Full Transcript Below:
Bradley Chalupski: In today’s competitive e-commerce environment, it’s never been more important to earn and maintain the trust of your customers. Merchant Fraud Journal’s “To Catch a Fraudster Podcast” is supported by Sift – the leader in digital trust and safety. Sift empowers companies to stop fraud and grow without risk. Visit sift.com/assessment to schedule a consultation with Sift’s trust and safety architects – industry experts who have decades of fraud-fighting experience at companies like Facebook, Square, and Google. They’ll help create a custom plan for your business with an emphasis on technology, organizational structure, and process. Visit sift.com/assessment today.
Bradley Chalupski: Chris, thanks so much for joining us.
Chris Mangold: My pleasure, Bradley.
Bradley Chalupski: Chris Mangold – also known as Christian Mangold– is the CEO of Fraugster and the Managing Director there as well. So, Chris, why don’t you tell us a little bit about yourself; who you are, where you come from, who you represent? And then we’ll jump right in.
Chris Mangold: I’m really excited about this interview today. I’m representing Fraugster, a Berlin-based, German, Israeli payment intelligence company. I’ll come to that in a second – what we’re doing. Who am I? I’m the Fraugster, as the CEO since 12 months now, pretty precisely. My background in internet payments and e-payments reaches now 12 years in the past. I started out with a company called SOFORT when I was in Germany, not known so much in an international context. But to give you some context, this was one of the first A2A, Account-to-Account transfer schemes. And these Account-to-Account transfer scheme actually back then have been even an anti-fraud method more than a payment scheme. It had some conversion problems. But in the very risky areas, they always have been the payment of last resort for gambling, digital contents, remittence, and these kinds of things. After two years back then, we successfully sold that company to Klarna; what you also didn’t know. And then I served for two years as a Managing Director for Klarna in the back area, rolling out back then Klarna’s checkout, which was at most disruptive back then to the merchant’s checkouts besides the PayPal Express, and other things stations in credit card acquiring, in crypto, Fraugster and myself found ourself 12 months ago. I was brought in the first interview when I looked at the deep AI technology we have here and the UX. So, cutting fraud is one thing and avoiding false positives is another thing. And the third thing is also to have the fraud operator in the center so that she or he can really combine the roots, backtest everything, and [02:59 inaudible] nice cockpit and dashboard. This is what we bring to the market.
Bradley Chalupski: Well, I really appreciate you joining us. You left out the most important part, which is that you have flawless English grammar with an amazing German accent, which is just amusing to my ears. It is just a joy to have you. It is a pleasure to have your baritone voice on this program. Thank you so much for joining us. So, we’re going to jump right in. We spoke previously for a little bit about how we like to do things here. So, why don’t we just jump right in with your first crazy fraud story?
Chris Mangold: I was thinking about it, and the pity is fraud is interesting, fraud is measurable, but it’s not always that crazy. So, what really erodes quality and security in the internet is perhaps completed in a second, all these small and repetitive business cases. So, being a fraudster is a job, being a fraudster is hard work. So, it’s not only this Hollywood style kind of coups [04:08 inaudible] but sometimes it’s really business-case-driven hard work, and that’s why they’re also somewhat predictable. But I was checking our fraud fights and one of our largest fraud cases we had here. Fortunately, it was not our fault, but the merchant was informed but they still failed to revoke the delivery in time – it was Christmas time. It was a sauna, worthwhile 14,000 euro. And still a sauna is not a good that you easily intercept and run away with that. It shows us that fraud is not always predictable. So, you must, as a merchant or as a bank, not feel safe that something is crazy enough to not be stolen, even saunas can be stolen. On the other end, we have with one luxury label legitimate purchases up to $50k, sometimes also in pre-Christmas time, sometimes it’s making a bit what he or she did that you need such a ticket.
Bradley Chalupski: This idea of high ticket items is definitely something that we’ve covered recently on the publication. So, I’m interested to hear from your perspective – you said that even a sauna, this is an item that most people would feel like who’s going to steal a sauna because how are you going to steal a sauna? How is that going to work exactly? So, I’m curious to hear from your perspective, how you go about assessing risk specifically with these types of high-ticket items. Are you looking at anything differently than you would for other type of items? And if you could take me through that process a little bit on your end, it would be, I think, really insightful for our viewers or listeners.
Chris Mangold: High-ticket items is definitely the area where latest rules fail. Since, as a rule, you would always cut very high purchases at nighttime or whatever. But the other thing is, this can be the most lucrative transaction since perhaps you have this checkout at the luxury trader, at two o’clock in the morning, when you had two or three too many, or when you just got the call from your spouse. So, the beauty about high-ticket is, on the other hand, that most often, they are verified investing a bit. These are the ones where automation sometimes has an end, where it makes sense to go to a manual flow, to look into these transactions, they have all kinds of visualizations. You see, in all the dashboards of all of our competitors pops up. There you get everything – you get the full story, you get the credit card details, you get the IP, you see where this person pretends to sit and where this person wants to get the goods delivered to. And this gives you a lot of background. The other thing is you see linking. This is something you actually only achieved with AI else you are not fast enough. You will see graphically the link; has this credit card, has this type of good, has this IP already been successful in the past? Or has there been flaws, or rejected transactions, or even chargebacks in the past? And this is what it take together to come to an educated decision. Another thing is, normally, time is the absolute essence since nobody waits in the checkout till the transaction is accepted. This can change a bit if it’s such a high ticket, customized good, since most often then the consumer will not get this as the next shop is just a mouse click away. So, this gives some leeway there still [07:40 crosstalk] hitting the reputation of the trader.
Bradley Chalupski: I have a question on that because we talk a lot about friction in the process, obviously. We’re at a point of zero friction at this place in time. I would think that for luxury people, luxury goods, that that’s even more important, that people who are paying $15,000 for a sauna on the internet are really expecting to just be accepted and just move forward. So, how are you balancing that need for a reduced customer friction experience with the idea that you have these types of luxury customers who are expecting a luxury experience? Or do you think that there’s just no way to avoid, given the the dollar values adding friction into the process?
Chris Mangold: Beautiful thing. Don’t quote me, but when in my former station, we launched back then and there was not the tool with requirements of the services directive already in place. We, back then, eight years ago, rolled out a real one-click checkout where we said, “Without registration, you can purchase up to the limit, you can purchase only by punching in your zip code and your email address.” And then it was accepted. And let me set for more critical – I’ll come to the point in a second – risk profiles. We have to not have this one click, we show them rather a variety of payment methods – so, if they go to a final payment method like a trustee or so forth, we would not have to score them. And if they would go for a “buy now, pay later” function, we would have to score them, and probably even to throw them away. And this flow converted just better than the one-click flow. And snappiness and seamless checkout is one thing, but if this is thousands of dollars, it can feel insecure to the consumer – this is one aspect. So, interesting enough, this was definitely not stipulating one-time passwords and these kinds of things. But we had these, we measured the doubt. And I said, “All these other party poopers, this will be too fast, people will not like it. Measure will result with millions of transactions.” And to see, there’s all there. I could take the credit card, I could take PayPal, but I want to opt-in “buy now, pay later.” Back then, really, get it now and later on sort out how I paid and converted better. And another point is education. Since, for high value checkouts, many users know that there might be a second step. Business travelers – I don’t know what I did last time I had my trip to Las Vegas and my credit card was accepted right away – interesting. But, most often, for instance, business travelers know if they have a business trip, which is several thousand dollars or euros, that there might be a confirmation and could be a disbalance, even a bit. For me, it felt more awkward to just have a few thousand dollars on my card without anything – just “Goodbye. Thank you.” It could be even more comforting to have this trip [10:52 inaudible]. On the other hand, to add on what you said, what’s most disappointing, obviously, for a fluent or high-net worth customer is to make up her mind to buy these goods, to go through the checkout, to perhaps even undergo some extra verification or authentication steps, and then being rejected by timeout, by whatever. This is a catastrophe, since then, they’d really go shop somewhere else.
Bradley Chalupski: That’s really interesting. That’s the first time that I’ve ever heard anyone say anything like that. Usually, we’d just hear about “faster, faster, faster.” But I could see how that higher-end [11:30 crosstalk]. So, I’m curious – this is maybe a little bit of a side question here – what do you do to test these processes? How are you deciding what the optimal friction to create is, the optimal way to create that friction? And how are you balancing that with your need to complete each of these purchases but also get information on what actually is most effective? Are you running A/B tests? What are you doing?
Chris Mangold: Yeah, there’s this A/B testing. Now, you have the opportunities to do in-vitro tests, EIP tracking, and all kinds of things. Another interesting finding what surprised me very much in 12 months is reviews of e-payment. And you would also perhaps not have expected that whenever you change something in the flow, even if you change something that obviously improves the flow, conversion – they first drop a bit. This is still a popular e-payment, eCommerce shopping. Meanwhile, many of the consumers shop, really, especially with one shop – just see how often you buy at your sneaker shop. This is not once per month so that you would always feel what the flow is. And the other thing is the high rollers, which are shopping really in electronics. There are some tinkerers, they’re shopping practically by the day. And they concentrate a lot of volume. And for those, it’s disturbing whenever something changes. They’d rather get used to clunky checkout than appreciate that. With all beauty of EIP tracking, and so you optimize the flow, [13:07 inaudible] comes back. It was a learning. The first time, it always dips. And this is important, since then you say, “How do you measure that?” You have to communicate it to your merchants if you try something new. Obviously, there should not be a huge dip. But if you try something new, there will always be a small dip first before it improves and goes up.
Bradley Chalupski: So, on that subject, how are you recommending to merchants that they inform their customers of this, just to say, “There’s going to be a new checkout experience. You’re going to experience X and Y and Z.” Do you find that oversharing? What’s going to come makes people not want to try it to begin with? What are you seeing with that communication between yourself and the merchant, and then the merchants and their customers, since you’re kind of playing a game of telephone with the end users from your end?
Chris Mangold: Considering the average checkout and average margin per user, it hardly makes sense to reach out actively there, that’s rather creating facts which have to be attractive. It was back then big blue, now big pink – Klarna. You always, obviously, had discussions with the merchants – PayPal did the same – and merchants said, “No, the checkout is ours and we do not want to have different brands in there between. And you could preach again and again. We are only earning money. If you are earning money, we are only being paid for successful transactions. You were doing great 100 million of volume, we are doing even great – tens, back then, now hundreds of billions of volume. Believe us, we have 50 people working for the checkout, you have five. We are investing in the checkout.” But still, you do not want to have people interfere in your customer interaction in your story. Now, I guess, it’s sooner or later. Today, I checkout, which is not offering PayPal and “buy now, pay later” like Klarna or something, that will be mistrusted by consumers. You cannot educate consumers. Many have their likes. We tried to educate them for more than 10 years, but you have to consider the fact that sometimes the more payment methods you offer, the more the likelihood that everybody finds the one they tend to prefer, and then they do it, and they find out, and there will be a hard effect to the ones which are really working.
Bradley Chalupski: Really insightful. Thank you. So, let’s hear the next great story.
Chris Mangold: Do you know how much a credit card is in the dark web?
Bradley Chalupski: The cost for credentials? I think it’s like $1 or something, right?
Chris Mangold: No, it’s much more expensive. First of all, it depends. It depends on whether you just want to have a MasterCard with public mode, you can call it, MasterCard with pin mimicked – it was perhaps $15 or something like that. And then when it’s a quality card with the availability of at least $1,500, it can go up to $150. And this is why fraud is hard work. And criminals – they have to live from what they earn. They’re very, very good entrepreneurs. This means that whenever you increase the threshold, you are likely to not be hit since they will always go for the easiest way to commercialize this credit card. If you buy now a credit card for $150 on the internet, you have only two or three shots since then, each and any velocity rule goes up and this thing will be blocked or somebody finds out. So, we have to buy it and you go to find the easiest target. This leads them to fraud patterns, which one might perhaps not expect. For instance, in airline fraud. Airline fraud – it’s always tricky to steal a ticket and then fly with it under a fake name – just to say, I would not enjoy this experience. But why are they so many victim airlines, for instance, to third-party fraud as opposed to first-party fraud that people make up the mind not to take the travel and then cancel, it’s a different topic. But for card testing, for instance, since if I want to test that this card really has $1,000 limit. If I test this at my local grocery shop, could well be that I’ve been blocked since this is just saying this checkout is way too high, it’s suspicious. And we’ll never find out whether this limit is on there, and then it goes to reject me as a mismatch of merchant purchase. Airline is unsuspicious – they have very high ATVs, Average Transaction Values. So, this is very good for card testing. Nobody wants to take a ride with this ticket. And still, the economic damage is huge since the [17:49 inaudible] is blocked, you cannot resell it; you’re flying empty; you’ll have the cost of the chargeback later when the consumer that’s being mistreated there finds out; catastrophe, economically, and for the fraudster, not really a business case other than preoperative fraud.
Bradley Chalupski: I think this whole world of card testing is something that we don’t talk enough about. Take us through some of the more interesting examples that you’ve seen of what fraudsters are doing that is maybe out of the ordinary that people can identify as a potential fraud testing scheme that you’ve seen a lot as fraudsters are trying to determine that these cards are legitimate and that they have the limits that they think that they do.
Chris Mangold: Card testing – but if you really steal an identity in the card, the next step is then – when the test is fulfilled – to get hands on the goods. This is the next thing where you have to see then how you’re going to commercialize. So, ideally, the best thing is to steal any vouchers or something like that. This is tricky with the cards instead of all systems we hit you. The other thing is then if it’s physically good, you have to intercept them. Then you have to drill into the process from the intro, or other thing most popular is what one might find out without knowing that your credentials are used, your illegitimate credentials are used since then your card. If your delivery address is connected or the fraud screening method says “okay,” and then the fraudster tries to log in into your account later to derail the goods to an anonymous parcel station or a pickup point somewhere at a retailer’s because they never check the ID. They are just like, “I’m not paid for checking IDs.” They just [19:35 inaudible]. Then all of a sudden you find out there has been a transaction, even a receipt of delivery, under the name of you, and then you have to fight back the liability against the credit card or the “buy now, pay later” or whoever was a victim besides you to this identity fraud. And this is where the hard work comes, this is not the Hollywood.
Bradley Chalupski: So, you mentioned vouchers. Because I’m always curious in this world, when you buy these cards – there’s no honor among thieves, as they say. So, you go on the internet and you buy whatever off the internet, and you want to make sure that you’re getting what you want to get. But at the same time, you need to make sure, like you said, that you’re not making it known that the card has been stolen. So, I’m just curious – you mentioned vouchers – if there are other specific things that people should be looking out for, for patterns of behavior that will indicate card testing on certain types of products or services that are targeted by these fraudsters. So, airlines, you’re saying, are really prolific because there’s a high ATV, so it’s not a big deal. I guess, vouchers, I’m assuming, you’re talking about maybe lower-priced vouchers because you could just make sure that the card works. Are there other things that you’re seeing in there? Or are those really the two major instances?
Chris Mangold: The beauty with vouchers is that you get hand on the voucher, and then you can really steal the money. So, this card testing in airlines, most often, perhaps the consumer will not find out since a clever enough fraudster will perhaps even abort the transaction once it looks good – so, not to damage his expensive limit when he sees that it goes through. But again, it makes perfect sense to go through your statements, your card statements by month, and to go through that and just chase it for anomalies. Suspiciously high amounts, as you say, you will find out quickly, but suspiciously low amounts, you will also find out. And then you’ll see very soon, normally it pumps up. So, this card testing thing, it could be Amazon things or whatever was bought, or any merchants most often from the digital some phony gambling thing, where you might perhaps even think that you [22:10 inaudible] not that you should give them hold of the card, but it happens again and again. And then there is 50 cents, two euro, six euros, eight euros, 17 euros, 70 euros. So, this is how tests and escalates, and this is also mimicking normal behavior. [22:30 inaudible] since it could be that somebody is now not addicted but liking this game and buys more and more. And these things you find out – even if it’s abused, they would put it with similar. And they would not jump in with the highest possible purchase at a gaming or at a voucher shop – they would start to mimic normal behavior. And this is then also when your card is already spoofed, this is also for the antifraud scheme. The topic where we have to go, “How do we find the mimics?” And then you need many attributes to see what’s wrong with this? Ideally, automatically, what could be wrong with this transaction? Again, this is a busy man which does not exist. It’s the phone number. So, there are many layers. And most often, again, the entrepreneurial fraudster has to go, they have the least resistance, and they will not make a super deepfake. They will [23:26 inaudible] scratch a little bit on the surface to fake a social media account. But then, if you automatically or manually open it, you will find out there is no followers or there is no pictures in there. So, it’s always worth the effort to go a tiny bit of extra mile to, again, increase the threshold for the flux up, and then, at least, you are protected and they will try to somebody else.
Bradley Chalupski: So, I have a bit of an existential question off of something that you said there, which is: Do you think that these types of attacks always leave traces? Because you said they try to mimic regular behavior. But if you dig into it, you can usually find something. Is it possible that it would be undetectable?
Chris Mangold: Yeah, it’s a question of time. If you have all time of the world, it will always be detectable. But if you have all the money, this leads them to manual follow-ups, and this has been seen. I saw extreme cases – again, this is in luxury goods in extreme cases, they could even feel it as a service if somebody calls and says, “My friend has a huge checkout, we want to make sure everything’s in order. Can you do that?” So, the question is that you cannot just crawl through all the transactions. So, you have to make sure you detect the majority besides edge cases, which we then can cut with rules. But [24:44 inaudible] you detect and prevent the majority of the damage automatically. You cannot afford a 15-headed manual follow-up team who goes to each and every transaction, and you have to somehow be so clever that if you have not 15 but five manual follow-up team that you only give them the transactions, which are really somewhat worthwhile to look at [25:08 inaudible] swamped them. And then again, if you swamp them, again, you risk that something goes through. So, with all time and money, you will always detect it. But you have to meet in the middle. And again, the fraudster also does not have all the time and money, they will not make a deep fake dare to steal just a few game vouchers or whatever for card testing.
Bradley Chalupski: All right, awesome. Next anecdote here, next story.
Chris Mangold: What I find interesting, actually, what’s now, and I don’t know whether you have this [25:38 inaudible] also is 2FA, 2 Factor Authentication, hacking, and all these things. This has taught me a bit since this is really tricky, that without the contribution of the consumer, perhaps sometimes fraud – fraud is there, fraud is real. But sometimes it’s also a hoax. It was so much in the press, just recently on spoofing or hacking of two-factor authentication. It was really interesting. I read this article, and it was, by the way, from one of the big five – or it’s now big four – auditors who actually have a footprint in the market. And I said, “How can it be? How can you actually hack that without getting hold of the –” And then it was an edge case like some faking or some spoofing that you call the mobile operator and say, “Hi, it’s Bradley and I lost my phone. Can you send me…” So, it was just sometimes perhaps a fraud pattern which could be reported but which is not really there. Again, we talked about business cases, how effortful is this. Avoid manual interference also. So, how should a fraudster have the time without the lousy accent, like I have as you said, call a British mobile operator to say this story, and then you will again say, “Okay, I sent you a one-time password.” So, one should not believe everything since in business, the fear or use of fear is always great. And you can say, “We had so much back then in our account-to-account scheme with so forth.” Or the trustees of the [27:25 inaudible], they spoof your account, they check on your creditworthiness, they store the data. Why should they store data they don’t need? So, sometimes one has to also reflect on what they really did.
Bradley Chalupski: So, where do you see this bio authentication? This is something that’s becoming more and more prominent now. We’ve had the password apps for a long time now where you need to put in the extra code from your Google Authenticator. But now we’re moving more and more not just from thumbprints and face but I’ve heard things about possible voice recognition, I’ve heard retinal. We’re getting into James Bond territory here – things that are crazy. I guess here with your visionary CEO cap on here, do you think that this is where we’re going in the future that the password will eventually be retired and everybody will just be expected to scan their retinas? Or do you think that that type of world is too invasive, too much of a hassle to set up? And maybe people don’t want their retina data out there on the dark web versus just like a password, which is very unpersonal. Where do you see us going with that in the future?
Chris Mangold: I think there is a huge lobby for that. And this is the whole regulated business since it’s a nightmare, and actually it is even an unfair advantage of incumbent players in banking, in account keeping. Since it is difficult for new [28:56 inaudible] to get customers, this is an important process. The regulatory onboarding of a consumer – leave alone a business, but thereby will not be so far – is actually protecting very much [29:07 inaudible] 10 years. And then you had mail identity. I don’t know that you have this in the States or in UK, but you had to go to the post office earlier. Now we have this video identity, which is still really cumbersome – you have to hold up your ID and you flip it and you show. And then this is also cumbersome for the consumers since they are not only opening their main account, perhaps they were trying to just apply for a small retail loan for the next holiday or something like that, and they get it, and they’re happy, and then they have to go [29:38 inaudible], and they let it. Leave alone gambling, for instance, age verification, where they want to be quick. So, it this seems a bit spooky, I think there is a huge market because market participants needed to have a frictionless — We spoke about the checkout, but to have a frictionless onboarding. It’s not far away since this is really difficult to fake, it’s not far away from being regulatory accepted. Many financial authorities already accepting that, step [30:06 inaudible] anymore. And that’s why I expect a lot there.
Bradley Chalupski: On the consumer end, do you think that consumers are willing to participate in these things. I know consumers will always trend towards less friction. But to me, I don’t know, maybe I’m just getting too old now, I don’t know. But I don’t know that I would want my biometric information being stored in a company database. It’s one thing to have a password, and I understand those are going to get stolen – it’s one thing to have a Google Authenticator app and this. But now when you’re talking about Bank of America or Deutsche Bank is holding a scan of my retina or however else they’re doing this, I don’t know that I would be so comfortable with that, Chris. I mean, maybe I’m just showing my age. But do you think customers would just look past that because it would be more convenient? Or do you think that there would be some discussion between businesses and consumers about how this type of information is going to be held, handled, and ultimately, used and protected?
Chris Mangold: Yeah, obviously. First of all, it’s hopefully not a magnitude of companies having the data of your retina. This will be, hopefully, very well-controlled and high quality protected, and service providers or people in the middle who are doing that. And the other thing is, I guess, we are really showing our age there, both of us. I had this 15 years ago, where epayments just came and everybody, again, showing age and said, “How can you reveal your credit card data on the internet?” My argument always be like, “Guys, this was magnetic stripes back then.” In each and any delay, you give your card to somebody who disappears and comes back 15 minutes later with your card[32:02 inaudible] phone or you give your card, and how much more effort is it to spoof your card on the internet than there? There’s a better fraud that can happen. It’s showing age. And I have to say, especially, for instance, fingerprint – not the Google Authenticator but fingerprint – access to many of my life accounts. Even my normal, I cannot do a transaction, but access to my account is fingerprint. And this is much easier, and I have a lot of medium security apps there. I will be in a bad place if this would be switched off and I would have to remember or to reinstall all these passwords. So, sometimes, ease of access. And it’s two-layer – means, one thing is access, and next thing is, technically how precise your fingerprint has to be stored there for access to the account. And then when it comes to transactions or to emptying your account, most often that’s the next thing. So, data privacy is something. Sometimes this is even technically explainable that this is a fingerprint scan, but only, for instance, three or four data points, and you could not even reproduce the fingerprint from that. This is what consumers need to be educated on, but this is how it works. You know that, I know that, and I guess, convenience will help a lot there. There is what’s possible. And actually, it’s good. But you know that I spoke to a South African micro-lending company, they do it mobile-based. So, they get micro-loans. If you’re a trailer, you’re selling apples or potatoes or whatever, you get a micro-loan. And by the way, there is low risk here. And there, to apply, you had to open your location history in Google, they read location history, they find out whether she or he obviously has a mobility profile that indicates a daily go-to work, and they make this as basis of the credit decision, where we would say, “How can they be? How much can you trespass into the private hemisphere?” On the other hand, they get access to the funds. So, sometimes it just develops.
Bradley Chalupski: Yeah, convenience is definitely a huge motivator for consumers as we have seen definitely time and time again – so, I guess that this would be no different. So, you have time for one more?
Chris Mangold: Yeah, if you have.
Bradley Chalupski: Yeah, this is great.
Chris Mangold: We love good linking, this is really something, but [34:22 inaudible] to next thing in fraud prevention and artificial intelligence [34:27 inaudible] fraud prevention. The good thing is, this is pretty data protected with entity recognition, with behavioral linking, and so you do not have to have a customer file. We are not tracking and our competitors also not – the next-gen providers. We’re not judging on the behavior of a Bradley or a Chris. We’re seeing linking, we’re seeing all the signals. We see whether this is potentially advantageous – graphically – or malignant entity, not behavior. You are somewhat data slim – you do not have to store data since the only good data that cannot be stolen is the one you do not store – at least for this data, everything can be hashed. But this is really beautiful. And then you can even develop self-improving rules and all these kinds of things. This is the next thing.
Bradley Chalupski: So, take me through the science of this. I’m curious to hear when you’re able to use the linking to develop better models. What does that process look like, at a high level, obviously?
Chris Mangold: It’s big data. It’s detaching the data points totally from the attributes and actually collecting and calculating only with the attributes. And then, one is also reverse engineering what the machine learns – it’s obviously backtesting. By the way, you cannot do this totally without feedback – what was good and was rather inaccurate. But you [35:55 inaudible] the relevance of specific attributes and data signals. In the checkout, you get 40 or 50 attributes, you blow up to 2000 or 3000. And then, again, it’s a way for you to mainly detect them. And then you learn for specific demand situations, verticals, geographies, which are the most relevant data points, attributes for the score, then you link them to this linking first and linking the second time. And this makes it really powerful. Since still, you have to react in milliseconds, so you cannot crowdsource. It’s data design. And also you have to quickly decide which attributes you take into consideration.
Bradley Chalupski: And I assume how much weight to give to each one, based on a different situation. Well, Chris, we really, really appreciate your time here on the pocket. You started off by challenging everything that I had ever heard. So, right off the bat, this has been absolutely amazing to hear your thoughts on customer friction and on the future of technology. We really, really appreciate it. So, I’ll let you sign off, let everyone know where they can find you and Fraugster on the internet, and then we’ll call it a wrap.
Bradley Chalupski: Thank you so much, Chris. Really appreciate it.
Chris Mangold: Bradley, it was a pleasure.