The Special Investigations Unit at Central Insurance is known industry-wide for its contributions to the insurance fraud detection process. Jeff Lieberman, Central’s Director of Special Investigations & Recovery, is at the heart of that work.
Over his career, Lieberman has developed strategic partnerships with technology companies, government entities, and even other insurance carriers to unify data, integrate systems, and work together to identify, mitigate, and prevent insurance fraud.
In this article, we dive into the two distinct factors contributing to Lieberman’s fraud analytics system, discuss the extensive benefits of this approach, and explore the impact this one-of-a-kind model has made on the fraud detection process at Central and beyond.
The Two Key Components of Central’s Fraud Detection Model
Lieberman’s fraud investigation model strategically layers a claimant’s historical data with relevant externally sourced information. Below, we dive into each of these components and explore how they interact to provide a full scope of information on each claim and detect potentially fraudulent behavior early on.
Get insights like this right in your inbox. Subscribe to the Central Blog below.
Part #1: Historical Data and Link Analysis in Fraud Detection
Early in his career, Lieberman realized that data could be used strategically in fraud detection. By examining factors like the number of claims, type of claims, and payouts for past claims an individual had filed over a specific period, fraud teams could start identifying informative patterns.
“One of the first things I knew we had to do as an industry was develop a data-forward fraud and subrogation program,” Lieberman says. “So, early on, I partnered with ISO Statistical Service, a Verisk company.”
ISO Statistical Services is an industry-leading company that collects and stores four billion detailed records of insurance premiums collected and losses paid annually to customers into a single database.
When an insured is flagged as potentially fraudulent, Central’s fraud experts begin by referencing ISO’s ClaimDirector analytics program to review an insured’s historical data.
ClaimDirector is a rules-based analytics program that feeds off of the ISO database. It can pull up information about a certain individual’s insurance history based on a series of pre-set rules.
Adding Context: Queries a fraud analyst might run in ClaimDirector include checking to see if an insured has had more than two fires at their home in a year or if they’ve had six or more Workers’ Compensation claims in the past three years.
This information can be crucial in determining a customer’s patterns when it comes to their insurance and has quickly become an integral tool in Central’s fraud detection processes.
Mapping and Analyzing Claimant Connections
Sharing information into systems like ISO not only creates a database from which companies can pull customers’ historical data, but also provides a shared space for carriers to input their historical claimant information.
For Example: Central’s Special Investigations Unit (SIU) frequently contributes its data on customers’ claim histories to the ISO database in hopes that if a fraudulent claimant were to jump from one insurance company to the next, that carrier would still be able to access the customer’s history and identify a fraudulent pattern.
To easily track shared information across carriers, Central uses a tool called Netmap. “Netmap is a link analysis tool that gives us the ability to ingest large amounts of claim data,” Lieberman says. “We can take hundreds of thousands of data sources and input that into our system, and it begins to tell us the story of the customer.”
Lieberman describes Netmap as a digital suspect board in a police station that shows suspects’ photos and has pieces of string stretched between them to signify connections. In the same way, he says, Netmap pulls out information from historical data that shows how different individuals relate to one another.
The system can pull details about claimant “vehicles, what addresses someone was at, where money is going, and who the kingpin in all of that is,” he explains. “It helps us tell the story of this claim more fully, which often leads us to identify organized fraud ring-related activity.”
Did You Know: Lieberman launched a Major Case Program at Central in 2022 as a way to find and put a stop to ring-related insurance fraud specifically.
The integration of ClaimDirector and Netmap has substantially impacted Central’s fraud investigation practices. “We’ve detected a lot from these solutions,” Lieberman says. “Today, close to 40% of our referrals [from the claims department to the fraud unit at Central] come from these automated detection practices.”
Part #2: External Data Sources and Artificial Intelligence
Lieberman’s next step in developing the fraud analytics program at Central was to layer some of the advances in artificial technology and machine learning to his historical data model.
To launch this initiative, he approached France-based AI company Shift, and posed the idea for an integration.
One of Central’s core uses of Shift is to track an issue or accident that resulted in a claim back to its true source. This practice is known as subrogation, and often results in garnering reimbursement for funds lost due to negligence.
For Example: Imagine you are a Central customer who purchases a refrigerator and, two months down the road, the air compressor shorts out and creates a fire that burns down your home. You file a claim to get your home rebuilt, and Central pays that claim so you can get back on your feet. After you’re taken care of, however, our Subrogation Unit will contact the refrigerator manufacturer whose air compressor caused your fire and hold them responsible for paying the claim.
The more data Central’s team contributes to Shift, the more accurate the system is in flagging fraudulent behavior. The two key data points the fraud detection team uses to help teach the system these patterns are the “input” and “output” of a claim. The input is the reason why the claims department referred the case to the Special Investigations Unit in the first place, and the output is the final result of the SIU’s investigation.
“We tell the system if the claim ends up being referred to the Department of Insurance, if it was subrogated against, etc.” Lieberman says. “All those results help teach the computer and the machine learning algorithms the proper detection practices.”
5 External Data Sources and their Impact on Insurance Fraud Prevention
“At this point, we have a lot of different data sources that Shift is constantly analyzing via artificial intelligence,” Lieberman says. “So when a claim gets filed, it interacts with our Shift model and pings out to all those different sources to help determine if something is fraudulent or needs to be subrogated.”
The graphic below represents the wide array of external sources Central’s fraud prevention model currently pulls from. In the next section, we dive deeper into five of these data sources to better understand how they’re being used to help identify and mitigate fraud.
When Central receives a claim from an individual, Shift automatically uses AI to run through data from this national organization. Its database tracks information on any current or past insurance-related crimes and can alert Central if the individual filing the current claim is associated with fraudulent activity.
Central’s model also considers data from TransUnion when determining fraud cases. Because money is at the root of most insurance scams, having insight into a claimant’s current financial standing can provide insight into potential fraud.
“To be clear, we’re not running our claimant’s credit reports or anything like that,” Lieberman explains. “We’re just looking at the information that’s on the public record in regard to their finances, such as liens, judgments, bankruptcies, criminal convictions specific to white collar crimes, or if they are undergoing any form of financial distress that would lead them to commit a crime.”
Another data source that has proved crucial to Central’s fraud detection processes is the Geospatial Insurance Consortium. Described by Lieberman as “Google Earth on steroids,” this group’s Geospatial tool provides aerial images and geospatial information for insurers. The strategic use of these images in fraud detection “has separated us from all insurance carriers in the industry,” Lieberman says.
“No carrier had ever done it before. We were the first ones, and GIC saw tremendous value in that,” Lieberman continues. “I helped them bring the two industries together, and began to develop that integration into the model we’re currently using today.”
Central uses the low-altitude, high-resolution photos from Geospatial to help determine fraud on claims that might otherwise be hard to mitigate. For example, if a customer claims that their roof suffered damage during a hurricane, this historical imagery can prove whether or not that is true.
“Especially in catastrophic losses [such as a national weather emergency], the planes that document these images go up right away,” Lieberman says. “But they’re also consistently flying and taking photos of every area of the country so that we have historical imagery we can use in a claim dispute.”
Data collected from the National Recall Database can be used to help determine the root cause of a loss. When Central’s claims representatives gather information from a claimant on a fire caused by a washing machine malfunction, for example, they are trained to ask for the make and model of the appliance. Central’s AI model can then run that information through the National Recall database and immediately report on whether there’s been a national recall alert on that item. If there is, that often results in subrogation of the claim.
5. Social Media
Central also uses social media as a third-party data source when investigating insurance fraud. Specifically, the SIU leverages Skopenow—an AI software used to search, collect, and analyze open-source data—to review information or images a claimant might share on social media.
“If a claimant says they had a slip and fall at a hardware store and sustained all these injuries, [Skopenow] is going to sift through that person’s social media automatically,” Lieberman says. “It will look through their Facebook, Twitter, [and] LinkedIn, and look for images or mentions of that person at a yoga class or downhill skiing…basically them doing anything that proves they filed a false claim.”
Information collected via Skopenow is then routed back to the Shift system and taken into account when determining a claim’s fraud status.
Top 5 Benefits of Central’s Fraud Detection Program
While the most significant benefit of such a well-established fraud analytics system is the ability to stop insurance fraud in its tracks, there are other positive outcomes from Central’s investment in this cause. Below, we explore five of the most substantial Lieberman has experienced working in the Special Investigations Unit.
Benefit #1: Efficiency
While many companies are already utilizing data sources in their fraud analytics, Lieberman points out that few are taking advantage of the automation capabilities available through current AI technology. It is these automation systems, however, that are defining the fraud detection process for Central.
“The manual aspects of fraud detection can take a lot of time,” he explains. “It’s always been easy enough to run a comprehensive report, but the time you would then have to spend deciphering it really adds up.”
By adopting a fraud detection system that’s always working in the background, Central has been able to reallocate resources. Now, SIU members have the time to handle fraudulent cases instead of drowning in endless files and reports.
“We don’t have to be the ones to search for recalls anymore, for example,” Lieberman says. “Our fraud analytics program is doing it with AI instead. This leads to early detection, which means better protection of everyone involved.”
Benefit #2: Accuracy
Central’s fraud analytics program increases accuracy across the board. Not only does it allow the SIU to cross-reference information across a plethora of new and in-depth data sources, it also eliminates the potential for human error or misinterpretation of information.
As a result, the team has developed a higher alert score and enhanced their credibility as fraud detectors in the field.
Benefit #3: Financial Savings
Fraud detection programs that capitalize on data analytics and automated systems can bring major financial savings. “From an efficiency standpoint and a cost perspective, the benefits of automation are astronomical,” Lieberman says.
First, this type of analytics model reduces the need for employees who must manually handle data, which reduces hiring costs. Additionally, a more accurate and efficient detection program results in identifying and mitigating more fraudulent claims.
“We’re saving money because we’re not paying the claims that are fraudulent,” Lieberman explains, adding that the more cases of fraud that can be stopped, the more money an insurance carrier can save in the long run.
Benefit #4: Early Detection
Central’s fraud analytics program allows the SIU to address flagged claims much sooner than the average carrier.
“Our system gives us the ability to validate a person’s claim, which is of the utmost importance,” Lieberman says. “This early detection of questionable claims allows us to better protect the assets of not only the company but our policyholders, as well.”
From an efficiency standpoint, he adds that it’s “much easier to detect a fraudulent claim early rather than after we’ve already paid it.”
Benefit #5: A More Unified Industry
In bringing together data from multiple sources—including other insurance carriers—Lieberman and his team have connected multiple groups with different perspectives and outcomes under one common goal: stopping insurance fraud.
What’s more, where other carriers might keep such a successful fraud detection model to themselves for a competitive advantage, Central is committed to sharing the discoveries the Special Investigations Unit has made and the ground they have covered in the industry. Lieberman is especially passionate about continuing to develop partnerships and systems that work together to detect fraud.
The Future of Central’s Fraud Detection Model
Upon his arrival at Central in January 2019, Lieberman took stock of the standard approach to fraud analytics and determined more could be done. This inspired him to develop the signature layered approach of historical claims data and automation that defines Central’s state-of-the-art fraud analytics program today.
Of his success in the industry, Lieberman is quick to say that “anyone can start an analytics program or go buy an off-the-shelf platform; it’s thinking outside of the box that has continued to set us apart.”
And while the existing fraud detection model is already proving incredibly effective at identifying potential insurance scams, Lieberman knows there is more work to be done.
“A fraud detection model needs to constantly adapt to new fraud trends and incorporate new fraud-fighting tools to help it evolve with the times,” he says.
In fact, Central’s team is already in the process of developing and improving its AI model.
“The more data sources we can bring into it, the better our [detection practices] will be,” Lieberman says. “There is so much data out there, and I want to make sure we’re using everything at our disposal to put an end to insurance fraud.”