Fighting Account Opening Fraud With Big Data
By Robert McGarvey
When it comes to account openings, credit unions increasingly find themselves caught between the proverbial Scylla and Charybdis, the fancy Greek mythology way to say between a rock and a hard place. In either direction lies complete devastation.
That’s because on one side, where credit union account opening barriers are high to prevent bogus accounts opened by crooks, big banks clean up by opening a lot more accounts. Credit unions too often simply drive new business away.
On the other hand, when barriers are lowered, sometimes criminals pile on – like sharks sensing blood in troubled waters – and they get busy defrauding a credit union.
Indications are plentiful that criminals, increasingly, are honing in on account opening fraud. A lot of their activity is shifting into that realm.
Credit unions, seeing that data, have doubled down on erecting barriers to fraud at account opening.
The upshots: potential members are frustrated when account opening barriers are too high and credit unions, too, are frustrated because they know they aren’t opening enough accounts.
There is a big data driven escape hatch that can save credit unions. But lots remain clueless.
They will suffer because what they are presently doing just isn’t current grade.
Personally, I know about the difficulties involved in opening a new account. It took me some weeks, of sheer tedious frustration, to open a new account at a credit union two blocks from my house in Phoenix. If I hadn’t been a credit union fan, and if I weren’t something of a student of credit union practices, I would have walked long before I succeeded in opening the account. Always it seemed another “proof” was required.
My drivers license had a different address than my present address and that was the hitch.
But five years earlier, soon after I had moved to Arizona, I had opened an account at Chase with a driver’s license from another state and, in the bargain, Chase gave me a check for $300, just for opening a checking account. The process was accomplished online. It took a couple days, max.
With the credit union, I eventually surrendered and visited the branch. How 20th century. I did get the account open but the process left a bad taste of obsolescence.
Think about that. Chase paid me to open an account, the credit union inflicted aggravations upon me to do so.
Where will most potential customers wind up?
By now you may be saying, what should the credit union have done differently? What could it have done differently given government insistence on KYC?
Lots.
It could start by joining the 21st century.
Start by reading a very short ebook from Feedzai, a data science company with roots back to Carnegie Mellon University.
It puts forth a lot of frightening ideas. For instance: criminals, increasingly, are opening crooked accounts but leaving them sleeping for some time – typically until a financial institution has stopped watching it as a worrisome new account. Think about that sophistication. The FI eyeballs a new account, crooks discover that, so they wait out the FI.
It gets worse.
Criminals know financial institutions love rules and so they learn them, they obey them – until the FI’s anxieties pass. Then they strike.
Machine learning can still give a credit union the edge, said Feedzai. “Banks can begin to keep pace with fraudsters with hypergranular, continuously updated profiles that pinpoint and connect fraud signals. There can be thousands of fraud indicators that alone aren’t enough to make a conclusive decision of fraud, but which a machine can thread into a complete view and an accurate decision.”
What Feedzai is using is very big data – but that’s how to thwart crooks. It’s found oddities that correlate with criminality. For instance: “Feedzai has found that devices with high battery power are correlated with high fraud. Fraud is three to four percentage points above the average fraud rate when emails have two to four consecutive digits. Certain email domains are more correlated with fraud behavior, including public domains. And when a device name is unknown or null, the likelihood of fraud is 78%.”
Machines can also hunt for data across multiple channels to support, or deny, a new account opening. In a matter of minutes, a machine can make a thin file fat and take a lot of worry out of an account opening.
In assessing a new account, a smart machine looks at both traditional sources (credit bureaus for instance) but also non traditional sources such as social media channels. A full picture results, quickly.
That’s the thing. Many institutions continue to use only a small, limited data set in opening new accounts. Smart, contemporary institutions are using an array of 21st century data and, despite the volume, it is swiftly processed when machines do the heavy lifting.
How does this work in practice? Feedzai pointed to a case study – involving a large bank – where the institution had been approving only 40% of new account applications. With machine learning and big data on the case, it upped that number by 74% – with no increase in fraud. None. And the process itself was streamlined.
Can you say similar about your account onboarding? Talk with any of a growing number of companies such as Feedzai that are targeting account opening fraud and how to do the screening better, faster and smarter. Many companies – mainly small, very tech in orientation – are now in this space. Talk with a few.
If you can’t open accounts as well as big competitors, you won’t long stay in business.
Put the machines on your side and stay in the game.