Real-Time Fraud Detection & Exabyte Analytics

MICHAEL KRIGSMAN: We’re discussing real-time data with Eric Haller from Experian. Eric, how are you? It’s great to see you again. 

ERIC HALLER: Hey, same here. It’s good to be back on your show. 

MICHAEL KRIGSMAN: Eric, tell us about Experian, and tell us about your role. 

Sure. Well, Experian is a global information services company, we do business in over 40 countries, have about 18,000 employees, and are in a lot of different industries, but you might know us mostly by credit. And here in the United States we own one of the three major US credit bureaus that’s consumer-facing. 

And so most people know us by that. But we’re in a lot of different markets. It can be auto, health care, marketing, across the board. My role is I’m a business leader. I’m executive vice president, I manage our identity and fraud P&L globally, as well as R&D, which we call our DataLabs and I manage DataLabs around our globe. 

MICHAEL KRIGSMAN: So obviously, real-time data is central to your business. 

ERIC HALLER: Absolutely. In any of the enterprises, we do a lot of work on behalf of our customers behind the scenes, where having up to the minute, the freshest, most accurate information available, is critical to making the best decision at that time. 

Particularly when it comes to credit risk or fraud, as the world goes into a digital environment, that information to be able to collect, gather, and analyze, and respond back literally in milliseconds, becomes a critical part of the process. 

MICHAEL KRIGSMAN: So the real-time aspect is then absolutely crucial? 

ERIC HALLER: Yeah, so you have to think about just data in general. And if you’re we’ll say, geek in the data world, you kind of segment out the world into things that you have time to think about, and things where you’ve got to answer in the moment. 

And so in the world of answering in the moment, some of that information is captured, analyzed, collected, and teed up. It takes a lot of time to do that, so it’s all teed up, so when you have to make a decision, it’s ready and available right away. In a lot of our decisions, that’s a big part of it. 

But there are some decisions like in the world of fraud, and in a digital environment, where the actual data that’s captured as a part of the transaction or part of that process, is absolutely critical in making the best decision at the time. So you’re taking what’s happening now in the moment, and what you’ve known about in some kind of historical perspective, and analyzing, scoring it, and coming up with the decision. 

And that’s the nexus of what I call real-time. Many decisions are made in the moment. It gets more challenging and more complicated, and where we have to really exercise I’ll say some analytical and technical muscle, to actually leverage information that’s being captured as part of that process, where that’s part of the scoring mechanism or the machine learning process of arriving at the right decision. 

MICHAEL KRIGSMAN: So at a high level then, the real-time aspect is vitally important for certain segments of your business, more importantly for what your customers are doing. How much data are we talking about? What’s the kind of scale that you operate at? 

ERIC HALLER: Well, again, it’s really different types of ways of looking at the data. So like in R&D, we’re very comfortable leveraging large data sets, terabytes of data. I think collectively, our four labs are probably dealing with about an exabyte of data regularly across four labs, that’s about 1,000 petabytes. So we’re used to using large amounts of data. But that’s typically when you’re analyzing a large corpus of information to try to extract out certain behaviors that you can leverage. Like, hey, this is a people group that has never been lent to before in Africa. So how much information can we capture, so we can extract a signal, so we can understand how to lend to them? 

In that real-time environment that you are talking to, the payloads are smaller but they’re very, very concise. Like every data element is important. So the amount of data is not a lot. Even if it’s 1,000 bytes of data, which doesn’t sound like a lot, because there is a lot of information there that you’re going to have to try to make a decision on. 

The sheer magnitude of data in a real-time environment might not be as large as kind of the breadth of information that’s coming in, and I’ll give you an example. So kind of as a consumer, you’re buying your item on the internet, you’ve gone through, you get to the shopping cart. 

And what you might not realize, is there’s a lot of things happening behind the scenes real-time. And that can be from, hey, we’re looking at the device that you’re using. And there’s information, the characteristics that are captured is just part of a one device communicating with another on the internet, and they’re saying, have we ever seen this device before? 

There are sensors on your device that capture how fast you’re doing something. If you’re on your mobile device, how many pixels you might be firing up on your screen at any time. That information, believe it or not, actually is very insightful to know whether or not you’re a real person or if you’re a bot. 

You know, bots may send information without firing hardly any pixels at all off on your phone– maybe one or two pixels that it’ll do as part of the message is coming through, and we know that a finger touching a screen will fire off a minimum of like 110 pixels. So it’s that kind of information that’s captured. It’s small bursts, but very meaningful. And I could probably– 

I don’t know if your eyes glaze over, you find it fascinating about all the different aspects of data that we capture during that online process, just to validate that you’re a real human being, hopefully that we’ve worked with you before, that our clients work with you before, and there’s a certain comfort level to know that when you buy that item, you’re not a fraudster, you’re not a bot perpetrating fraud, that you’re an actual consumer that wants the goods and services and it’s going to be sent at home. 

99% of the time everything is good. It’s the 1% or less than 1% of the time, when you’re dealing with actual fraud. And that’s where all that real-time information and validation and assessment becomes very critical. 

MICHAEL KRIGSMAN: And how fast is all this taking place? 

ERIC HALLER: Milliseconds. Milliseconds. Literally less than a blink of an eye. I think some of our faster solutions will run at 40 milliseconds. I mean, I’m trying to recall my visa days, I used to be a payments guy and I think typically, if you’re in a retail environment you need to run at about 200 milliseconds. So that about one fifth of a second, and we’ll do one fifth of that. So, extremely fast. Yeah, that’s the environment that we’re in. 

MICHAEL KRIGSMAN: So let me get this straight. So you’re gathering– for each transaction, you’re gathering dozens, maybe hundreds of data points to first validate the authenticity– the identity of that transactor– that person, to ensure that it’s an authentic, legitimate transaction. And it’s only at that point, that you start analyze– collecting the real data relating to the transaction, processing that data, and then returning it back to the device. 

ERIC HALLER: Yeah, depending on the process that we’re talking about. So since we just are camped right now on doing things online, which is what most people do these days anyway, that transactional assessment happens multiple times during a session. Of which you are not even aware of, until it comes time to actually hit at buy this item or whatever it is you’re doing online. 

So there is a cut from the moment you engage. And this isn’t just– Experian does this kind of work, but there are a lot of folks in the industry that do this, including our clients, but you’re going to be looked at or watched as you go through a website. 

It’s very peculiar, if somebody is on a website and they all of a sudden want to look at source code. How is this website put together? I want to go flip over and see if I can get to the source code of that. That would be a red flag. Normal shoppers don’t do that. I’ve never done that, for example. So it’s not a normal behavior. 

But how you traverse from the moment you land on a site to the end game of actually buying something, may go back and forth between our client and us multiple times to assess. 

Then when you get to the final transaction, and you’re providing some information– an account number, or a name, an address you’re going to ship an item to, whatever those things are, then it goes to a whole another assessment. Are these credentials valid? Is this matched with what we’ve seen before? Taking into consideration so many aspects. 

Ultimately, believe it or not, as crazy as all these hoops and checks and validations and going against databases and running models and all this happening in milliseconds, as crazy as all that sounds, we don’t want to stop something unless we’re absolutely convinced there’s a lot of risk in this. Most often there isn’t a lot of risk. Most often the happy path is successful purchase. 

And that’s what our clients care about too. They don’t want to disrupt their customers experience because of the risk, unless the risk is high enough or certain enough that they want to take an action. So all of that behind the scenes– in theory, you could have an environment where your experience is choppy or you’re asked a lot of questions along the way, and then that’s no fun for anybody. Everybody likes the convenience of being online. 

It would become less convenient to going into a store in some ways, if you took that kind of approach. And that’s not where the world’s headed. The world wants to make it as easy as possible. So it’s using all that information, and validating that behind the scenes. And that makes life a lot easier for all of us. 

On the other hand, if you don’t get it right, there’s a very high cost to allowing a transaction through, that should not have gone through because it is in fact, fraudulent, masquerading as a legitimate transaction. 

That’s why it’s so important. So our clients– we go through this ourselves, but our clients go through it with us, which is evaluating the efficacy of the approach itself, and assessing– like I always say, it’s the data you don’t see, is what usually catches you. If you’re going to get hit by fraud, it’s because it’s data that you’re not analyzing, you’re not seeing. 

And so the different approaches that are out in the market, we take a very– I’ll say, broad, holistic approach, where we’re looking at as many data sets as possible. In fact, in our environment, we’re building a platform right now, that just does nothing but fueled with new data coming in from all different– if it’s been identified and can be captured, and legally we can pull it in. 

Obviously there’s legal hurdles and regulatory aspects of this, but assuming that it’s OK, we put it in a hub, and we’re constantly evaluating for new signals that might identify risk, so that we can put that through to our clients in the solutions that we have in the market today. 

So you’re right. If you get it wrong, you’re going to take a financial hit. So the objective is to stay as much on that edge as possible, in terms of capturing the data and making sure you can see that fraud, before you get hit. 

MICHAEL KRIGSMAN: And again, this is real-time so it’s happening faster than that. 

Yeah, right. Yeah. I mean, that’s exactly right. I’ve always had  trouble– I’ve always had trouble getting my head around how fast this happens. When I was younger and I did my own coding, you get excited about processing a file faster. Like you’d put a file in and maybe one month it would take you all night to run, and then they upgraded your system, and then it only took an hour. 

And so the idea of being able to understand the magnitude of what you are trying to do in a quick speed is easier I think, than when you’re confronted now where we’re processing literally millions of transactions every day in millisecond time frame. Maybe because I’m in my 50s, it’s just hard for me to imagine it. I know we do it and I see it, but it’s pretty amazing. 

MICHAEL KRIGSMAN: I totally agree. It is absolutely amazing. So not to get too geeky here, but the database technologies, where does that come into play? Do you think about that, or is that just in the back end? 

ERIC HALLER: For sure. No, database technology has become a big part of things. We were early adopters for Hadoop early adopters for Spark. We’re early adopters in a lot of different database technologies to see how they play out. I think a few things like– so we’re talking about online and fraud, and one of those things that’s not really done real-time, it’s done in batch, but it’s updated on the– we’ll call it on the margin real-time, is an identity graph. 

So that is the aspect of matching up all these different data elements that might describe, ascribe our identity, and trying to draw the probabilistic relationships between those. So we’re all familiar with Social Security number, name, address, telephone number, birthday, that’s what we’re used to. In the online world that would be augmented by maybe an email address, a mobile advertising ID, an IP address, device ID. 

And so all of these pieces– gosh, even now my dongle that’s attached to my TV, that’s a Roku dongle or Google Chrome dongle or Amazon, all these devices become a factor in our identity– our household identity or individual identity. And so that is done in batch. 

And I bring that up, because you bring up database technologies. A database technology like Neo4j, which may not necessarily be– I’ll say it on the lips of everybody who’s in tech, but in the ID graph world, that actually became a go-to database for us, because of its ability to calculate and draw those relationships. 

It’s a lot of work and what I would call high dimensional space, where you’re trying to draw relationships between data where there’s no relationship that exists, but you’re going through the machinations of developing relationships across your entire corpus of data, and that has to be reduced down to what we call low dimensionality or dense data where those relationships exist. Neo4j is very good at that. 

So I do think the world of database technology is definitely evolving. Even when I’m talking about reduction in dimensionality– hopefully I’m not getting too geeky here, but there are a lot of shortcuts that one does as a data scientist to try to reduce down the amount of computational effort required, depending on how many relationships and how much data, and it’s only getting bigger and bigger. 

I think database technologies will start picking up those tricks themselves, make life a bit easier for us. It’s always typical that the innovation starts at individuals and eventually works their way into larger software platforms and computer chips, so that automation of some of that, I’ll call it human ingenuity will happen. 

Individual profiles are getting far more complicated now than they ever were before. I think that’s another added perspective on database technology, where that’s going to evolve as well. But yeah, I mean, I’m a kid that grew up with pre-Oracle and dBASE IV, and so I’m an old timer in the world as it’s constantly evolving and getting smarter. 

We can talk a little bit about blockchain also, as well as the database technology. That actually I will say, even though I was a quick learner on blockchain, I’m a slow adopter and I have been. We’re playing in that space a little bit but not aggressively. I do see some– I’ll call it, long term potential value in blockchain and decentralized finance, the DeFi space is blowing up. 

But I’ll say, it’s like really, in my mind it’s a bit more abstract. Because when you’re talking about decentralized, it’s more about things like every business has a certain number of calculations that they do on data X, Y, or Z. And if you evaluate all businesses doing all their work, how much overlap there actually is. 

And that’s why I say it’s a bit more abstract, because as you start going to the blockchain world, as those calculation– that calculation funnel might be able to actually be made more efficient over time, across a broad array of businesses, then there’s a financial value to that. 

You could argue there’s a climate impact– a positive climate– there’s a sustainability value there in terms of burning up electricity and energy and heating things up. So there is value there, I just think it’s a bit more long term and maybe a little more abstract than running an immediate P&L and seeing the value there. 

MICHAEL KRIGSMAN: From a real-time data perspective, do you consider blockchain and decentralized finance to be real-time in the way that your transactional data is real-time? 

ERIC HALLER: So from what I’m exposed to, the answer is not really. But I do know that it’s moving in that direction. So I won’t say anything disparaging across the many, many businesses right now that are running at that. 

I do believe that when there’s a lot of people focused on something and a lot of money put into something, there’s a good chance that these hurdles like blockchain and real-time and high volumes has always been kind of one of those hurdles. 

I can say right now, decentralized finance is probably more along the lines of like a mortgage or something where there are pieces and it is something that takes a bit of time to work through. 

But I’m not going to say it’s not going to get there someday. That would be my moment where I say the mobile phone isn’t going to go very far, that was the Steve Ballmer, the mobile phone– I’m not going to say that, because I think it would be a mistake. 

MICHAEL KRIGSMAN: Well, as you said, when you have a lot of money and a lot of time and people putting energy and resources into something, there’s a reasonable chance that at the other end something important will emerge. 

You were talking about the evolution of database technology. Do you think about the evolution of cloud, cloud storage, hybrid cloud, multi-cloud and where all of that’s going? Because obviously, that’s an important part of the overall data equation as well. 

ERIC HALLER: Yeah, it is. Although I tend to be more on the solution side and less on the IT side of it, but I’ll tell you where we see this going. So a hybrid cloud or multi-cloud– I do some things on-prem, I do some things in the cloud, or I do things across multiple clouds. Some of that’s a financial game, some of it’s a security game, and some of it’s– I’ll just say some of its financial and some of its security. And I’ll explain why. 

So in a multi-cloud environment, I want to make sure that if I develop something in AWS, that I can move it to Google or I can move it to Azure with Microsoft, and so I want that flexibility. So I might build something in a container in Kubernetes, and then I can move it from cloud to cloud. That’s I say, more a financial game, right? 

Might be different things. I may want to be able to negotiate with one company over the other, and if I can maintain some flexibility or some portability, then it gives me some leverage. There may be another aspect to that, which is not all clouds are created equal in all markets. 

Some clouds perform much better, international markets will say non-US and other clouds. And so you may want to be multi-cloud just because like Experian, you do business in a lot of countries, and you want to make sure that you have that ability to deliver the best solution possible in whatever market that you’re in. 

So that would be, I would say, the financial aspect. On the security side, depends who you are. I’ve talked to a lot of folks, we’ll call them business/technologies that appreciate the security features of being in one of the big clouds, knowing that they’re completely obsessed and with the best tools and keeping that up. 

Maintaining cybersecurity defense is a lot of effort, work, know-how, you have to get the best people, you have to be constantly on it. Experian happens to be one of those companies where we invest a lot in making sure that we are out there on that edge. But there is some value of being in the cloud from a cyber risk perspective. I would say– so there’s a risk side of it. 

On the financial side again, flipping back, managing R&D, we do a lot on-prem, that we just did I’ll say, the cost analysis to say, if we did all of this in the cloud versus we do this on-prem, where would we be better off because of all the– I’ll call it the computational intensity, and the data analysis that goes back and forth. We actually have to analyze a lot of the data itself. And so that in and out, in and out, has an expense to it. 

But where does that all head, it’s a different question. So that’s kind of where we’re at now, where it all heads. So I have my own opinion– this is more of an Eric Haller opinion, more than anything else. 

So I want to say about six years ago, I had done a market map of the clouds and where they were headed, just trying to figure out what should be Experian’s role in all of this. Where we should have an opinion, the board will ask me for my opinion, I got to have an opinion. 

So in my perspective, was that each cloud provider has to look at their ability to differentiate themselves, not just on the fact that there are a better financial equation or whatever. And so I went to, it’s either going to be analytical tools which Google’s tried to do obviously, or data, which I think Azure has tried. 

And I actually believe that the data aspect of the clouds over time, will become more of the area of secret sauce. And I’ll give you an example. You’ve got multiple businesses all aligned on one cloud infrastructure. We’re talking about real-time data sharing, talking about maybe identity graphs, those types of things. 

Today we go through those processes maybe in developing a consortium, a group of businesses that are willing to work together to do it, or you develop a certain amount of infrastructure that has access to that data. So I think in a cloud environment, they’re kind of naturally on that door to be able to drive consortium-related things, where the added value would be very easy to demonstrate because of the infrastructure and the environment that they’re in. 

I think that’s all upside benefit to everybody by the way. I don’t see that as– I could argue all the upside benefits of that over time. Again, I guess this would be my abstract thoughts than anything else. 

MICHAEL KRIGSMAN: Eric, we’ve been talking about the technology. This podcast is really about the business impact, the consumer impact of real-time data and innovation. And so let’s talk about the role of data in innovation at Experian. So how do those pieces connect? Innovation, real-time data, and Experian’s business goals. 

ERIC HALLER: So a couple of things. One is– and you might be aware of this. I was in a book written by Greg Satell called Mapping Innovation, and he did a chapter on our labs and how we innovate in the labs. One of the things that he had earmarked was that all innovation centers around solving business challenges. And I say that first because the role of data– data in my mind is typically treasure that you tap to solve a problem. But it’s not your go-to to innovate from, it’s your go-to to solve a particular problem. 

So like with our labs across the globe, we are highly engaged with our markets, our client base. In fact, we measure– we have a funnel that starts with our clients on top and how often we engage with them. And we actually measure our level of engagement, quarter over quarter, region by region, just because all ideas come from solving our clients’ problems. That is kind of the top of the funnel. 

When you start getting into the how. When you start getting into the how. So I’ve got a particular challenge, it might be, gosh, I’m trying to– I’ve got a fraud problem, and it’s when people are using their mobile devices, and it’s only mobile devices, and it’s for this kind of app that I’m using. I mean, they can get very specific. They can get very abstract. 

Then you go to the data to try to figure that out. You go, OK, what can I learn from the data? In our labs what we try to do, is create– I’ll call it like Three Rivers Stadium, all the rivers kind of lead– but we try to get all the data rivers from all the businesses leading to one location. 

That way when it comes time to problem solve for an issue, we have as much data as possible, that we are aware of. If we in our problem solving think we don’t have the right kind of data to solve it, we actually have a team that goes out and talks to other companies that might have the kind of data we need, to try to pull that in. 

With a company like Experian where data and machine learning or analytics becomes kind of like the fuel for the problem solving, it becomes kind of a second nature for us. And the biggest– I’d say the biggest guardrails on it is our regulatory environment that we work within. 

So like I mentioned all of our businesses, the credit business has got the Fair Credit Reporting Act, the auto business has the Driver’s Privacy and Protection Act, even multiple businesses have Gramm-Leach-Bliley, our health care business has HIPAA. All of our data scientists have to learn the law, they’re tested regularly on the law. Our audits go in place to make sure we’re in compliance with the law. 

So when I say all of this, and maybe just in case anybody was curious about this and go, Experian has got a lot of data, what are you doing with it? So we have to go through all those guardrails to ensure that when we do go to solve a problem we use as much data as we can, that we’re doing everything in a legal and compliant manner. 

But that’s second nature for us. Like I can tell you that with all the processes we have in place, we appoint people to do nothing but assess these things, but in most companies that wouldn’t be the case. I think in most companies it takes a lot of work to set up an environment where you’ve got that steady flow of data, you’ve got people that are aware of what’s available, what they can use. 

Sometimes it is the human ingenuity and creativity, sometimes it’s– when we first started our lab 11 years ago, we built what was a JavaScript that basically turned data into attributes. And we took any corpus of data and to turn them into thousands of attributes, thousands of attributes. 

And then– so you didn’t even have to know. The logic was, you don’t even need to know the data, let’s see what pops. Let’s run it through, we’ll throw some machine learning at it, maybe it’s– who knows? Boosted trees, or whatever it is, but we’ll see what pops, and then that will guide our direction. 

And sometimes that works. Sometimes that’s super helpful. Sometimes it’s more than that. Sometimes you have to think through what is you’re trying to solve. And I almost call some of our attributes like they are mini-models all of themselves, and you have to think through that, and you’ll get more out of it. You get more juice from the grape. 

So you said that innovation comes from– I’m interpreting here– comes from your customers’ challenges and the customer problems that they’re trying to solve, and your customer goals. And you’re innovating off of that, as opposed to data being the source of innovation. Data is the treasure, as you described it. But there must be times when you have access to certain kinds of data, and that then spurs you to drive innovation. 

You know, it’s interesting posit, because my gut tells me you’ve got to be right. But I’m just thinking about practical applications in our business. The thing is you would love to have maybe the bandwidth to go through kind of let your heart go wild on this data set and see what you come up with. The last time– I’m trying to think the last time we did that. 

See the problem Michael, is a lot of times if you do that, sometimes you might get the one in a million needle in a haystack I found a winner, but you’re more likely to find winners if you’re chasing a particular problem to solve. Because you know if you solve it, I mean, the theory is somebody will pay for it. And that’s the thing there. 

I’m just thinking about COVID. That might be one of the areas. We did a lot of work in COVID when the pandemic hit in all of our labs, where we came with the initial premises. The current models, the SEIR model, the IHME model, the models that were being used to assess how quickly the virus was spreading and what the impacts were, were important, but we wanted to bring it a little bit closer to home around economic impact. 

Because the closure of businesses in a particular geography, people becoming unemployed or having to go to unemployment, so we felt that Experian was in a nice spot to help hospitals, help governments, municipal governments, federal government, in assessing the building up of hot spots across the country. And we wound up doing this in multiple countries actually. 

But in that case, again, it was a problem where we thought we might have a solution on, but our net for data got cast pretty wide. And data assets that– pulling GPS data from MARC Advertising– online advertising networks. 

Like you wouldn’t think to use something like that maybe initially, unless you started and want to say, well, how compliant are people with stay-at-home orders and that kind of thing, then all of a sudden you use that. And if there’s more movement in an area, we can model and show that the virus would spread more and you’d have a bigger impact on business and economy in a negative way. 

In Brazil, we realized that we didn’t have the data to solve the problem. And we engaged in a very ambitious network of companies– we wound up partnering with the United Nations and the World Health Organization, and guys like Amazon and Microsoft, universities like University of Chicago in Brazil– even though they’re in the US, they helped with this in Brazil, along with like University of Sao Paulo and others, to build out a network of data. 

And because we didn’t have it all within Experian but we had some pieces of it, and we had this ambition that said, the world needs help right now. And we should do our part, and eventually it will feed back to the things that we’re good at, and helping our customers figure out business problems, but in that case, maybe there was a bigger picture there. 

But again, it wasn’t like kind of going through the data as much as listen to the problem. And maybe where the human creativity is, is trying to assess where– like I will tell you, this is like human creativity aspect. 

When we are brainstorming around what data might be helpful in tracking the virus, I mean, you wind up realizing some very specific things were helpful. Like– we call them social determinants, like proximity to public transportation. Like that actually is a very useful statistic in tracking virus spread. 

But in our brainstorming, we went as nutty as, hey, can we look at by nine digit zip or seven digit zip, people buying cough syrup and thermometers and things like– because they’re sick and they’re going to the store and buying it. Will that be enough up– will it tell us enough, that it would be like an early indicator before they took a COVID test, that there’s somebody sick. Turned out not to be useful, by the way, just so you know, it’s noise. 

But that’s what I’m saying is, you want the problem to drive your net, capture whatever data you can or can think of or can I get access to, and then let the data tell you how to solve the problem. And that’s more of a– even though it doesn’t sound data-centric, I would say that is a data-centric approach. 

MICHAEL KRIGSMAN: It’s very clear that you’re looking at real problems, and figuring out what kinds of data you have available, or might you need or be able to get in order to address those problems. And Eric as we finish up, you were talking about the social benefit of data. And so how can organizations, governments, use large scale real-time data to the benefit of customers, to us citizens, to the social benefit of our environment? 

ERIC HALLER: So you know, I guess that your listeners are at all different levels, all different kinds of companies. I will tell you that it’s not like big pharma. You know like when I think of big pharma, I think of, hey, you don’t really get into the big pharma game unless you got like a few billion dollars to spend on infrastructure, otherwise you’re not going to really get into the game. 

And with data, it’s not there yet. I’m talking about all the infrastructure we have at our disposal at Experian. So we’re pretty fortunate that way, and I’ll call it the tier one companies that have that kind of infrastructure. 

I think if I was doing a startup and I was chasing what I think we say is social good challenge, like deepfakes– I’ll take deepfakes. I’m like you know, you see Experian a lot in the press around deepfakes, because I’ve got the labs and our fraud and ID group putting a lot of time and effort, thought into that, because I think that’s an emerging challenge for the world. I think three years from now, it’s going to be much more challenging than it is now. 

And if you’re a startup, you could still chase a noble purpose around deepfakes. I mean, but you’ve got to be able to persuade others in your fight, that it’s worth it to jump on board. It’s a different challenge if you’re brilliant like Mark Zuckerberg and come up with a platform that everybody just glomps onto and then you’ve got the data that you can– that’s the dream, right? And if you come up with that, that’s fantastic. 

Otherwise, you’re going to have to get stakeholders to jump in, and they have to share in the same dream and the vision. And I’d say just try to do good, like stick with what’s legal. You know, we got a lot of clarity– I think governments can always be more clear of course. 

But I think what you’d see as a world trend, is government trying to right-size legislation to protect people, but allow innovation and our lives to be impacted in a positive way to continue to grow. 

But yeah, that’s what I would say, is be ready to work with a lot of other folks. I mean, starting with a company like Experian is always a good thing. If it’s a big idea we work on things with other companies all the time around big ideas. But generally speaking, I think that’s the key– stakeholders. 

MICHAEL KRIGSMAN: Collaborate. 

Collaborate. That’s the word I was looking for. That’s a better word. 

MICHAEL KRIGSMAN: Eric Haller, thank you so much. This has been a very insightful and inspiring conversation. Thank you so much for taking time to talk with us today. 

ERIC HALLER:: Hey, thanks Michael. I always appreciate being on your show. So thank you. 

We’ve been speaking with Eric Haller from Experian. Everybody, thanks so much for listening. Be sure to check out the next episode.