How Real Time Fraud Analytics Enables Real Time Decisions

When talking to senior counter-fraud professionals, we frequently hear that 20% of their staff detects 80% of all claims fraud; and the interesting part is that generally there’s no discernible difference between the portfolios of the employees. Some claims handlers, either through training or instinct, simply seem to have a better nose for potential fraud. When FRISS works with insurers, one of our main goals is to create consistency for these companies. As Gartner research indicates, holistic fraud management must include the entire policy lifecycle. That’s why we help prevent and detect fraud across the whole portfolio. Our question is, though- what is the other 80% of that staff missing that we can help solve?   As we all know, fraudsters are continuously evolving and evading even the best detection efforts from insurers. The experienced ones use their agility to stay ahead of traditional companies playing catch-up, and the number of steps ahead pretty much entirely depends on the quality of technology the insurer is able to deploy.  

What Features Should I Be Looking For? 

The best detection combines a number of these resources:

1. Models to detect potentially fraudulent behaviour 

2. Feedback loops to improve accuracy over time 

3. Expert knowledge rules 

4. Artificial intelligence 

5. Text mining (and natural language processing) 

6. Device recognition 

7. Network analytics 

Given the speed of the evolution of the fraudsters, the insurer needs to be able to use these technologies to make a real-time decision on a claim. Often though, the detection is based solely on a batch process, typically on a daily or weekly basis, against either an AI model or the network analytics. As customer demands for straight-through processing continue to increase, insurers are forced to deliver on that technology. Without such, flaws in the traditional process can be easily exploited by today’s savvy fraudsters.  

So, is moving to a real-time decision-making the only answer? Well…yes (and no). Obviously, being able to make a real-time decisions is a step in the right direction, but the key point is that the conclusions must be based on relevant data that is updated in real time. For example, a real-time decision based on network analytics updated the week prior is a weakness asking to be exploited.  

But What About Data Sources? 

To really amp up capability and create an holistic view on potential risks, it’s crucial to access both internal and external data sources. 

Internal data sources can come in the form of:

1. Internal Document Store – data rendered from Optical Character Recognition to be used for spotting the language patterns associated with fraud and entities from within the documents that are not yet captured within the core system.  

2. Image Screening – taking all the images submitted with a claim and checking for consistency and anomalies.  

External Data Sources may look like:

1. Contributory fraud data 

2. Claim and policy data 

3. Credit data 

Main Takeaways 

However, something to keep in mind is that the most important factor in assessing such data is adopting a champion/challenger mode, making sure the data is predictive and adds value. Knowing if/at what point of the claim the data adds value will help to manage the cost of the data source and maximize the ROI.   By harnessing data sources in this fashion, we can create the ever-improving cycle of fraud detection, reacting to new typologies and levering the data in real-time to inform a fast, and accurate, decision. This allows your claims handlers to “Be Brilliant In The Moment” when your customers need them, allowing them to deal with claims promptly at the moment of truth.