China ad fraud has been in the headlines recently, and we wanted to have a show about it. As expected, getting someone on the show to discuss this sensitive topic has been impossible. But we finally found one: Dr. Augustine Fou. He is the Founder of FouAnalytics and a leading expert in the field, boasting a profound understanding of digital marketing, cybersecurity, and advertising. With a Ph.D. from MIT, Dr. Fou has been on the frontline of programmatic and digital marketing for three decades.
1. What is ad fraud? What is it, and how does it work?
2. What’s the motivation to buy fraudulent ads?
3. Are impressions still an important measure of advertisement effectiveness?
4. Can we buy on the big platforms where the humans go?
5. Why do they engage in ad fraud if they’re not getting the results? The marketing/sales disconnect
6. Is there an opportunity for new and emerging brands to use advertising as it was meant to be used?
7. Can AI help relieve the ad fraud problem? The answer is no.
8. How do brand relieve their addiction to ad fraud practices?
9. How do you drop in tags that may reveal ad fraud (that may ultimately rejected by some publishers)?
10. Where do you think ad fraud relief will go? Enter the CFO.
11. A/B Test: Humans, 1 Billion, Platforms, and Humans
Dr Fou on LinkedIn: https://www.linkedin.com/in/augustinefou/
Try FouAnalytics Platform Yourself (it’s free): https://fouanalytics.com/
ShanghaiZhan Theme Music: by Bryce Whitwam
https://soundcloud.com/bryce-r-whitwam/bad-cough-syrup?si=cfb30a6e0c0e459da78b912bf60825ac
Bryce on Linkedin: https://www.linkedin.com/in/brycewhitwam/
Ali on Linkedin: https://www.linkedin.com/in/alikazmi/
TRANSCRIPT OF THE SHOW:
Welcome to Shanghai Zhan, a raw and lively regular debate about China tech advertising, creativity, platforms and the intersection of it all. Join us each session for timely and relevant discussions on all things China marketing. We will be joined by an entire spectrum of China experts and you can learn more about ShanghaiZhan, all of our past episodes on zhanstation.com. I’m Bryce Whitwam and I’m Ali Kazmi. As always, we like to thank everyone listening for their continued support. And if you really like the show, make sure that you let your friends know, copy and paste the links uh onto your favorite social pod social media platform or better yet. Go on Apple podcast, Spotify, Xiao Yi Zhou and give us a five star review, Ali, we’re talking about your favorite topic in the whole world ad fraud today. And uh this is a topic that we’ve always wanted to talk about on the podcast. But as we found out it’s a bit difficult to find people in China that will need to come open in public to talk about ad fraud. But today, we’ve really lucked out. We’ve got an amazing guest, Doctor Augustine Fou is our special guest. Doctor Fou is a distinguished expert in this field uh based in New York City, boasting a profound understanding of digital marketing cybersecurity and online advertising. Doctor Fou has been in the front lines of digital marketing for almost three decades. It’s from that vantage point. He studied and documented the nexus of cybercrime and ad fraud. As an investigator, Doctor Fou assists government and regulatory bodies as a consultant. He helps clients strengthen cybersecurity mitigate threats and risks including the flow of ad dollars that fund further criminal activity and improve the effectiveness of digital marketing campaigns. He’s also the founder of FouAnalytics, which is an online platform that everyone can use. Uh, He’s got a Ph.D. from MIT, an impressive career, worked at Omnicom before, and has parallel insights into the challenges faced by the digital industry. Uh, He’s a sought-after speaker, consultant, and thought leader. We’re really happy to have him here. Looking forward to this conversation. Doctor Fou. Welcome to the show. Thank you very much Bryce. Uh Very glad to be on the show with you. That’s great. So let’s just start with the basics. I mean, ad fraud is prevalent, it’s prevalent in China. Uh statistics from actually from WPP say it’s close to 60% of advertising programmatic is driven by bots. It’s definitely a problem not just in China but all over the world. Doctor Fou, could you just talk about what is ad fraud? Uh some of our audience might be novices at this. So we can kind of give them a quick explanation about what it is and how it works. So, digital ad fraud is basically ads that are not seen by humans, right? So the way digital advertising works is when people visit websites, when they load the page, the ads load. Uh So that’s how it’s supposed to work. But when you have a whole bunch of smaller websites that don’t have a lot of human audiences, some of those sites might resort to cheating. And so they go out and buy traffic. So literally, if you google the term, buy traffic for my website, you will find hundreds and hundreds of vendors ready to sell you traffic, right? You can say I want to buy 10 million page views and they’ll sell it to you. And common sense will tell you that there’s not a whole bunch of humans sitting around with nothing to do but to go to your website when you tell them to. So basically that is all automated activity, which is a nice word uh for bots, right? Bots are just automated browsers that developers can code and tell it to repeatedly load web pages. In that case, the reason we call it fraud is because the advertisers are not getting what they paid for, right? So advertisers think they are showing ads to humans which will then drive a marketing outcome for them. But now when the ads are being shown to automated browsers, obviously, it’s not going to drive any business outcomes for the advertisers. So that’s why we call it fraud. And that’s kind of how it grew up in the past when most ads were shown on websites. Uh those websites would just buy, bought traffic and juice their numbers. But we’ve also now moved into mobile and mobile apps and they’re doing basically the same thing. So instead of websites, they’re using mobile apps and there are something called mobile emulators, right, which emulate a mobile device. So they can load up 507 109 100 mobile apps and play them continuously. So then the ads just load continuously. And again, it’s a disservice to the advertisers because they’re not getting ads shown to humans, they’re just getting huge volumes of ad impressions and they think it’s being shown to humans. That’s why we’re calling it fraud. Maybe this is a silly question. But why would people reach out to these companies to buy, buy fraudulent traffic? What’s what’s the motivation? Well, obviously they’re not deliberately buying fraudulent traffic, fraudulent ads, they’re kind of tricked into doing that. So a lot of the advertisers, you know, coming out of traditional channels like TV, or print or radio, uh a lot of the advertisers think in terms of reach and frequency, right? So if you want to reach a very large audience and you want to show them more ads more frequently, uh the those ads would achieve a branding effect, right? It would cause the users to remember their product and then maybe hopefully go buy their product. So similarly, when they take that mindset and translate that into digital, what they’re trying to buy is as many ad impressions as possible, shown to as many humans or many as many users as possible, right? They’re still using that region frequency mentality. So the media agencies that they typically employ to buy the media for them also takes that mindset into digital and it typically translates into oh if I can buy 2 billion ads, that’s got to be better than 1 billion ads because that means I got more reach or more frequency, right? So over the years and in fact, over the last 10 years, the advertisers have just demanded irrational numbers or irrational quantities of impressions. There’s absolutely no way the 7 billion or 8 billion humans on earth can be generating 20 trillion ad impressions per year uh for the advertisers to buy, right? So when we’re talking about trillions upon trillions of ad impressions available for sale, you can understand that that is completely out of whack from reality, right? And because the advertisers just want the larger quantities, uh they’re now also somewhat willing to look the other way and say, oh, well, that’s got to be real, right? Even in China, I can count many times when our clients would actually deliberately tell us, they would imply that they wanted to be able to get more impressions from the campaign. And they look to us to think of ways to do that, which we thought was just a sign for us to go out and buy some. I guess the one question I always have is and maybe we can get into details is impression still an important measurement of effectiveness of advertising because the numbers are so huge. Uh And I remember doing a campaign for IKEA on Doyen and we counted 3 billion impressions. So that, that, that, that the ad was watched, seen by at least every Chinese person twice. Uh So I knew that wasn’t happening. So is impression still a useful measurement of advertising effectiveness? No, it’s actually never been useful. Uh But let me be a little more specific in the difference between traditional channels like TV, print and radio versus digital channels. So when you’re talking about TV, think about a 60 minute episode or 30 minute episode, there’s only so many minutes out of the 30 minute episode that you can use for ads, right? So, you know, there’s some examples here in North America where, you know, 11 or 12 or 15 minutes out of 30 minutes is used for advertising, right? But in those cases, there’s actually scarcity, right? So there’s only a finite amount of air time that you can use for advertising. But in digital, we’ve now kind of dissociated ourselves from reality. Um So now we can have unlimited fake websites generate unlimited fake ad impressions. So that’s why it’s problematic to take the reach and frequency mentality into digital because we are no longer constrained by physics, right? We’re no longer constrained by the physical limitations of the physical world. So in digital, it is actually even worse for the advertisers to think in terms of how many billions of impressions they can get now beyond that quantity metric, right? So one of the reasons it’s used so often, right, not as a measure of success, but just as a measure of how the how much reach the advertiser got, it also comes along with a lower C PM, right? Lower prices. So instead of buying from legitimate publishers which have to pay editors and journalists to write content. And so they have a finite human audience, some of these fake websites, they can generate as many impressions as you want to buy, right? They can literally eat up your entire budget for you. So it becomes a game of big numbers. And I think the the common thought is like, oh more is better. So they said, you know, 10 billion is better than 5 billion. So if you can buy me 10 at better C PM prices, I’ll buy from you. That has just kind of happened repeatedly over the last 10 years. And so now we’ve gotten to a point where I mentioned earlier, trillions upon trillions of ad impressions are sold and it’s so out of whack from reality that, you know, sometimes people are hesitant to go back and question it. Right. Because again, they’ve already told their boss last year we bought, you know, 5 billion this year, we’re going to be buying 8 billion. Right. We’re going to do even better than we did last year. So, because those numbers are easy to record and report, um people have been using those to report up the line. And so that just makes it, you know, compounds the problem. So when we talk about trillions of impressions intuitively, you would probably as an advertiser gravitate towards any of the six leading platforms. And so, you know, in the China context, we’re talking about doin we’re talking about Alibaba, we’re talking about Ken uh Xiao Hong Shu in the US. You have an equivalent and why wouldn’t an advertiser just stick to the top six publishers and just advertise on those larger publishers? Yeah, where the humans go. So what’s what’s why, why do we have this problem? Why traffic if 90%? I don’t know. I, I I’m assuming that 80 90% of the audience is going to stick to the six big platforms. Yeah, that’s right. A very simple exercise that I run in class. So I teach a marketing class, undergrad and graduate students. I ask a simple question. I asked them to name or write down as many websites as they possibly can in one minute. Right. Maybe name off 10 websites that they visit every day. So they’ll rattle off the 1st 3456. And then by the time they get to seven or eight they start slowing down, they can’t even get to 10 websites that they visit every day. And then I do the same thing with mobile apps. I asked them what name of the apps that you use every day, right. So they can’t even get the tent. So most humans go to the large publishers, the large websites, the large social media networks repeatedly, right? There might be some that go to the long tail websites once in a while, right? You might look up a recipe for baking sugar, cookies or something and you visit that, you look at the one recipe and then you’re gone, you’re not going back to that site over and over every single day. So in that case, the human behavior is that you’re right, they’re, they’re going to a very small number of very large websites. But that being said, what the advertisers have been taught to buy by their agencies is, oh, you know, these publishers don’t have enough ad impressions or the the number of impressions don’t grow fast enough, right? So we’re going to go help you buy ads to programmatic ad exchanges, ad exchanges, which we have tens of thousands or, you know, hundreds of thousands of websites and mobile apps that sell their inventory through the exchange and we can get you much more quantity. So the advertisers over the last 10 years have become addicted to irrational quantities. Right. Literally, those quantities don’t even make any sense. But again, because it’s happened for so long and along with that, the fraud detection companies of the past 10 years have not marked most of those as fraud. Right. In fact, they marked only 1% of that as fraud. So the advertisers think, oh, wow, we got so much more quantity to buy over here instead of buying from the main platforms of the main publishers. And oh, by the way, the CP MS are much, much lower. We’re getting a way better deal over here. Let’s go move more money over here. So, in fact, if you look at the, what’s happened to the mainstream publishers right here in North America and also around the world, we’ve seen revenue declines across all of the major publishers and also margin compression. Because if you’re in New York Times, it’s going to be very hard for you to sell ads at $35 CP MS when you have competitors selling them at 30 cent CP MS, right? So the advertisers and the media agencies again are shifting their dollars over here chasing the large quantities and low prices. Like you said, Ali it’s common sense they should do that and I’m actually hopeful that we’re going to kind of swing back to that where we’re starting to buy from fewer and fewer websites where, you know, the actual humans are, are congregating. Right. But it’s a very hard bad habit to break after 10 years of that, I guess, Dr Fu, if they’re putting money into these lower price outcomes, but actually not getting the results, why would they continue to invest there? Wouldn’t I want to see the results as well as, you know, cost savings? Yes, one thing but also am I getting the results? Well, in this case, I’m not because uh I’m buying path on platforms that are, are basically, you know, it’s all, it’s all bot related. Yeah, it’s all fake accounts and automated. But here here’s where some of those uh digital marketing mix models come into play. So to the uninformed marketer, right? Who really didn’t get past high school algebra because they’re English majors, their art majors or whatever. Yeah, I am kind of insulting to the marketers, probably insulting us too. But basically for them it’s like, oh here’s these fancy models that tell us we got so many sales. They actually never connected it to the actual sales. Let me, let me give you a very specific example when you’re dealing with a F MC G, right? A consumer packaged goods manufacturer. Uh they rarely get to see the lift in sales that occur in the grocery store, right in the very last mile where the consumer comes into the store and buys their product, they typically sell a very large quantities at wholesale to regional distributors, right? And then those distributors sell it to the grocery stores and the other retail stores and things like that. So they are almost like one or two steps removed from the end purchaser in the, in the physical stores. Ok. So in that case, it makes it doubly hard for an F MC G advertiser to measure whether there was a actually any lift in sales. So what they do is these kind of hand wavy mathematical models to approximate that. But the problem is you can’t really draw a cause and effect, right? You can’t draw a straight line of cause and effect, like run these digital ads and it cause these sales, you can’t do that. So everything is approximated but how convenient because now they can say, oh yeah, we believe that this lift in sales was caused by the digital marketing that we did over here. The only way they can tell whether the digital marketing actually drove those sales is what I’m going to call, turn off experiments, right? They literally should turn off the digital marketing and see if there’s any impact, any negative impact on the sales and those have occurred. But they’re extremely rare because most marketers don’t have the courage to run those kind of experiments, right? Most marketers believe that if they don’t spend it. All right, if they don’t spend their budget in full. They’re not going to get as much budget to spend next year. So, their top priority through the year is to ensure that their agencies help them spend it all. But in the few examples that we do have, I don’t know if you remember uh Procter and Gamble, I believe in 2018 or so turned off $200 million of digital ad spending and they saw no change in business outcomes. Uh Similarly Chase Bank, they turned off the majority of their uh programmatic websites, right. So they were showing ads across 400,000 websites. They reduced that all the way down to 5000 websites. So that’s a 99 99% decrease in the number of sites showing their ads and they saw no change in business activity on their websites. So those are extremely rare and very few advertisers are willing to run those kinds of experiments. And in lieu of that, they’re just using these fancy mathematical models to approximate. We may have gotten all these sales due to the digital advertising running. But what they don’t say is that, oh, yeah, we also have TV ads running. We also have print ads running. We also have billboards running, right. So you really can’t assign a cause and effect to the digital ads themselves. But because they’re willing to believe that the ads did something, they’re willing to continue to buy large quantities of low cost ads online. I was just going to say, aren’t they already preconditioned doing that? Because living your TV was also measured based on just boxes on top of television that gave you an approximation of how many people. That’s right. It’s all approximated. So there, there it’s good enough for them. No one’s going to really question the mathematical methodology of those things. No one’s going to even ask. OK. Well, is this even a representative sample? I mean, if you have 100 and 50 households, does that actually represent the entire nation? I mean, literally, I think the panels are, you know, 5000 households. To me, that’s too small of a sample. It’s not going to be representative gave reference to PNG over here as well. And PNG obviously is a home care business and they turned off advertising and that had little to no effect on a business outcome. Are we then saying that there’s an opportunity for new and emerging advertisers or products that want to penetrate this this home category to use digital advertising in the way it’s meant to be used? Or are we saying that home care categories just doesn’t have any impact on consumer behavior? No, no, it’s, it’s any advertiser regardless of industry, right? If your ads are not shown to human beings, the human beings are not going to be aware of your product and therefore they’re not going to buy your product. Now, for some of those legacy brands, right? The ones that are already established, they’re also wasting money on the advertising because I’m going to go buy my shampoo that I’ve been buying for the last 10 years and I’m going to buy it anyway. So more, more advertising, even if I saw the ads, right, it’s not going to cause me to buy more. Right? I have one example I use in my writing. It’s like I can only drink, my family can only drink four quarts of milk every week. No matter how much advertising the Milk Council does. I’m not going to buy 1/5 quart of milk because we physically can’t drink it. Ok. So there’s a lot of this kind of stuff where they’re just advertising way the heck, too much, they’re trying to buy way the heck too much, too many digital ads. And they’re not really asking hard questions about whether those ads are even being shown to humans. Right. So now going back to your question, I’ve actually seen categorically that small and medium businesses uh who have very little budget to spend in advertising are the ones who are the most vigilant about their spending. So they’re constantly looking at the analytics, they’re constantly looking at whether any ad spending is driving sales because if they spend $100 or $1000 on ads and it didn’t drive any sales, they can’t spend the next $1000 on it. Right. They are a small business owner, a medium business owner. Right. Whereas a big company like PNG, I believe a few years ago, their digital budgets were $2 billion. Right. There’s also very large banks, airlines, uh and other types of companies where they have hundreds of millions of dollars to spend. There literally is not enough real stuff, real ad inventory for them to spend all that money on. Right. So, because they have too much money to spend, they end up kind of incentivizing the fraud, right? If you go to a New York Times and said, and tell them, I want to buy a billion ad impressions. New York Times can’t sell them a billion impressions because they don’t have enough humans visiting the websites and generating the page views, right? The human might go read one or two articles, look at two or three pages and then leave. So they can’t magically, you know, come up with billions upon billions, more ad impressions to sell. But the sites and apps that belong to uh ad exchanges can magically come up with as much ad inventory as you want to buy. So do you see how, how those are literally, you know, if you use a little bit of common sense, you can tell that that can’t be possible, right? You can’t just manifest billions upon billions of ad impressions without using some kind of cheating, like using bot activity or automated browsers to do it. So would a smaller advertiser will be more skewed towards advertising on one of the top sort of six platforms. Are they going to be more skewed towards long Tail? Absolutely. I’ve seen very, very successful examples where, you know, if you think about uh search, right? We, we commonly say people Google things, right. So if you advertise on the main property of Google, people are going to see your ad when they’re searching, right? When they’re typing a keyword in, that’s a great place to put an ad in front of them, right? At the moment, they are looking for something, right? That’s the ideal time and make sure you turn off the search partner network, which are all the sites outside of the main property, which also run Google ads, right? Because that’s where the majority of the fraud is going to be. And then similarly, uh think about Facebook, there’s a lot of humans that are still logged into Facebook all day long, logged into Instagram all day long. So if you want to get your ads in front of them, that’s a great place to put your ad, right? You don’t need huge quantities even, right? But that’s a great place to get your ad in front of humans. And I’ve clicked on some stuff that I’ve discovered on Instagram and I ended up buying something, right? So it kind of ties to your earlier question with small businesses and start ups. Be able to compete against the giant giants Right. Yes, they absolutely will be able to compete and they don’t even need to spend that much. In fact, the more strict they are in terms of the money they need to spend in digital and also where they put their ads, they’re going to be so much more effective than the largest of advertisers who have just way, way too much money to spend. So, one thing that continues to kind of plague me a little bit is as a smaller advertiser, wouldn’t I be more, would I not be kind of skewed towards buying programmatic advertising? I’ll try to explain some of the thinking. And then, and then you tell me where the flaws are in my thinking. If the advertiser, if the small advertiser were just chasing low prices like low CP MS, it would be, I mean, let me explain the thinking, let’s stick to the conversation on home care product. And then maybe we talk about, you know, I don’t know, uh laundry detergent, for example. But if you’re, you know, if you’re looking for laundry detergent, it may be that you’re going to go to a local retailer that does delivery of, of detergent and home care products directly to your home. And so as a result, you’re skewed towards buying a product on a lesser known platform or a lesser known website or lesser known retailer versus going to a Google. And so as a result, you know, I, I can think of other categories as well. For example, you know, if you think about cars, right, you wanna, you wanna buy a car from a local dealership as opposed to going directly to a Volkswagen or Tesla. And so you end up making decisions that are a lot closer to home. And as a result trying to get that or capture that type of traffic, are you therefore going to be skewed towards buying through a programmatic platform and and and consequence of that you end up getting lower C PM. Does that make sense or? No? I mean, you can do the same like if you want to do Facebook ads and just limit it to a very specific geography like a city or maybe even your zip code, you, you can do that. So you don’t need to buy programmatic ads to do that, right? You can buy it on the platforms and and do a very specific geography. And so for, you know, local businesses like hairdressers, bakeries, donut shops, restaurants or whatever, uh they can use Facebook advertising to great effect, right? And they can also add the social component to it where they’re posting pictures of the food, the menu items and whatever and all of that creates kind of like a little community around it. So the problem with programmatic is even though it has very low C PM prices and you know, some of these ad tech vendors have been touting almost magical technology like oh the right ad to the right person at the right time. What most people don’t ask about is the accuracy of the data or I should say the lack of accuracy of the data. So in most cases, the websites and the mobile apps that are selling the programmatic ad inventory, they don’t have access to the GPS in the phone, right? So they’re not reading the actual location, they’re simply approximating it. By contrast, Facebook and Instagram, the users typically give that those apps the permission to read the actual GPS location. So not only can you confirm it’s a real person logged into Facebook all day long, but you can also confirm that the location is accurate because they’re reading it from the device sensor. So other than the problem of fake accounts on Facebook, if you’re advertising on Facebook and doing local ads, uh those will be very impactful for small businesses or local businesses, why would you need, why do you need doctor F? No, no, you don’t. You don’t. I I basically tell people if, if you’re buying Facebook ads and you have turned off audience network if you’re buying Google ads and search ads and you’ve turned off search partner network, if you’re buying youtube video ads and you’ve turned off Google Video Partner Network or buying Tik Tok and turn off Pango, right? Which is the audience network you don’t need me. I mean, you’ve avoided 95% of the fraud just by doing that part of what, what I’m doing right now is also helping them with things that are outside of fraud, right? So with the analytics to measure your digital media, right? There’s Google analytics, but it’s made for websites, you really can’t use it for the ads. So food analytics is meant for the digital ads where we can measure them. The idea is that there are other things that are not fraud, but that are still sub optimal for the campaign. For example, if the advertiser forgot to turn off the overnight hours, then they’re wasting a lot of impressions between 1 a.m. and 5 a.m. when most humans are sleeping, right, they might as well turn off those hours and save the money so that more of the ads can run during waking hours, right? So that’s not a fraud issue. It’s simply they forgot to turn off the overnight hours. There might be pacing issues where the majority of their ads are used up right after midnight, right? Because there’s probably like a daily budget and once the day ticks over, you’ll see a huge surge in volume at midnight, 1 a.m. and 2 a.m. and by 3 a.m. they’ve used up their daily budget. So there’s no ads to run during the rest of the day anyway. So again, that’s the role of analytics, right? I started off doing fraud detection, but now it’s like a more mature kind of platform, right. These big advertisers and small advertisers might have used double verify or integral ad signs these fraud, these legacy fraud verification vendors, which basically said it’s 1% fraud for the last seven years. We now need to upgrade the tools so that the advertisers big or small have better analytics to look at for their digital media and their websites. So it’s not just a bot or not kind of issue anymore. What do you think of the some of the innovations in, in the, in the ad detection fraud industry? And, and is there a role for A I in doing this? I, I would assume that there would be, I’ll, I’ll basically nip that in the bud and say there’s absolutely no role for A I in fighting fraud. OK? The reason I say that is the A I is basically set of algorithms and it needs to be trained on something. OK now, and, and I’ve looked into this for 11 years and still to this point, there’s no good A I, they can reliably detect that because what the A I is trying to do is detect patterns that it’s seen before, right? So you can basically train it with data that says, oh, here’s what bots look like, right? So they’ll have that training data, here’s what humans look like. So they have that training data, what the A I is going to be good at doing is basically finding the stuff you told it to find. It is very, not good at finding new stuff that you didn’t know to anticipate. OK. And the other thing is even though a lot of these A I proponents or machine learning proponents will say, oh, it’s so magical, it’s so magical. And you know, machine learning is supposed to learn and grow and evolve and all that kind of stuff. It’s all BS, it’s all if, then statements that they basically call A I and it’s not going to catch or outsmart the bot makers who have lots and lots of experience. So the bot makers are humans, right? They know that when you’re looking for these types of things. So for example, if we’re looking at mouse movement, if we see mouse movement, you would think that it’s humans. So the bot maker then tells the bot to go simulate the mouse movements, right? So now they can appear to be human. And if your A I is trained to say, oh mouse, if we see mouse movements, we’re going to label it as humans. If we don’t see mouse movements or whatever, then we, we label it as bots. If they’re trained on that, then in fact, they’re going to be assisting in covering up the fraud more than they’re going to actually discover and find the fraud. So I really want to stress here that A I is not the solution to everything and it would be a very bad idea if the advertisers were duped into thinking that A I can actually detect and solve the fraud problem. And to be more clear, ad fraud is not a tech issue. It’s an incentives issue. Right. So let me be more clear. The advertisers wanted to buy the large quantities. All the middlemen like the exchanges wanted to sell the large quantities because the more volume that runs through their system, the more money they make the media agencies that buy this stuff, want to buy the larger quantities and want to buy the lower cost stuff. And the only places to find that would be the fraudulent sites, you can’t find low cost large volumes from legitimate publishers, right? Because they’re limited by the human audiences that actually visit the site. So the only place you can get as much quantity as you want to buy and as low prices as you want to buy is through programmatic exchanges, most of which is powered by automated activity, right? So it kind of goes in a full circle and you know, basically don’t buy into the hype to put this into full circle into the China context because I think what you just said was spot on in terms of the culture that Ali and I, you know, for the last 20 years have been involved in it is simply that the agency’s responsibility is to make the client look as good as possible. And uh we work for global agencies, we are entrusted with the fact that we cannot at any means times participate in any kind of fraudulent or, or seemingly fraudulent activities. But despite that, there’s an incentive especially amongst even foreign companies in China to go to local companies who will simply look the other way. The reason for that is that everybody’s neck is on the line. It’s no different in China as it is in the West is that the CMO has the shortest time frame of a job in a company and that you are entrusted to be able to deliver results at any cost possible. Even if it’s BS results, how can we start to stem the tide and starting to bring back marketing advertising is to, to humans not to, not to bots. I think the the short answer is help, the people who want to be helped. Ok. There are too many large advertisers who don’t want to be helped. And I’ve seen this firsthand for the last 11 years in the US. Right. There are many large advertisers who don’t want to hear about fraud. They like to be taken out to fancy dinners by their agency and they’ll commit. Oh yeah, we’re going to commit to spending $200 million in digital advertising and we’re gonna, you know, sign the agency as our agency of record to spend all our money for us. So there’s those kind of advertisers, you really can’t help. And I think if we try to solve it for the industry. It’s too monumental a task. So what I’ve done over the years is basically, I try to show uh the data in the charts and all the writing that I published and some advertisers will actually reach out to me and ask for help because it, their gut feeling says, OK, well, something is still wrong here, right? We’re, we’re getting all these clicks. But when I look at my Google Analytics, it’s like it’s 99% bounce. I saw all the clicks that got to our website, they left right away. That’s kind of not characteristic of a human, right? If a human deliberately clicked on an ad, that means they wanted to go to the site, right? And so they’ll probably do something there if we see clicks that land on the site and then nothing else happens after that. Uh That’s a low value click, right? We’re not even talking about bot versus human, right? The user just didn’t do anything subsequent to arriving on the site. So in those cases, we can start asking questions, right? And that and that’s how we can start to educate the advertisers. And then when they take the reins and say yes, we want to do something about this. Yes, we want to make some change. Um They have to lead that because if ever the agency got in the middle of it, the agency is going to resist all those changes, right? The advertiser has to be strong enough. They have to have the courage to lead the change for the better. And some of that involves understanding that when we reduce the fraudulent volumes, right, the volumes are going to go down, the clicks are gonna go down and the C PM prices are gonna go up. Right. So those things have to change and sometimes we can’t change them overnight. So I’ll also tell you some of the large advertisers that I work with, you know, I leave it up to them to uh kind of effect the change as quickly as they want or as slowly as they want. Right? So one advertiser might be able to clean up most of their campaigns in six months. Another advertiser might want to actually kind of do it more slowly and take 12 to 18 months to clean up their campaigns because they can’t all of a sudden go back to their bosses and say, whoops, we just, you know, blew 80% of our digital budgets on fraud, right? They can’t do that. So we have to help them slowly and progressively clean up the campaigns. What does that mean? Sorry. What do you mean by clean up the campaigns? Like, how does that happen? Is that, is that, is that, is that, I mean, what is it looking metrics? Is it looking at the types of inventory that you’re investing in? Is it what’s the process of doing workshops is it training? What is it, what does it take for um for you to help uh you know, that type of transformation um within an advertiser? Yeah. So when we put the food analytics tag in the ads themselves, we can see where the ads went and whether a bot or some other form of fraud caused the ad to load. Right? So the first thing is if we see that the ads are going to not good places like porn sites or, or piracy websites or whatever, uh then we basically tell the agency you got to block these sites, right? So that helps to make it a little bit better. The second part is if, if we’re seeing a lot of bot activity, automated activity, loading the ads, we can see which sites and which mobile apps are, are doing that, right? Some of the mobile apps like a flashlight app should not be loading ads continuously throughout the day, right? How often or how long do you actually use a flashlight app or how long do you use an alarm clock app? Right. Probably for a minute at night, setting up the alarm clock and then a minute in less than a minute in the morning, turning the alarm off, right? But when those apps are fraudulent apps, they’re going to continuously load ads in the background and obviously none of those are seen by human, right? So when we see those kind of behavior, right? When we use the Food Analytics platform to measure the ad impressions, we can say here’s a list of websites and here’s a list of mobile apps that we want the agency to add to a block list. So you can do that cleaning, say once a month, once a week, however often you want to do that. And that’s what I mean by progressively cleaning the campaign. Right? We can’t chop off 90% of the volume because it’s fraudulent. We can’t do it overnight. Right. We need a kind of a transition plan and that’s how I work with the advertisers to help them make their campaigns better and more effective. The sound bite that I want to add to this as well is that we uh when we did any kind of ad fraud detection, um one of the challenges that we had was also just dropping in tags on the inventory that we bought from some of the leading publishers. So, so food analytics tag, it’s very likely it’s going to be outright rejected by a publisher because of the amount of transparency that you’re willing to provide the advertiser. Have you come across that in the US as well? Absolutely. Right. And, but I actually use that to my advantage because a legitimate publisher who’s not deliberately trying to cheat and commit fraud. Uh Yes, we have to go through approvals to get the code on the page, but they’re not going to resist. Right. But if you, if a fraudulent publisher, right knows they’re going to commit fraud, they, they will outright refuse to put my tag on the page, right? Because they know I’m gonna, I’m gonna catch them. So what some s SPS are doing with food analytics, they’re actually using it behind the scenes across the entire platform and they use it in two different ways. First is to vet the publisher before they let them into the SSP into the exchange, right? So we want to measure on their website to see if they, you know what percentage of the traffic is fake. Um And then we also tell them you have to keep food analytics on your site as long as you want to keep making money from the SSP because we’ve seen cases where the website or the publisher is on its best behavior before they get on boarded. And then the moment they get on boarded, they start cheating, right? They start buying traffic again. So that’s a way for us to monitor that and make sure they’re not cheating. And then the second thing is uh we continuously monitor across the entire SSP so that we can see if any of the publishers start to do some shady things and we can give them warnings. So it’s kind of like in baseball, you have three strikes, right? We say we’re seeing this kind of shady activity or bo traffic or other other things like pixel stuffing or ad stacking uh what’s going on with your site and if there’s three strikes, then we basically kick them out and don’t let them monetize it anymore, right? So there’s different ways you can use the platform to help monitor that kind of stuff. This is really fascinating and I think I can see where it’s not just about, it’s just not about fraud detection and proper analytics. There’s also a consultative element to it as well. And literally, uh it’s almost like like fraudulent rehab, where you have to gradually take uh the, the, the brands uh away from the, you know, fraudulent practices in a gradual way. My question, I guess is, where do you think this is going in the next, you know, 10 to 12 months? Yeah, that’s a great question. Um I really see the CEOs and CFOs stepping in and playing a much greater role because previously they would kind of give an annual budget to a CMO and the marketing organization and the marketing organization would basically go spend it and then report back with success metrics and stuff like that. But I think, you know, over the last 10 years, I’ve seen too many of those success metrics consist of vanity metrics, like number of impressions, number of clicks traffic to the website, or how low of AC PM we got right this year. So in those cases, the CFO realizes that, you know, it’s not about how much traffic you got, it’s really more about, did you drive any incremental business? Right. So, for every dollar you put in, did you drive more than a dollar of sales or outcomes for the business? So, I think the marketers haven’t been strict enough on that and they haven’t pushed hard enough on that in the past. They’ve just accepted these vanity metrics. So going forward, you know, I see CEO S and CFO stepping in more aggressively to ask the harder questions and to really kind of force the CMO the marketing organization to refocus on business outcomes, not how many clicks they got, right? And I do see that the marketer will need to um you know, make some changes. They can’t just say, oh, well, we bought this many billions of ads and these fraud verification vendors, right? These legacy fraud verification vendors said, oh, there’s no fraud, don’t worry about it. Just keep spending. They can’t do that anymore, right? It’s not free spending anymore. And we all know that budgets are going to get tighter, going forward. So they really have to prove that they are driving business outcomes. And I think that’s where we’re seeing the transition from fraud verification to better analytics, right? So in the past, when you have a fraud verification vendor tell you it’s 1% bots, what do you do with that? Right? There’s really not enough detail, you know, first of all, it’s 1%. So you think it’s, it’s low enough, it’s fine. I don’t even need to do anything about it. Um But they also don’t give you enough details like which sites were fraudulent or which apps were fraudulent. Right? So you don’t have that data to then go tell your agency. Can you please block these sites and apps? So, with better analytics, not only can you find the fraud and then assemble a block list, right? You can tell your agency here’s a whole bunch of flashlight apps and alarm clock apps and whatever that you need to block because we don’t want to buy from them anymore. They’re just committing fraud, right? So with sufficient detailed details in the analytics, you can take uh better action, right? You can see better so that you can do better. And that’s why it’s important to have analytics for your digital media. And do you remember the examples before or earlier in the show where I said, you know, sometimes the agency simply forgot to set frequency caps or day partying, right? Where they turn off the overnight hours or they forget, forgot to set pacing. So all the ads are used up between midnight and 2 a.m. right? So those are again, not fraudulent things, but that these are things that the analytics will tell you. Uh then you know, when you see that you can have your agency make those corrections, right? So again, collectively, I see a shift from legacy fraud verification to better analytics for the digital media. And I see the CEO S and CFO S stepping in and kind of guiding or forcing in some cases the CMO to pay more attention to the business outcomes for from the digital marketing. It’s really interesting. Uh uh That’s and yes, absolutely relevant to the China market as well. Uh I can see that what you’ve suggested here is a way forward to uh really taking a crack at, at this massive fraud problem that exists. Ali are we ready for the A B test? Yes, we are. We’re ready for the A B test. I’m gonna throw two words at you or there might be two phrases you have to make a choice between one or the other. Um Whatever comes first to mind, impressions are clicks, clicks uh reach or frequency reach. Humans are bots humans, cookies or cream cookies. 100,000 or a billion, a billion. That was actually what I didn’t say is I was going to put a dollar figure next to it, but you’re going after the bigger budgets platforms or programmatic platforms. Excellent humans or A I Humans brand or business CMOs or CFOs fraud verification or analytical insight, analytical insight, Dr. Fou. Thank you for being on the show. Sharing us with this interesting insight. It’s we really wanted to have a discussion about ad fraud and it’s a great way to start the conversation. So Thanks for being on the show. Thank you. It was a really great show. Happy to, you know, we’ve covered a lot so very happy to chat with both of you, Bryce and Ali. And, uh, thank you for having me and thank you for everyone for joining us on today’s episode. Join us in a few weeks for another exciting show and to all our listeners until then, have a great day.