The Power to Predict
Posted March 22, 2018
In this episode Armen is joined by Frank Teruel, CFO of ThreatMetrix. They unpack the theme of the Digital Identity Summit 2018: The Power to Predict.
Armen: Hi, thanks for joining today’s episode of Digital Identity 360. I am sitting here with Frank Teruel, CFO of ThreatMetrix, and we are here today to talk about the power to predict. Pretty dramatic, huh?
Frank: Very dramatic. Hey everybody. It’s a pleasure.
Armen: So “The Power to Predict” is the theme for our upcoming ThreatMetrix Digital Identity Summit series. No coincidence that we selected this theme. There’s a lot of anchor points around the ThreatMetrix business, some product investments that we’ve made, some of the trends that we are seeing in the marketplace around predictive analytics. And, I thought it would be good to focus on that theme as it relates to this episode. Ready to dive into it?
Frank: Yeah, for sure. It’s interesting, Armen, if you think about the power to predict and what we’ve done as a company, think of the idea that for many, many years we’ve been able to instill confidence scores about people. I’m fairly confident that when you draw up onto a website, you’re Armen. When you use an app, you’re Armen, based on what we do here as a company, but adding this element of trust. You know, I have confidence Armen is Armen, but can I trust Armen based on his behavior right now in a global, real time engine?
Armen: Well, hopefully, the answer to both of those is, “Yes.”
Frank: Yes. Hopefully. It makes it really cool and interesting is you can’t accomplish that unless you’ve got this massive data set that in real time is refreshing and stopping. Cause I can arrive in confidence that you are who you say you are, but I don’t know how you behave unless you’ve got a holistic global view of your behavior online. So, very cool, very interesting, and very predictive.
Armen: Yeah, we talk about deterministic versus probabilistic … It’s kind of the fusion of both. Where, yes we can determine at a confidence level that we absolutely know who you are-
Frank: That’s correct.
Armen: Or we know that you are not who you say you are. And then the trust component is dynamic. It’s based on signals that we’re seeing across the network. There is a bit of probabilistic, power to predict there. So I think that metaphor is a great one.
Frank: For sure.
Armen: In many ways, really defines ThreatMetrix moving forward. We introduced this big product innovation called ThreatMetrix ID at the last summit series. That really, I would say has set in motion this new focus on ThreatMetrix driving these decisions with somewhat disparate data points that we’re bringing together.
Frank: Yeah. Absolutely. The digital ID is interesting because it visualizes the first time the interrelationships between you and different kinds of transactional elements. And I think once you saw those elements, customers look at it and say, “Wow, I now have a view, optically, of is Armen compromised or not? And that really led to this idea that I can be predictive, just by simply looking at someone’s identity and going “Wow, it’s changed dramatically.” Yes, it’s Armen, confidence, but wow … It’s this device and credential combination is being used all over the place in real time. Obviously, my confidence is that it’s Armen, but my trust is very low because you’ve been compromised. So, I think that visualization was the very first set when people that aren’t customers looking at it and saying, “Wow, I can actually see for the first time this incredible relationship between the transactional elements.” It’s very cool.
Armen: Yeah, and when you think about the use cases, there’s a very broad array of use cases. You know, obviously, we are not part of the broader ThreatMetrix … I’m sorry, the Lexus Nexus Risk Solutions business where there’s a broad array of industry applicability, use case applicability. Even sort of transaction aside, right, with a very large considered mortgage/loan/home loan all the way to this long tail of digital transactions where we have historically played really well. This power to predict really can play across the spectrum, driving informed decisions.
Frank: For sure, think of the continuum of authentication. On one end, is where we play, and at some point dollar value and risk dictates that you do something else, right? We’re very good at real time decisions, but at some point people get nervous and say, “Hey, Armen’s applying for a rocket loan, he needs a mortgage in seven minutes.” Am I gonna simply relay on this digital attribution as a mechanism to authenticate Armen? The cool part of combining physical attribution is it really bolsters the trust score. So think back to the confidence versus trust. Armen’s applying for a loan, which is great. I now have confidence that this is Armen, but I know because of physical attribution that can increase my trust in that person. Because I’ve now been able, in an anonymous fashion … still tying that attribution back to digital identity, and that allows us to kind of cover the whole spectrum of no step-ups required, no further outside sources required. We can handle the entire thing using that attribution.
Armen: Yeah, that’s a powerful combination and this was part of the messaging when the two companies combined. I think it does open up the possibility for just even broader applicability. Especially with some of these maybe larger more considered transactions that aren’t purely digitally native. So-
Frank: By the way, transactions have to happen fast, right-
Armen: They still have to happen fast.
Frank: If you look at lending, for example, where the imperative is I want to get a loan out in the marketplace in less than seven minutes. I mean, in that kind of real speed, real time decisioning on a very large transaction with high risk, our digital identity with physical attribution really strengthens that decision.
Armen: Yeah. So, another dimension to this conversation is, you know, you had mentioned smart authentication … Again, that continuum of risk-based authentication versus strong authentication, bound to a device or bound to something certain. The reality is, there’s a spectrum there. The philosophy that we’ve brought to the table at ThreatMetrix is “Hey, depending upon the circumstances and the context, you need an array.” You need an array of choices with which to authenticate the customer in the moment. And, I think that very much defines where we are today from a product perspective. When you mash up these physical attributes with the digital attributes, it provides really an endless array of possibilities for how we authenticate in the digital age. You’re gonna cross a spectrum of use cases.
Frank: There’s a feeling, Armen, that strong can replace risk-based. And the answer is, it can’t. There’s an inherent vulnerability in strong attribution, or strong authentication. If I buy a new cell phone, and I’ve screened your account, you credential information, I can now associate my biometric on that new phone with your information. And so simply relying on strong is a false sense of security. When I bring to bear risk-based authentication, what we do … the probabilistic approach in determining whether you really are Armen … There are situations where strong will help, but it helps in concert with risk-based authentication. If you do it by itself, it’s a real risk.
Armen: It is. I think all roads point to having a broad set of capabilities that are rooted in what I would call a “digital first” framework.
Frank: That’s right.
Armen: It provides a really strong foundation to serve our customer community, which is what we’ve been doing for many, many years. I think about all the online lenders that we serve. You gave the example of “Hey, how do you deliver a lending decision in seven minutes or less?” It’s pretty amazing, and many of our customers doing this at scale, and you can’t do that just with traditional forms of authentication alone, right? You’ve got to match that up with digital.
Frank: Think of the numbers, Armen. We protect today. Billions of dollars that our lending community has put into the marketplace are being protected and issued into the market using our approach, our digital identities and our Digital Identity Network. It’s powerful and it allows people in real time to make those decisions, for sure.
Armen: Yeah. I guess sort of shifting gears a little bit, I want to think about some of the broader Lexus Nexus Risk Solutions. Just cooperate messaging, around hey, you know there’s a higher order value they’re bringing to the world. Around helping the under-banked around the world where there might be some digital markers and fewer physical, traditional, no home address, no ship-to address … How do you enable those folks to participate in the global economy; the digital economy?
Frank: In virtually every millennial, right? (laughing)
Armen: Yeah, exactly-
Frank: Who has established credit? Think of the old paradigm of “I established credit based on how long I’ve been at my address, do I have a landline, how long have I worked at one job.” You have an entire category of under-banked people that aren’t necessarily destitute. We tend to associate under-banked with underprivileged, and it’s not the case. Many under-banked people are simply people who haven’t established credit. So, what we allow our customers to do is to look at their behavior across the global network, and see these are in fact real people, with real credit. You can trust them. Other customers are allowing them to authenticate, to transact with them. You should have that same approach, and it really allows us to layer on a broader coverage to people like our lending customers, who want to establish and reach those markets.
Armen: It’s a powerful message, right? Above enabling the digital economy for our customers, we’re really serving a higher order purpose here, which is kinda cool when you think about what we do and the markets that we serve. Real life changing activities. I guess one final topic related to all this … the power to predict. Obviously there’s a big AI and machine learning dimension to this topic. These are words that are heavily used and maybe even overused in the industry. What’s your perspective on that? The role of good machine-learning with respect to predictive analytics.
Frank: Sure. I think it’s essential, here’s why. Think of the evolution of our company. It took us four years to get to fifty million transactions a day. It took us 18 months to go from fifty million to a hundred million transactions a day, and 72 days to go from a hundred to a hundred and fifty million. There’s this massive tsunami of data that hits us daily. The ability to interact with that data in a physical world is impractical and probably impossible. So, what good machine learning does is it helps us build and understand those trends, right? It helps us use algorithms and machines and science to say “I’m seeing vectors that are manifesting themselves across our global network, and I can respond to those vectors with machine learning in a much better way than if I had practitioners in their accounts and try to figure this out by themselves.” The other thing it does for us is very, very cool. Once you’ve learned those trends, it normalizes the trust score. So, behavior that otherwise physically might appear anomalous. What we can do is say, you know, this isn’t anomalous. The machine has learned, has realized it’s real. If Armen decides for the very first time to send money to Costa Rica, and you’ve never done it before, the machine realizes this is no longer anomalous behavior, it’s part of Armen’s behavior once we’ve authenticated using our system. Machine learning helps us to be quicker, helps us to be more accurate, helps us to address bigger problems in real time and in many cases helps our customers because we can address those problems and see them way before they ever see them.
Armen: Yeah, and I’d say it helps us scale. You drew a picture of the force multiplier that we’ve been seeing in the network growth.
Frank: Massive and fast.
Armen: It’s massive and fast in having good learning systems that can adapt to what is and what is not anomalous behavior, it’s pretty powerful stuff. Just to put a bow on this, these themes that we’re talking about are very much going to be themes that are going to be present at our Digital Identity Summit coming up.
Frank: Please show up!
Armen: May in Paris and then September in Los Angeles, we’re looking forward to having our customer community there, our business partner community, the extended LNRS, the ThreatMetrix community. It’s gonna be a fun time and some really good learning content there.
Frank: A powerful theme, applicable theme, I think a theme that will make our customers be able to interact with their customers quicker, better, and have more trust.
Armen: Excellent. Alright, well you heard it here, please join us at the Digital Identity Summit. Thanks very much to the audience.
Frank: Thanks everyone.