Grasshopper Bank partners with fintech to outsmart fraudsters

Grasshopper FiVerity.jpg
Grasshopper Bank CEO Mike Butler recently announced he would be partnering with FiVerity and the fraud prevention firm's CEO Greg Woolf for an aggressive gatekeeping operation at account opening.
Complimentary Access Pill
Enjoy complimentary access to top ideas and insights — selected by our editors.

Mike Butler, the CEO of Grasshopper Bank, likens the role of banks in fraud detection to the role of bouncers at bars.

Smart bouncers know a fake ID when they see one, but the ones who don’t care let underage teens drink anyway, Butler said. Whether a business is illegally letting adolescents drink or letting fraudsters open bank accounts, the consequences can be major.

Butler recently announced that the Boston-based fintech FiVerity and its CEO, Greg Woolf, is among the partners he has selected — a bouncer of sorts — for fraud prevention services at his digital bank for small businesses.

New York City-based Grasshopper relaunched last year after hiring Butler as its CEO, and it has expanded from small-business banking into yacht financing and commercial real estate. The bank said recently it has total assets of approximately $300 million.

Woolf said FiVerity differentiates itself by aggregating the fraud signals its customer banks use, without compromising the privacy of legitimate customers. The approach helps identify fraud tactics that criminals use at multiple institutions, allowing FiVerity customers to learn from fraud attempts on other FiVerity customers.

The Federal Deposit Insurance Corp. and the Financial Crimes Enforcement Network recently selected FiVerity to compete in a tech sprint focused on measuring the effectiveness of digital identity proofing. The company is among the 60 selected.

With FiVerity stationed at the door during account opening, Butler said Grasshopper would be able to take “an aggressive approach to fraud detection.” The company has a particular focus on synthetic identity theft, a pernicious kind of fraud in which criminals synthesize real and fabricated identity components to evade conventional fraud detection methods.

Woolf and Butler spoke with American Banker to discuss how they collaborate to fight fraud.


Mike, you said on a recent episode of the American Banker podcast that Grasshopper is a “partner shop” rather than a “build shop,” so you focus on your core expertise, then look around the industry to partner with others on the rest. It does seem like a core business operation of a bank is to prevent fraud, so why would you outsource something so important to FiVerity?

MIKE BUTLER: An interesting evolution inside the industry has been how we interact with clients. In the early stages, people predominantly walked into a branch and met with an individual. That was banks’ way of reducing fraud: through the fact that we interacted individually, in person.

We've quickly adapted as an industry to the world of interacting digitally. Frankly, I don't think the industry has been able to keep pace in terms of competency with the likes of FiVerity and others.

You do want to have a core competency of understanding fraud, but when you look at what we do to avoid it, we use a variety of gates through which we bring clients — know your customer, identification sources. We then use our internal experiences to help us stack those gates in a way that makes meaningful sense for us.

On top of that, we create models with Greg’s help to be able to use data to search for the right clients and avoid fraud. So, we're not downplaying the need to be experts in it. But, we are also recognizing that there are other companies that are far more advanced in fraud detection, who have been doing it for much longer in the digital world.

An important element of training artificial intelligence is providing it lots of data. You need to be able to provide a lot of inputs into the model to be able to get a good output. How does FiVerity do that? Are you only relying on the decisions of Grasshopper analysts looking at what they see as possible fraud? Are you also taking inputs from other companies?

GREG WOOLF: I started my career in one of the big four accounting firms, and through that experience, I learned that ultimately a human analyst — somebody who has years of experience, an education, and a gut feel for what they're looking at — is going to outperform a machine.

We're taking the know-how of the analysts working at Grasshopper and combining that with the know-how of the community of all banks we work with and coming up with a collaborative model that's learning from all the top analysts across the industry.

We've seen in the past our human-in-the-loop model has outperformed your standard “black box” AI model by 70-75% in some cases because the big challenge is that fraudsters are hugely innovative. We have found that humans are very good at changing and adapting.

It’s funny to me, the idea of having analysts feed into this model that cannot feasibly replace them but to some extent is ensuring you don’t need so many analysts to do the job.

Woolf: No, it’s the other way around. We’re not looking to replace the analyst. Analysts are the most important, precious piece of the system.

What we're doing is scaling the analyst. We're allowing Michael to have a team of five analysts that can behave like 50 analysts because his growth plans are so aggressive.

We're enabling technology that allows his core team to grow faster — exponential growth in the number of customers with linear growth in the number of people. That's what he's trying to achieve.

Butler: What we're trying to do with artificial intelligence and some of this really cool data gathering is eliminate or speed up the manual work associated with the transaction, leaving the analyst with the complicated decision, who as Greg says can always beat the machine.

Greg and his team have provided us with this model that comes from a variety of different people, who use the data to help us turn the knobs to get the right outcomes. That's where the analyst becomes so important.

What makes FiVerity different? You can do supervised machine learning on all kinds of things with the right people; what is FiVerity’s secret sauce?

Woolf: There’s two key elements. The first is that we have this collaborative machine learning capability, which is part of our intellectual property. We've figured out how to leverage the know-how of the analyst to create models that can be shared across multiple institutions in a network-type environment.

The second piece, which is really important in the space, is privacy — data privacy and security. If we're sharing patterns and potentially identities of known fraudsters, we have to fall within the regulatory compliance infrastructure that banks are subject to.

The ability to combine the holistic platform with collaborative models in a secure fashion is what's making us interesting and differentiated and frankly, giving us lift in our models.

Any closing thoughts?

Butler: Think about teens using fake IDs to get into bars. The smart bouncer can figure it out; the one who doesn’t care lets them in anyway.

At the end of the day, we have a situation where I’m bringing on clients that could cause me a lot of damage. If I’m the bouncer, if I let in the wrong people, I could lose my liquor license. In this business, I can lose a lot letting the wrong people in, or not letting the right people in.

It’s top of mind that this is what we have to do to evolve — to be able to do business. This is our competitive advantage. Our model is to partner first, build second. We think by partnering with companies like FiVerity, who can help us build models faster and with better data, that will be a huge competitive advantage for us.

Woolf: I'll echo that. I think collaboration is key. One of the themes of a tech sprint in which we are participating is around how industry, government, law enforcement — everybody — can work together on addressing fraud.

Fraudsters are going to collaborate by buying products and services on the dark web; industry needs to work together. The industry needs technology that can learn patterns across the community and share those patterns, and at the same time, secure the information of consumers — U.S. citizens — so that institutions don't run into any kind of regulatory issues.

For reprint and licensing requests for this article, click here.
Fraud prevention Fraud detection Cyber security Technology Digital Identity in 2022 and Beyond
MORE FROM AMERICAN BANKER