Is Fraud a Solved Problem?

Over the weekend I was listening to one of the sessions from LendIt Fintech USA 2018. The name of the session was Speed Matters: Making Real Time Credit Decisions which is pretty self-explanatory as to the topic. It was a really insightful panel with so many interesting nuggets of information. But one statement stood out to me. Jeff Stewart, the Chairman and Co-Founder of LenddoEFL said that “fraud is a solved problem”.

That is quite a bold statement. So, I reached out to Jeff yesterday to get some more color on what he really means here. He stood by what he said on the panel at LendIt. While we can’t get rid of 100% of fraud what we can do is catch fake identities, fraud rings, and large-scale theft of identity.

“The key is to make fraud unscalable, we want to make the cost of committing the fraud too much,” Jeff said in our phone call. “There is so much data in a person’s digital footprint today and the algorithms have become so good at analyzing this data that fake identities stand out.”

Better Algorithms Means Better Fraud Detection

It has become too difficult to accurately simulate real identities on a large scale. The algorithms have so much data that they can recognize unnatural patterns of interactions. Now, this doesn’t mean there is no fraud at all. There will still be nefarious individuals with reasonable credit who take out a loan with no intention of paying it back. Jeff added, “Out of character, ‘internal intent’ is much more difficult for AI to predict, because after the fact it is hard to know what was in the mind of the applicant but fraud on the other hand simply requires good data forensics after a default, so the behavior improves the fraud AI.”

Work is being done here as well with behavioral assessments and psychometric testing that can help root out the propensity to commit fraud. LenddoEFL calls it “algorithmic character”.  Jeff also shared interesting work being done in voice stress analysis. An example of this is where an voice enabled AI engine can call an applicant to confirm application details and assess the stress level in the person’s voice as they have the conversation.  Yet another tool that can be used in the long running game of cat and mouse between AI and fraudsters.

In today’s privacy-focused world I should point out LenddoEFL makes sure that users give permission to access their digital footprint. This is done during the loan application process. They also ensure that this data is never shared with anyone. And they only use this data for fraud detection and credit scoring, something that regulators appreciate in the various countries they operate in.

Fraud With Fake Identities is Rampant in China

It is not surprising that the type and amount of fraud varies between countries. We learned from Thomas Wang of China Rapid Finance in this same session that a staggering 97% of loan applications in China are fraudulent. Think about that for a moment. Only 3% of the applications that a Chinese online lender receives is from a real person. The rest is fraudulent activity often from established fraud rings.

So, if Chinese lenders were not good at fraud detection they would all be out of business. Another interesting thing that Thomas said was that the way he is fighting fraud today is different to the way it was done six months ago. His company has to evolve as the criminals seek new avenues to commit fraud.

The cat and mouse game of fraud/fraud detection will continue but given the success of companies like LenddoEFL and China Rapid Finance (as well as many others) we are winning the battle. While it is not unheard of to learn of a company that has failed due to problems with borrower fraud it is certainly not commonplace, even in China.

The challenge used to be that there was not enough data on the underserved to make good credit decisions. With the proliferation of smartphones all over the world that is no longer the case. We have an ocean of data on the majority of the world’s population. The challenge now is analyzing this data most effectively.

Getting fraud down to zero is probably an unachievable goal but stopping large scale fraud rings is now possible. This is what Jeff means when he says that fraud is a solved problem. It doesn’t mean we can sit back and relax, we need to keep innovating. But there are solutions available today to help lenders win the fight against institutional fraud.

 

Subscribe
Notify of
0 Comments
Inline Feedbacks
View all comments