Fighting Fraud in Real Time: 3 Recommendations

Insurers are obsessed with cycle time. They count the days it takes to make a claim adjudication decision, the minutes it takes to complete the loss intake process and the seconds it takes to process a transaction. Especially in high-volume environments, time is money.

In the wisdom of insurance claims executives, faster claim payments generally equate to better customer satisfaction and loyalty. Anything that slows the process is burdensome and costly.

Of course, accuracy is important too. Just printing checks for anyone who calls in to the claim center would be quick, but not terribly accurate. So the key to all great claims organizations is to strike the right balance between speed and accuracy.

When it comes to fraud investigation, historically the process has been anything but quick. Many organizations still rely on a manual process where adjusters identify red flags and notify the Special Investigation Unit (SIU) by email or even a paper form. There is some sort of triage process, and then an assignment to an investigator within the company or from a vendor partner. And then of course, it takes time to do the actual investigation: Schedule and reschedule appointments for interviews, track down witnesses, review evidence and document the findings.

James Ruotolo

Many organizations are implementing analytics to help streamline the fraud detection process. The Coalition Against Insurance Fraud reports that more than 80% of US insurers are using some kind of fraud detection technology today, and nearly one-third expect increases in technology budgets with predictive models being a top area of investment.

How fast is too fast?

Analytical fraud detection models provide insurers with a great advantage. They are ever-vigilant, always scanning the data and not letting anything fall through the cracks. They can quickly identify risk flags in new information as it is added to a claim file. Models can look across large numbers of claims to see patterns and identify relationships that would not be detectable by a human.

But one of the greatest benefits is speed. Optimized models can scan an entire book of business very quickly. Inevitably, when implementing this technology, the subject of "real-time" processing will come up. While speed is a key benefit of analytics, insurers must be mindful of how the analytics will be deployed. Insurers should define what is meant by "real-time." For many technology providers, real-time transaction processing involves sub-second response times most often used in credit card processing. While it is possible to design a similar solution for the insurance environment, it is expensive and often unnecessary. When evaluating processing needs, insurers should ask themselves a key question: Even if it is possible to get a "real-time" response, are we prepared to consume the results in real time? If not, other less expensive options might be preferable.

[Prevoiusly from Ruotolo: 3 Places to Look for Underwriting Fraud Risk]

When considering options for implementation of an analytical fraud detection platform, there are a few options for processing. Real-time processing provides instantaneous response times, often measured in milliseconds. It is generally appropriate for high-volume transactions with a limited number of highly consistent variables when an immediate decision is required.

Near-real-time processing provides a short delay in response time, often measured in seconds or minutes. It can often be done by running intra-day batch cycles.

Batch processing is generally used when processing time takes minutes, hours, or even days. It can accommodate very large or very complex data and computations at a reasonable cost.

In many implementations, a combination of these approaches can be used. For example, fraud scoring on long-tail workers' compensation claims could be run on a weekly batch basis while short-tail auto damage claims could use an intra-day batch process that runs every 15 minutes.

Another approach is to use batch processing for very complex calculations like advanced network building and text analytics while using faster processing engines for transactional claim scoring.

Where speed really matters

In the claims environment, most organizations can get by with batch processing for fraud detection scoring. However, there are several places where real-time detection can pay big dividends. Here are three recommendations:

1. Point of sale processing. During the application and underwriting process, especially with more insurers expanding their direct Internet channel, real-time interdiction is critical. If a high-risk application is flagged in real time, it can be routed for more thorough validation and underwriting before unnecessary risk is taken.

2. First Notice of Loss (FNOL) processing. Insurers are always looking to streamline their loss intake process. For claims accuracy and customer satisfaction purposes, it is best to get the correct resources assigned to the claim as fast as possible. During intake, it would be advantageous to run fraud risk scoring models, which could direct the intake processor to ask for additional information or automatically route the file to a more experienced adjuster or SIU contact.

3. Claims workflow processing. As information is added to the file throughout the life cycle of a claim, new decisions need to be made. Supporting resources for medical case management, SIU, subrogation, etc. need to be assigned. This is traditionally handled by libraries of business rules. But more advanced analytical approaches can predict the need for these types of resources. Implementing real-time analytics that work in conjunction with claim system workflow processing engines can yield the best results.

The future is faster

Technology is ever-changing, and the current focus on big data analytics is driving innovation, especially in the ability to process large and complex data sets very quickly. High performance analytics takes advantage of improvements in grid computing, in-database and in-memory processing, and low-cost commodity hardware. In the future, insurers may not have to worry as much about the tradeoffs between speed and cost. But for now, it pays to make an informed decision.

About the author: James Ruotolo is an insurance fraud technologist, thought leader and the principal for insurance fraud solutions at SAS. Connect with him on Twitter.

[Moving to the cloud is not done overnight. Where is your insurance company when it comes to migrating to the cloud? Learn how one enterprise is making the shift. Attend the An Enterprise Journey to Hybrid Cloud session at Interop 2014 in Las Vegas, March 31-April 4.
You can also REGISTER FOR INTEROP HERE.]