Fraud detection is in many ways a numbers game, but that doesn’t mean that more data means more accuracy. As Nuno Sebastião, CEO of Feedzai, recently discussed with MPD CEO Karen Webster, the most effective fraud detection technology might be that which takes a micro view of a very macro problem, using intelligence and insight to address the issue on a personal level.
KW: Feedzai recently shared some exciting news, having raised $17.5 million to really scale what is already a powerful fraud detection platform. Two questions come to mind. First, what did investors find particularly attractive about your platform, given that the space is now crowded with billions of players that are clambering for visibility? Second, what do you plan to do with the money?
NS: The first question is one we’ve been asking ourselves — what do we have in our business that really sets us apart from the competition?
There is a need for a technology that can really solve the problem. I would say that a lot of the solutions out there, they’re working along one angle: that being, “I’ll get some sort of data at the social level and whatnot, and then I’ll try to use it to understand whether or not it’s fraudulent."
We come at it from a completely different angle. Ours is a purely technological approach that is based on particular clients’ data — social data, for example — used to understand what behaviors are indicative of fraud. While there is a lot of singularity among players — merchants, for instance — the fact is that each one has a specific set of data. We focus more on the technological engine rather than having a ton of data to then share with everyone.
Our numbers demonstrate what kind of results we can get, and that materializes into clients and some very long contracts. Our average contract is five years, and since we operate on a yearly revenue model, that is very attractive for investors, in the sense that we can forecast a five-year horizon on how much money we’re going to be making. It’s a business model that can sustain long-term growth and revenue; last year, for example, we had a 300 percent year-over-year growth. And this year, we’re going to sustain that rate — that’s why investors wrote us a check.
KW: You said those investors found value in feedback because your technology can solve “the” problem. What is "the" problem? There are so many of them — cybersecurity, keeping data safe, keeping networks and systems safe. What’s the problem that you solve?
NS: We started by solving just the transactional risk fraud. Then, we realized that the internal data we were using on a client could help assess not only a particular transaction but also how it compared to previous behaviors from that client, as well as to different types of clients. That helped us realize that what we actually do is risk management in a wider scenario.
Now we have use cases in which we’re utilized for online account opening, where we determine not only whether or not a person should be allowed to come into the system, but the degree of risk in that decision. We also work with acquirers who offer our services as a tool to their merchants, where it becomes a management system that allows merchants to understand whether or not they should ship a particular item, as well as allowing them to perform continuous merchant underwriting — continuously tracking merchant behavior to assess potential fraud.
KW: What is really very unique, then, is that you’re not looking at millions of data points that aggregate a spectrum of clients. You’re evaluating for a particular customer, based on his or her patterns of behavior in a specific environment.
NS: That’s correct, and that’s why our target clients are ones that do a volume business.
That’s not to say that we don’t use external data sources to enrich our process. What we’ve demonstrated, rather, is that there is more value in some internal data that clients are currently willing to share, more than what they share for consortium purposes.
We’re able to do a better job than just using the so-called consortium data that is what every client is willing to give to everyone. We’ve seen that in some of the organizations we work with, there’s such a richness of information inside that, just by using that — and, of course, by going in some external data sources — we’re able know the customer, to profile him or her and understand if what’s happening is some sort of deviation that might be indicative of fraud. We consistently are able to avoid what is the plague of this industry — instances of false positives.
KW: When you look at the specific client — for lack of a better term — engine, or database, do you look for patterns that might be helpful to inform the action for other clients? Do you use client data to inform other client data?
NS: It depends. We deploy our client’s data center. We don’t even have access to the production machines. Some clients — typically smaller ones — do allow us to share. What we do not share is the data itself; rather, we share the measurements that we’ve done on the data — what we call profiles — which in and of themselves contain no personal information and no bank statistics information. The measurements provide us with patterns that allow us to take care of the privacy concerns of the client.
KW: What are some of the outcomes that a client has achieved in working with you and deploying your technology to their situation? We talk about this as being a much broader solution than just fraud detection, really more risk management as well. Give us a sense of some of the ROI a client has shared with you as a result of working with you.
NS: I can address that without disclosing any information about a particular client.
Today, the rejection rate for opening an online account, a deposit account, or trying to get insurance online is really quite high — a typical range is from 40 to 60 percent. That’s because the screening process is not really good, and available information is lacking. Imagine if you’re able to go from a 40 to 60 percent rejection rate to one where 80 or 90 percent of the people trying to get those online services are successful. That can be achieved, and it can be done by managing their risk and the kinds of things that they can or cannot do. And that’s the kind of situation that provides a measurable ROI.
The types of use cases in which we’re limiting transactional fraud are all about false positives. The game is how much more fraud can you detect without detracting the good people, the good behaviors, that not only impact the network you’re working with, but also impact reputations.
Another example is that, on the acquiring side, our technology enabled us to detect for a particular client that a merchant was going to go bad many days in advance. That’s a very valuable service, especially as so many people are going after chart track management. We’ve actually had conversations with people who tell us that we’re not charging enough.
KW: $17.5 million — that’s a big check. How are you planning to deploy the funds?
NS: I’ve been working with some of the largest organizations out there. One thing we cannot do is screw up, in terms of support and SLA, so the funding is going to be used to really build top-notch support for our clients.
We’re also going to really strengthen our sales and our local market team. We are a company with over 60 people, now; we have two salespeople, so really it’s been more inbound than outbound. We’re looking to change that; I think we’re now at a stage where we can start to really scale that side of the business, so the product is built, tested, and demonstrates the highest possible level of performance. With that, we can get into business with more clients faster.
We’re going to open an office in New York; we’ll be covering other regions, as well. We’re looking at India because we have a very big client who is based there.
With all of that going on, it starts to make the $17.5 million not look like much anymore.