Credit unions (CUs) face a delicate balancing act when it comes fighting fraud — a specific needle they need to thread, CO-OP Financial Services’ SVP of Fraud Products Fotis Konstantinidis told PYMNTS in a recent conversation.
On the one hand, banks and CUs want to keep fraudsters out and fraudulent transactions down. That is the obvious part. On the other hand, though, Konstantinidis noted, they’re trying to preserve and enhance the customer’s experience, and an overly sensitive fraud detection system runs the risk of turning up false-positives that lead to card declines and other things that make the customer’s journey unpleasant.
“And it can depend on the credit unions and their specific needs and goals,” he said. “You have credit unions that are more concerned about the member experience, and false-positives worry them more. They don’t want to bombard the customer with calls and emails over innocent transactions. They may prefer to absorb more fraud losses so they don’t create a clunky member experience. On the other side of the spectrum, you have credit unions [that] would rather create more false-positives and weed out as much fraud as possible.”
CO-OP serves 3,000 credit unions, the vast majority of which are between both extremes — CUs less committed to one way of thinking or the other, and most interested in hitting the right balance. They want to catch as much of the fraud as possible in the most efficient way, while keeping the experience on the customer end as friction-free as possible.
Trying to hit the right balance lies in that effort, Konstantinidis noted. Artificial intelligence (AI) can do its best work in “coming to [the] rescue,” in many cases, by making it easier for institutions to understand their customers better — and, thus, better evaluate what is and isn’t a suspicious transaction.
CO-OP’s offering in this area is called COOPER, a platform that offers up what Konstantinidis described as a “menu of options” for its CU partners when it comes to leveraging machine learning (ML) as a fraud-fighting tool. That menu has just seen a new entry: the launch of the COOPER Fraud Analyzer, a new rules engine and case management tool that makes fraud easier to spot and address in real-time.
While fighting fraud is key, Konstantinidis added, it is really only the beginning of the opportunity for credit unions when taking an ML approach to fighting fraud. That’s because the data required to do better, smarter authentication of transactions can also teach CUs an awful lot about their customers — and how to service them better.
A Changing Fraud Front
The security picture has changed quite a bit for small banks, even in the last few years, Konstantinidis noted. There are more threats out there by volume and there are emerging threats they haven’t seen before.
“With the rise of P2P, real-time payments, different networks opening up over APIs and open banking, fraudsters are the first one[s] to see those changes from a payments standpoint. They adopt them, and then fraud patterns change,” he said.
On top of that, he added, there is a host of new channels opening. There are standard card payments, not to mention payments on mobile devices, wearables, smart devices, the web — this list, he said, keeps getting longer. With the predictions about the Internet of Things (IoT) laid out as they are, the reasonable expectation is that those channels are going to keep proliferating, which means the list is going to keep getting longer for quite some time.
That means, he said, we are moving into the realm where human intervention is being stretched to the limits of its usefulness. At some point, there is so much data pouring in from so many channels that even the best human fraud fighters in the world are going to be overwhelmed. Forget fighting the fraud, he noted. It will be too hard to even spot the new patterns as they emerge.
“Fraudsters are working hard to build computational ability to attack your financial institutions and [generate] a lot of fraud losses. This is where static rules are always reactive and not proactive,” he said.
That is the bad news.
The good news is that with AI intervention, those rules can be more proactive, because the AI can do more than try to apply rules transaction by transaction and, instead, take a more holistic view of the payments flow.
The Fraud Fighter
Rules are a perfect place to start — and the new COOPER Fraud Analyzer tool begins with a set of rules generated by security experts. That tool, he noted, is now built in with an automatic function for opening cases, so that when transactions are flagged, organizations can immediately open a case and move to customer outreach in determining if a transaction was real or fraudulent.
The bigger picture plan for the tool is for the rules to engage themselves and essentially be written by the AI. As the tool runs through more transactions and “learns,” Konstantinidis explained, the Fraud Analyzer is designed to be a flexible rules engine that can add new rules as the AI discovers and logs new fraud patterns in the market.
Today, he noted, the Fraud Analyzer tool is in pilot with five credit unions, though it is set to roll out to about a dozen more in the near term. The goal is to see the system online as an option for all 5,800 credit unions that CO-OP serves, though the firm wants to make sure the engine is fine-tuned so the experience of both the CUs and their members is where it’s supposed to be.
The early feedback has been strong, he noted, but making sure the AI engine is perfected and writing good rules is critical before it’s out and about in the wild. More critically, he said, CO-OP needs to make sure that the AI offered can live up to the firm’s larger vision for it, which is helping credit unions understand their own customers better.
More Data, More Insight
There are insights that CO-OP can make available to credit unions simply by giving them data that they have been able, traditionally, to access when it comes to fraud. As an example, CUs couldn’t go back to their security providers and ask why a specific transaction was denied — as that sort of information was part of an algorithmic black box.
But this data is something that CO-OP offers, with a full profile of organized data about member transactions. That’s because, Konstantinidis noted, the future of fighting fraud isn’t just about battling thieves; it’s about helping credit unions connect better overall with their members. Knowing how members spend, where they spend, what devices they use, how much they buy — those are useful data points in verifying a transaction, he said. However, that data is broadly useful for much more than that.
“Understanding members is really the North Star in so many ways, and fraud is just a single use case. There is also cross-selling products like mortgage or auto loans, or offering credit lines or consumer counseling. All of this stuff is a through line within member data and understanding,” he said.
Seeing fraudsters for who they are is critical for a credit union, and offering a security solution has to, at base, be able to offer that. The potential in that, he noted, isn’t only in seeing crooks better, but getting a much clearer look at legitimate customers.