10% of us is demonstrably lying online
Imagine you knew that a specific user repeatedly deleted his description in the field "Describe the insured event" and rewrote it again. Or - regarding car insurance - that user has changed the location of the insured event three times with a distance of over 1 km from previously changed values? You won't probably quickly approve those claims, will you?
But how to recognize suspicious behavior or even a lie in real-time?
Insurance companies usually use Machine Learning or Artificial Intelligence to detect potentially fraudulent claims in the whole process. However, these models are usually built on rule-based scoring and at the same time use only "hard" personal data and compare it for example with the IP address or third party registries.
While this may also lead to the detection of something suspicious, there is still a big blind spot in monitoring user flow while filling in an online application. And nowadays the rate of frauds coming out from online reporting is up to 40%! Well, you want to look for other solutions.
Device fingerprint is a good point to start with
There are AI-based products (like Zoe.AI) that are able to discover and tell you a lot about a particular user as soon as he enters an online form, without filling anything in yet. Now we are talking about the primary device and online data.
From these outputs, we compose a unique ID which we call device fingerprint. This serves to identify a specific visitor in the online environment despite the fact that he comes to the form several times repeatedly and changes his personal information. That is a good starting point to be able to use behavioral prescoring.
To be continued by a text about How to identify suspicious aspects of online behavior.