significant problem of losses caused by individual fraudsters and organized crime rings.

One way that insurers can do so is by taking advantage of a vast amount of data that, to date, has gone largely unexplored. Personal attribut e data on claimants and other individuals can be found across a variety of data sources – much of it within insurers’ own databases, as well as accessible external sources.