The insurance landscape looks significantly different today than it did 20, 15, or even 10 years ago. However, one thing remains constant: fraud. In the industry, fraud occurs at all stages from the initial application to the claims process.
By far the most common type of fraud is claims fraud. This occurs when a policyholder makes a claim that never happened, was staged, or exaggerated. It can happen with personal injuries, such as slip-and-fall claims, or include a policyholder’s property such as their car or home. No matter the circumstance, insurance fraud not only impacts companies but that company’s everyday, non-fraudulent customers. According to the FBI, the total cost of non-health insurance fraud is estimated to be more than $40 billion a year in the US, costing the average family between $400 and $700 per year in increased premiums.1
While this issue has persisted over time, insurance companies today have more efficient tools at their disposal to deal with it. In many ways, data is improving the insurance industry in every way from customer retention and talent acquisition to mitigating risk and enhancing the customer experience. It even has the power to help companies more efficiently detect and prevent fraudulent claims as long as companies know how to use it properly.
First, we outline some of the most common problems insurance companies have regarding fraud detection and prevention. Then, we offer four specific ways data could be leveraged to improve these practices.
There are several areas of concern in regards to companies being able to effectively detect or prevent insurance fraud. First, there’s the issue of the mass amounts of data companies receive from a variety of sources (direct, aggregator, agent, broker, bancassurer, etc.) Most of this is often first party data, which is not always reliable and allows fraudulent claims to slip through the cracks.
Although, it’s important to remember that for many insurance companies, the amount of data is not the issue. It’s the organization of it. Too often we see unstandardized data that makes it difficult for teams to manage and integrate it with variable third-party sources. When this is the case, it is easier for fraud to go unnoticed.
Finally, perhaps the biggest issue that has allowed fraud to continue for years in the industry is the fact that insurers are just human. Human beings can only detect so much. Those who commit insurance fraud count on this, but fortunately, there are now non-human tools available to help insurers catch what the regular human eye cannot.
The Covid-19 pandemic did not help the issue. When pandemic relief funds started in spring 2020, insurance companies started to feel the pressure of the changing profile of claims. As it became more difficult to predict and segment claims, many insurance organizations ended up too weighed down and stretched too thin. This allowed people a greater opportunity to commit fraud.
In today’s world, there is a significant amount of data at our fingertips. It is up to insurance companies to leverage it effectively if they want to tackle the issues surrounding fraud detection and prevention. Here are four solutions that Data Axle employs to help Insurtech companies better utilize data to support insurance companies in their anti-fraud strategy.
When it comes to tackling the issues of unreliable first party data, insurance companies should consider data matching, or data enhancement, to help detect and prevent fraud more easily. This requires matching and appending a company’s customer data on file to a verifiable third-party source. It allows companies to find information and any potential discrepancies in a mass amount of data by creating a more accurate prediction of a customer profile.
If you are an insurance company that sells individual coverage such as home or auto insurance, you would be using data matching for demographics such as gender, age, location, marital status, number of children, income, etc. If you’re a B2B insurance company selling blocks of policy, data enhancement for you would involve variables such as company size, business type, sales volume, etc.
Oftentimes, people who commit fraud use an excessive amount of records to camouflage their activity, but data matching allows an insurance company to see through it with a clearer eye for fraud detection.
When it comes to improving data quality and accuracy, standardized data is critical. It doesn’t matter how great the data is, if you can’t make sense of it. That’s why Insurtech companies should seek out a qualified source for standardized data.
When working with a data provider like Data Axle, Insurtech companies want to take the provider’s standardized third-party data and match it on top of the company’s existing data to identify risk and ultimately prevent fraud. Unorganized data on either end can slow down the process and make it more difficult to gain relevant anti-fraud insights from the integration.
When choosing a source, look for a data provider to help smooth out the process. They should offer comprehensive customer service, technical support, and proper documentation to seamlessly onboard new data and ensure smooth product operation.
In today’s world, data changes quickly, and out-of-date data is expensive. The cost has been well documented over the years. In fact, IBM estimates that the cost of poor data quality to be around $3.1 trillion a year in the United States alone.2 This number is staggering, and while there is room for variation in it, insurance companies need to be aware of how important this is; and why data accuracy is key in not only identifying and detecting fraud, but also preventing it.
When choosing a data partner, look for a source that has an exhaustive verification process for the data they collect. Outdated data can not only lead to inaccurate policies or quotes, it can also open the door for fraudulent claims to sneak by unnoticed.
Unsurprisingly, fraudulent claims have gone unnoticed for years simply due to human error. After all, insurers and even data scientists are just human. Now, there are tools, like data analytics, to help detect what they cannot. This is revolutionizing the insurance industry across the board, but in particular, it can be useful in helping detect and prevent fraud with predictive modeling from machine learning.
With machine learning, data scientists can build predictive models that produce fraud propensity scores. Then, the system is able to score new applications and claims for their likelihood to be fraudulent. Once enough history is built up to allow for an increase in statistical significance, predictive modeling can be extremely effective. For example, predictive modeling is able to help insurers more quickly spot application fraud, arguably the most common type committed. It works across all types of insurances. Whether someone is providing false medical information on a life insurance claim or over-exaggerating the worth of their 10 year old car, AI is able to flag it for insurance companies to take a closer look at.
A study from the Certified Fraud Examiners found that in 2019 only 13% of organizations use machine learning to detect and deter fraud.3 However, it is clear that AI is gaining traction across the industry. We’re already seeing predictive modeling help insurers prevent negative events from happening in the first place.
For example, look at home insurance. Using AI, insurers are able to map weather and geographical data to pinpoint potential hazards for water leaks. This way they are able to quickly spot leaks and provide corrective action. As a result, costs will naturally be lowered. The model can also help assess a claim for fraud based on the leak’s location.
Insurance companies do not have to settle for traditional approaches to fraud detection and prevention. Instead, they can leverage data and technology to more effectively mitigate the risk of fraud across the entire process.
Tanner Lerdahl is an experienced business development professional. Tanner's expertise includes managing sales pipeline and communication with prospective customers to help better understand their data applications as well as meet their business needs. He has a Bachelor’s of Science in Business Administration and Marketing from Aurora University.