Brett (00:00)
Hey everybody, welcome back to Signal & Noise. We have a really exciting episode called Ad Fraud are the Bad Guys Winning, where Rio and I talk with Dr. Augustine Fou. And for those of you who don't know who Dr. Fou is, he's pretty well spoken about in the AdTech and MarTech world, but he's an oft quoted, occasionally controversial, and always deeply informative

digital marketing executive and ad fraud investigator. And he's been on the front lines of digital marketing for nearly three decades. So we had a very interesting conversation with him. He comes to us from his own firm called Fou Analytics. He was the group chief digital officer of Omnicom's healthcare consultancy group, which was about a hundred million dollar group with eight agencies serving mainly the pharma industry and healthcare clients.

He was also at McKinsey, was the SVP of Digital Strategy at McCann, MRM Worldwide, and he taught digital strategy and integrated marketing at Rutgers, ⁓ my home state of New Jersey, and NYU's School of Continuing and Professional Studies. So yeah, we're really excited to have him on. As an ad fraud investigator, he's done a ton of work with the government, regulatory bodies, he's been a consultant to clients, helping them to...

mitigate cybersecurity risks, ad fraud risks to really improve the effectiveness of their marketing campaigns. So I found the conversation deeply interesting. I know Rio did as well. ⁓ Sometimes controversial, but it's topics that I think we need to breach to really understand how we make the digital media ecosystem as effective and efficient ⁓ and safe, brand safe as possible.

And I think he uncovers a lot of the challenges that we still continue as an industry to see in those areas. So Rio, you had a couple of quick news ⁓ items to cover. Obviously, Cannes is coming up, and I know you're going to the South of France shortly. Plus, introduce the theme of the show and some of the key takeaways that we had.

Rio (02:03)
You got it, Brett. So I completely concur. Great conversation. Lots of meaty discussion about ad fraud, about some of his conclusions that, as you mentioned, some of it is controversial. We were joking in the episode that we're maybe expecting to get some hate mail. So we'll see. Hopefully not. But yeah, to your point, everyone's getting ready for a Cannes and because of that kind of a slower news cycle the past couple of weeks.

I leave tomorrow night, family goes to Turkey and Greece for a couple of weeks, and I actually fly straight to Nice and right off to Cannes.

So I'm looking forward to it. It's going to be great. already have a pretty packed agenda. Lots of really good stuff going on. So it's going to be fun. But I guess before we dive in, the one announcement I did hear, Brett, that I thought deserved some attention was

actually by Meta, you remember a couple of weeks ago, mean, Zuck's been saying a lot of interesting things about just give people a button they can push, tell us what outcomes you want, we'll do it for you. I mean, taking AI, applying it to the entire value chain and advertising, right? Everything from ad ops to trafficking to optimization, right? To audience development and planning, everything, right? So that's kind of his vision. He's not been shy about saying that. So I thought it was interesting how he really announced that

they plan to fully automate ad creation and targeting using AI by the end of 2026, so very aggressive timeline, right? Some of these things are just starting to roll out, right? So what this really means when you cut down to it is advertisers in theory should be able to just input a product image and budget and then let AI rip and it'll create the entire end-to-end campaign, including text images, videos, ⁓ audience selection, targeting, everything, right? So kind of wild, that'd be across Facebook and Instagram. ⁓

and I guess WhatsApp too. So pretty incredible. That's an aggressive timeline. I mean, we talked about some of the agency layoffs last time. There's gotta be some kind of crossover there, right? As a lot of these jobs are automated, does it break the challenge or break the FTE model? ⁓ What do some of these look like in the end? It's interesting, interesting discussions and Meta continues to make investments and make some pretty bold statements in this area. So I'm excited to watch that.

Brett (04:13)
Yeah.

Yeah, for sure. And it certainly, it certainly follows nicely after our conversation with Ben Wild at Georgian on AI and its implications for both the workforce, but automation and what's realistic with today's sort of ⁓ data management ecosystem, tech ecosystem and what that's going to look like in the coming years. So to see Zuckerberg come out with, and you and I have talked about that a couple of times in the pod, about ⁓ creative being one of those first strike areas where you can automate and version

and personalize, ⁓ you know, exponentially, ⁓ just at a high volume without the need for a small army of people to do it. It'll be interesting to see the quality that comes out of this, And some of the, there's obviously gonna be, there's some curation involved and management and control and obviously the planning and strategy that goes into these AI systems that develop out at creative has to be sound. So you're not eliminating all positions, but you might in a sense,

up leveling the core function that is managing this process and all the stuff that's more commodity work, right, ⁓ is the stuff that AI is going to be doing and automating, right?

Rio (05:25)
I guess my advice to young people

would be get familiar with these tools, learn them. You want to be the person using them, not the person who's, I guess, being victimized or made redundant by them. That would be my advice. I guess getting back to the episode, as you mentioned, important topic. It's not a new topic. I mean, we've been talking about digital ad fraud, right? For as long as there's been digital advertising, right? So going back 25 years.

Brett (05:37)
Yeah, yeah, totally.

Rio (05:52)
We thought this would be a great topic because there's been some interesting reporting recently on, including that kind of explosive piece in Wall Street Journal, right, where they claimed up to of web traffic is, quote unquote, fake users, which is a high number, certainly the higher end of what I've been seeing. And actually, Dr. Fou has some even higher estimates, which may be higher than most people are comfortable. And you could argue that maybe he's looking at specific use cases, not generally, but it's a good, it's a great discussion.

Brett (06:18)
Yes, some of his estimates always make me think He's got a 99% number that he throws out there So we lean in because as we talked to him we realized that's not really what he was claiming he was trying to make a kind of overarching point

Rio (06:18)
But look, look. Yeah.

Yeah, I

Yep.

Yeah, I he did kind of say it but he does have data to back it up but mean the counter argument or kind of factual to that is okay was he maybe looking at ...I mean... think was actually Adam Heimlich from Chalice was saying like some of food studies are looking at bot honey traps and maybe not representative of the general but that said he has the data to back it up, really smart guy, incredible CV so think everyone's gonna really enjoy it.

Brett (06:57)
he's over waiting the bot traffic is what that perspective seems to mention.

Rio (07:02)
I mean, in

theory, it depends on what you're sampling, right? If you're sampling pot honey traps, sure, but he's been saying this for a long time. And I think you look at over the years, a lot of the stuff he has been saying has been proven out, right? The number of, the amount of IVT or NHT, I mean, it's a lot of it, right? And it has an impact and money's being siphoned off. So I think we just have a really good productive discussion about what's behind it, what does it look like, what can we do to solve it? So I hope everyone, I'm sure everyone will enjoy it.

Brett (07:06)
Yeah.

Rio (07:31)
Let us know in your comments what you think.

Brett (07:34)
Yeah, for sure. And I think there's a lot of conversation about incentives and across the ecosystem, from the brands to the agencies that are doing the media planning and sometimes buying to the DSPs and ad tech middlemen that are taking a tax for every ad impression served and every ad served. But yeah, fascinating conversation, definitely tune in

Rio (07:54)
Enjoy.

Brett (07:56)
Dr. Fou a pleasure to have you on Signal & Noise. Welcome to the podcast.

Augustine Fou (08:00)
Thank you, Brett. Thank you, Rio. Glad to be here with you.

Brett (08:03)
And as we customarily do on Signal & Noise for those that have been listening out there, Rio is going to walk us through a little bit of what we're going to be talking about with Dr. Fou today, some very interesting ad fraud topics. And we've got a little thesis prepared to kind of set the stage for the conversation going forward. And then we'll go into a little bit of Dr. Fou's background and where he comes from, and then we'll get into it. So Rio, why don't you take us away?

Rio (08:31)
Thanks, Brett. Dr. Fou, excited to have you on this. This has been a long time coming. It's going be really fun discussion. So, it's the preface here and the reason why we call this episode, Are the Bad Guys Winning? Look, I spent a better part of two decades working in digital media, untangling. It's almost like a Rube Goldberg machine, right, that really is modern advertising and ad tech. And today we're going to be...

Brett (08:55)
I remember those cartoons.

Rio (08:57)
Yeah, yeah, those are wild. I mean, no one would have really built the system the way it is. It just kind of happened to be created this way, right? With all the different components and players. And today we're really to be looking at a problem that quietly siphons off quite literally billions of dollars a year and erodes trust every single day. And this is digital ad fraud. This is not a theoretical question, nor a hyperbolic one. For many brands,

it actually does feel, and I've seen this, doing media buys, it definitely sometimes feels like the bad guys are winning. You have bot farms masquerading as premium publishers. You have MFA sites boosting metrics. You have got what are supposedly safe channels, like CTV suddenly showing cracks. And let's face it, there's a lot of pervasive illicit activity there.

I think the problem is no one really agrees on exactly how much fraud is out there. Everyone does agree there's a lot.

I've seen some estimates recently in the billions of dollars. I think Wall Street Journal published an article a few weeks ago and they said it was 40 % of web traffic is made up of fake users, which is just wild to think about. But for anyone who's been working in this industry a long time, not too crazy, right? So the stakes are high and this is very important topic. And that's why we invited Dr. Augustine Fou, one of the industry's most relentless fraud hunters to help us cut through the noise.

So the next hour we're going to confront the uncomfortable truths, take a look at the latest tactics, and I think most importantly, we're going to attempt to outline some concrete steps that you can take before your next media plan goes live. So let's roll up our sleeves and see who's really winning, the bad guys or not.

Brett (10:44)
All right, well, Dr. Fou, again, pleasure to have you. You know, I know of you as a pretty well-known digital marketer, marketing executive, an ad fraud investigator, the CEO of Fou Analytics, which you honestly have the best tagline.

ever "see Fou yourself" with better analytics. I'm like great play on the name with my last name. I certainly have heard that more times than I can count with the last name of House But I saw you received, you you received a PhD at MIT for those of the folks that don't know Dr. Fou, the age of 23. Can't tell you what I was doing at 23, but certainly not that. You're a chief marketing science officer at the Advertising Research Foundation, a group CDO at Omnicom.

Rio (11:20)
Not getting a PhD.

Brett (11:29)
among a bunch of other pretty high profile roles. Why don't you tell the audience a little bit about yourself?

Augustine Fou (11:36)
I think ⁓ the way I would look back at my career, it's really, ⁓ I would call it science-driven marketing. So when I first started out, ⁓ my company is called Marketing Science Consulting Group. And the idea behind it was, how do we use data and analytics to actually see what's working in marketing? Marketing and advertising, there's a lot of creativity, which is great. And we all know the TV show Mad Men, where the creatives were

were the gods in the agencies. And that's all great, but as a scientist, I want to see how do I reproducibly do marketing well and how do I actually gauge its effectiveness. So it's really think of it as a scientific approach to ⁓ advertising and marketing. And I kind of fell into it because my PhD at MIT was in chemical engineering, material science. And you might have seen some of the Sony TVs or LG TVs. They have

O-LED, right, organic LED materials, right? So that's the material that I was working on in the lab 30, 35, 40 years ago, and it's finally made its way into consumer products. So it has nothing to do with what I'm doing today, but it's more the scientific method. It's running the experiments and seeing if I can prove things out ⁓ using my own data, my own experiments.

So it's really kind of translated into the digital marketing space. And that happens to be a perfect fit because digital is really the intersection between marketing and advertising, which is very creative and fungible and not concrete and data where we can actually measure a whole bunch of stuff. ⁓ Yeah, so.

Brett (13:14)
The art in the science, right?

Augustine Fou (13:16)
I kind of fell into it. ⁓ I left McKinsey in '95, ⁓ early '96 and fell into what is now known as Silicon Alley, which is New York's version of Silicon Valley. And a lot of it was driven by advertising because all the big advertising agencies were here. So New York has always been more advertising focused. Silicon Valley has been more tech focused. But in Silicon Alley, we started the mid 90s just building websites for people.

And back then, we still had to convince clients they needed a website because this was so new to them, they didn't really want it.

Brett (13:53)
Oh, you're making, you're just reminding

me of Rio and I kind of started there as well at about 1999. It was exactly that.

Augustine Fou (13:57)
Yeah.

Rio (13:58)
Yeah, we're trying to sell websites

to tile dealers at one

Augustine Fou (14:03)
Yeah, yeah. You know, in some agencies, you know, that was the first wave of the internet, right? They did very well selling a whole bunch of websites and hiring a whole bunch of people to design them and code them and stuff like that. I remember our first client was Victoria's Secret, and it took literally a year and a half to convince them to put up a website because they were deathly afraid of cannibalizing their existing channels, which were catalog and stores.

Brett (14:04)
The home improvement industry.

Augustine Fou (14:31)
But ultimately, we convinced them to do it and built the very first version. And then, as a lot of clients do, they quickly take that back in-house and do it themselves. But it was a great experience. And fast forward now 30 years, I've been doing digital marketing in various forms for that long. And a lot of it was hands-on keyboard, where I'm the one setting up the website. I'm the one publishing the WordPress blog when that first came along.

I'm the one looking at Google Analytics. I'm the one running LinkedIn campaigns, YouTube campaigns. So it's really from hands-on experience that I speak. So a lot of times what I'm finding these days is that there's a lot of marketers that are very new. And in fact, some of them weren't even born when I first started ⁓ doing digital marketing. So anyway, we're all old timers here. But the idea is

Brett (15:17)
Don't remind us.

Rio (15:19)
When you

start hiring people who were born

Augustine Fou (15:23)
There's still a lot of education that needs to happen and you hopefully more and more marketers are open to new ideas, new ways of thinking about it because a lot of times the sales people that are coming in and selling the media have an incentive to just sell them the media. You know, they don't really have an incentive to tell them half of it or more is just not real. I don't even blame those people because I would say that they literally don't know and even the people at the legacy fraud verification companies.

They don't know how the tech works. They don't know whether the tech catches everything. That's really not their expertise. They're a salesperson. They're not an engineer or a JavaScript person, so they don't understand. So the whole point is that, getting back to what you said earlier, fraud is very big. And no one really knows what it is, even if it's Wall Street Journal that publishes a number or eMarketer or any of these accounting firms. They don't actually know, because they didn't measure it themselves.

So everything you see out there is an approximation. And over the past 15 years studying the fraud problem, no matter how big the number, it really doesn't scare people into action because they just think it's someone else's problem. Oh yeah, well, when we look at our own campaigns, like our fraud vendor tells us it's 1%. So it's not affecting our stuff. It must be someone else's problem. those big numbers, unfortunately, don't actually do anything.

So I've kind of in recent years kind of resigned myself to helping those who want to be helped. And I actually think that finally we're seeing a turnaround, right? So for many, many years, there's a lot of big advertisers that didn't want to hear about it, right? They just wanted to buy more reach, more frequency, reach more people. yeah, our campaign, you know, spent 10 billion ad impressions per month or bought 10 billion ad impressions per month. And we spent this much money.

Some of the bigger players, I won't even name names, they go on the ANA stage, their badge of honors, how much money they spent. Shouldn't it be how much money you made from your digital marketing, not your ad budget? Yeah, so I think that kind of sets the stage for what we're going to get into today. But I mean, that's what I've witnessed in terms of the fraud in the last 15 years and the evolution of digital marketing over the last 30.

Brett (17:30)
Yeah, how many outcomes did you actually deliver as a result of that spend?

Rio (17:31)
Yeah. Yeah.

Yeah, it's interesting. mean, I kind of fear of finding out "fofo" right? I think there's a lot of that. ⁓ Marketers just don't want to know. I mean, I definitely see that. And then looking at the fraud itself, you mentioned the 1% study by the ANA. I remember you just eviscerating them online. That was actually kind fun to watch. think I actually, I did throw in some comments there too.

Augustine Fou (17:51)
Yeah.

Yeah, I've kind of made up with

the ANA recently, but I mean pointed out to them that they're citing numbers from the legacy vendors and the 1% is not all the fraud there is. The 1% is all the fraud they could catch. So that's a big difference, right? And so.

Brett (18:17)
Yeah. Didn't

they in 2023 put out a report? It was that sort of ad fraud, sort of wasted inventory report. Well, I read that they were saying ad fraud and sort of wasted inefficient media is equal to about 100 million. Or was it 100 billion? Yeah, that was... Yeah, yeah, yeah.

Rio (18:25)
think that was the 1 % one, right, in 2023.

Augustine Fou (18:35)
Well, that was kind of when I made up, ⁓ it was like $100 billion. But ⁓

what they were talking about is the MFA sites. So that's yet another ad tech acronym. It stands for Made for Advertising Sites. But it's kind of a not good acronym because all sites are made for advertising. That's their primary revenue model. What I see that is ⁓ kind of just ⁓ another ad tech acronym that kind of covers over their problem.

Brett (18:51)
Yeah.

Augustine Fou (19:05)
without talking about it directly. Call it fraud. Don't call it MFA or IVT or NHT or any of these other three-letter acronyms. IVT is invalid traffic. NHT is non-human traffic. MFA is just the latest version called Made for Advertising sites. Now, there are some sites that are completely fake. Over the past 15 years of studying the problem of fraud, back when WordPress templates made it easy for people to create websites,

the bad guys were creating 10,000 WordPress sites per day, just using WordPress templates and using 100 % plagiarized content. They just had bots. Yeah, they don't need to plagiarize it anymore. They're just going to use AI to generate it. But this has been a problem that's happened for 15 years. And only now, in 2023, the ANA said, yeah, we think the problem is a little bit bigger than the 1%.

Rio (19:46)
Well, I bet with AI, that's going to infinitely worse,

Brett (20:03)
Yeah, it's more like 25%. Yeah. Yeah.

Augustine Fou (20:03)
And it's called image. Yeah, something like that.

And even that's severely underestimated. So the article I put up yesterday was basically talking about the veil of verification. So the verification companies that advertisers have relied on for the past 15 years basically have said fraud is 1 % for most of that time period. And that essentially covered up the fraud.

If a bad guy were making 10,000 fake WordPress sites per day and putting them into the ad exchanges and they were all marked as 1 % fraud, the advertisers would not know that there's a problem. So that's been the problem, right? And it's further amplified because the trade groups like the ANA ⁓ kept putting out press releases for the last eight years saying the problem was 1%. So in essence,

Rio (20:53)
Well, how much

is it? I've done media audits for clients from time to time and I've never found frankly, like in terms of like non-human traffic invalid traffic, never found and not even talking about just missed optimization opportunities and just budget pissed away on random things. mean, it's rare offend less than 20, 25 % on these. Like how much is there?

Augustine Fou (21:14)
It depends on the context. So when I first look at campaigns for the largest of advertisers, it's 99%. The reason it's that bad is because the biggest of advertisers have too much budget they need to spend. So then they assign or task their agency, their media agency, to go buy too much inventory that doesn't actually exist. So that creates the incentive

for the bad guys and everyone in between to just manufacture as much ad inventory as needed to absorb all that budget. If we take a step back and go back to the beginning of digital advertising, the ads are supposed to load when someone actually visits a website and loads a web page. When the page loads, the ads are supposed to load, so thats.. the inventory doesn't exist before that. It is created when a person visits a website.

That's the way it's supposed to work. But now we have it completely backwards. The biggest advertiser, again, I won't name names, they say, yeah, we've got $2 billion we have to spend in one year in digital. Yeah, use it or lose it. ⁓ So go buy me. Go help me spend it all. So then it's like, OK, well, they go to one exchange. How much inventory can you sell me? OK, it's this much. They go to another exchange. How much inventory can you sell me? They go to every exchange. And they run out of stuff to buy.

Rio (22:25)
Just use it or lose it, right?

Augustine Fou (22:43)
So who else can give me more inventory? this is exchange I've never heard of before over here. They have all magically another 10 billion impressions for me to buy. Let's go buy. And on top of that, the agencies have to spend it all so they can earn their cut of it. So they're also incentivized to spend it all. So long story short is the amount of impressions that are being bought and sold are completely out of whack with reality.

Rio (22:54)
Totally legit, right?

Brett (22:57)
bit.

Yeah.

Augustine Fou (23:12)
and has been for the past 15 years. And the departure point, the point in time where we went wrong is the advent of programmatic advertising. So when programmatic became the dominant form of media buying, this was like 2013, 2015 timeframes, so about 10 to 15 years ago, that's when the fraud exploded. And that's because we lost sight of who the heck we were buying the ads from.

Brett (23:15)
Yep.

Augustine Fou (23:41)
Prior to Programmatic, prior to this whole bidding mechanism, we had to sit across the table from ESPN and negotiate a deal. I would need to buy ads from you, or I would buy ads from Conde Nast, from Hearst, from Meredith, from legitimate publishers. You could see the people that work there. Once we started allocating tens of millions of dollars to an exchange and said, OK, well, we don't know ahead of time where the heck these are going. We just need you to spend the $10 million for us. That's when everything went wrong.

Brett (23:57)
Yep, direct buys, yep.

Augustine Fou (24:10)
And so if you remember, AppNexus was one of the first exchanges. They would take the money, and they would supposedly bid it out to all these different places, and then the ads would get displayed. In 2016, there was an ad exchange article that was published to show a chart that says, AppNexus purged their own exchange. They had 260 billion impressions per month. They purged 92% of it.

down to $20 billion per month. So fraud has been a 92% problem. It's a greater than 90% problem since at least 2015. But again, no one else has done that after they did that. Everything else is manufactured. And AppNexus didn't do it again. But that was a data point that we had back in 2015 to say that ad fraud was already a 90-plus percent problem back then. It's only gotten worse from there.

Rio (25:08)
It's interesting you say about the incentives, right? I mean, because you mentioned the agency is taking their commission on it, but then you look at the entire tech tax DSPs, SSPs, ad verification, they're all taking a cut of it. It's based on the overall spend. It's not necessarily based on an outcome or anything like that. So, I mean, you could even argue that it's the incentive is a perverse disincentive to spend less and incentive to spend more. And even if there's ad fraud, everyone's still winning. Maybe the brand's not, but then if the brand doesn't care, I mean, like who's really accountable? I guess that's

Brett (25:09)
Yeah, and it seems...

Augustine Fou (25:23)
Correct.

Correct.

Rio (25:38)
Like, who should be accountable?

Augustine Fou (25:39)
Yeah, and so I kind

it kind of gets down to the way I summarize it is that ad fraud is an incentives problem. It's not a tech problem. And for years we have thrown more tech at it, right? So there's all these fraud verification companies, whether it's the two big ones or others that have sprung up to help police the problem, right? My tech included. But it's an incentives problem because a lot of people, like you said, don't want it solved. So even if the exchange or the DSP is not the one making the bots,

they're certainly looking the other way because the more volume that flows through their pipes, the more money they make. So if the fraud verification companies were saying, yeah, everything's 1 % fraud, don't worry about it. They're going to be so happy. They love those numbers because then they can continue trading. They can continue making money off of what essentially is made up stuff. Yeah. Yeah.

Brett (26:27)
It's a toll. Yeah, it's a toll. It's like managing a superhighway, right? ⁓

Yeah. And it's interesting that some of the reports that have to come out to surface some of these problems, like the analytics reports ⁓ right now at the Seattle International Film Festival, this festival. know you know about this, Dr. Fou, is there's a there's a ⁓ new movie called Unclickable, ⁓ An Ad Fraud Operation by ⁓ excuse me if I mispronounce the name, Babis Makridis and Guy Krief and

Augustine Fou (26:45)
Yeah. Yep.

Brett (26:56)
In order to really show how easy it is to exploit, how easy ad fraud it is to exploit, these two execs gathered a team of software developers and created a digital ad fraud operation right before our eyes in making clickbait websites, harvesting bots that wouldn't get flagged as bots, just to show how extensive the problem was. So they basically recreated the problem. And there was something in the trailer that I watched before the episode here.

⁓ where they really kind of knocked me off my seat was if you're asking how big ad fraud is, it's like asking how big the illegal drug market is in the United States, right? It's almost so large that it's hard to put a finger on.

Augustine Fou (27:31)
Yeah. Yeah.

And it's hard to believe, right? The number is so large that most people's default reaction would be to not believe it. Okay. Because that would wreck their worldview, right? That would wreck the entire premise. ⁓ yeah. Yeah. Yeah, exactly. So, you know, that would, that's a great example, right? Unclickable is the name of the documentary. You know, it's, it's been years and years in the making, right? It's finally come, come to ⁓ the film festival and all that. And it's great. And another reporter, Megan Graham.

Brett (27:47)
Yeah, that can't be true. That poss- it can't possibly be true, right?

Augustine Fou (28:05)
who used to write for CNBC ran her own experiment. This was in 2020 because she found that all of her articles were being plagiarized and used on fake websites to make money off of her articles. So she was pissed and she said, okay, well, I want to find out how easy it is for bad guys to basically set up a fake website and make money. So she did exactly that. She registered a domain, plagiarized 10 of her own articles just to have some pages, content on the pages.

and then applied to eight exchanges to see who would let her in. seven of the eight exchanges let her in, and she started making money the next week, the following week, literally. There was one exchange that didn't let her in. And can you guess which one that was?

Brett (28:51)
the Trade Desk.

Augustine Fou (28:52)
No, it's Google. So what Google did was basically they said, well, all of this is duplicate content. Because Google obviously has a search crawler. So they crawled it and said, OK, these are all exact identical articles to what we see on CNBC.com. So they didn't let her in. But seven of the eight let her in. And she started making money. And she showed ads from the biggest advertisers running on her completely fake site that anyone could have seen if they literally checked when that domain was registered.

Brett (28:55)
That was Google.

Yeah.

Augustine Fou (29:21)
You could say, OK, there's no possibility that this site or this domain that's 10 days old would be running ads. So anyway, so that's just yet another example that corroborates the documentary that you just mentioned, Unclickable. And like I said earlier, the bad guys are setting up tens of thousands of fake sites just using WordPress templates. Just populating it with whatever content they have doesn't actually matter.

And so they had a constant stream of these domains they could rotate through to just multiply their own revenue. So the whole unclickable documentary said, OK, they just ran five websites. What if they made 500? What if they made 5,000? They could literally multiply, scale their own revenue at will. It was that easy.

Brett (30:03)
start to scale it.

Yeah, yeah,

and just beyond the problem of dollars flowing away from legitimate news sites. I mean, that's a threat to more than just the advertising industry. That's a threat to, yeah, to publishing in general and our ability to, you know, information versus disinformation, right? Because these.

Rio (30:17)
It impacts publishing, yeah.

Augustine Fou (30:24)
Yeah,

let me add some stats on that. So Pointer is a nonprofit that tracks local news. So Pointer, I think a few years ago, published a study that said 1,500 local newspapers have died and gone away. So there's entire communities that no longer have a local newspaper because they were ad supported. When they transitioned from paper to their websites, they relied on ad revenue to make money.

But those are small because of their nature, right? They're just local to a particular town or city. They just couldn't make enough money to survive, so they've gone away. So like into that vacuum now, right? So there's a vacuum of local news into that vacuum. We now have entirely fake sites that have entirely fake content that's mostly fake news or deliberate disinformation making money on ads. And for them, because they don't actually have journalists or writers or people writing the content,

Even if it's a 5 cent CPM, 10 cent CPM, a dollar CPM, it's still profitable for the fakes. It's all plagiarized. So now the fake sites and disinformation sites have a new source of funding that they never had before. And so therefore, they could expand their disinformation operations. Once that was proven, now we see an influx of foreign actors, nation states.

Rio (31:25)
costs are so low, right? Yeah.

Brett (31:26)
Yeah,

it's gravy. Yeah.

Augustine Fou (31:47)
you know, countries intent on sowing disinformation, getting in on the game, because there's so much money, there's billions of dollars to be

And expand their operations. So that now spirals into what we're seeing now, which is a society that's most at odds with each other, the most divisive ever. We see so many people just not having access to real news or even fair and balanced news. It gets more and more polarized to a point where they're kind of incentivized to action. So now we actually see some of this harm

spill over into like real world harm to society. Right? And that's actually the outcome of the bad stuff that happened over the last 15 years.

Rio (32:30)
of this is,

Dr. Fou how much is being driven by the, you know, let's say nation states and like criminal cartels versus just random fraudsters? And how much of this is organized crime

Augustine Fou (32:41)
I think a larger percentage, obviously I don't have real numbers on that, but it's so lucrative because there's so many dollars. The pot is refilled every single year by the biggest of advertisers. And I think the IAB report for 2024 that just came out was like 250 something billion spending just the US on digital advertising, but most of which is going through digital and programmatic channels. So every year there's like $260 billion pot of gold just sitting right there by the...

highway, right? And there's no, not even a cover on that. You just go take the goal.

Rio (33:14)
Well, you

think about it, and a lot of people are making cut out of it, I mean, not necessarily bad actors, but all of these, you the exchanges are, certainly the DSPs, the SSPs are.

Augustine Fou (33:21)
Correct. So like Google being the largest,

they're making a ton of money on this. So I think that they intend to do the right thing, but they're not going to proactively do the right thing, which means there's so many of these apps that you see in the Google Play Store. Some of it, if you just literally look, the developer address is some letters and numbers at gmail.com. That's clearly something wrong with that. And then if you look up the mailing address, it's literally a mailbox somewhere or

you know, a Taco Bell restaurant address or something. So like some of these are so obvious that if you just look, you would know. And they're all ad supported.

Brett (33:57)
Yeah.

And the Wall Street Journal is covering this now, too. Right. I they just published an article, right. The Meta Battles an 'Epidemic of Scams' as criminal as criminals flood Instagram and Facebook, which was really interesting because it speaks to almost exactly what you just said. There is this wholesale company, Half-off Wholesale that doesn't sell direct to customers. ⁓ so they don't sell online.

Augustine Fou (34:06)
Yeah, yeah.

Brett (34:19)
people were impersonating them and they found that was largely the research was done in Australia and I think New Zealand mostly and people were impersonating them. There were 4,400 ads over like a year period for this particular wholesaler selling direct to consumers, promising things like pallets of, you know, Amazon return tools, like tools that were bought on Amazon that were returned, right?

And yet 15, I think of the 4,400 were actually wholesale ads by the brand itself. So everything was just this fraudulent activity to get people to send money in for this unbelievable deal to get merchandise. And then they would never receive it. And the return address was this Half-Off Wholesale.

Augustine Fou (35:08)
Yeah,

so let me differentiate the scam that you just talked about versus the ad fraud stuff because a little bit different right the stand that you just talked about targets real humans real people and there's a variation on that... That's just yeah, that's literally just regular fraud because what what I've seen happen over the years is that on Instagram as well. You see all these websites that promote you 90 % off your you know tennis shoes or

Brett (35:16)
Okay.

Rio (35:21)
That's just regular fraud, right?

Brett (35:23)
Yeah.

Augustine Fou (35:36)
or just clothing, whatever. And when you click on it and you buy something at 90 % off, you're saying, yeah, we got a great deal. But the product never arrives, and you've just given them your real credit card number and your real home address. So those were what I call credit card harvesting operations, where they never had an intention to actually mail you anything. They were just tricking people and using their greed instinct to say, yeah, I'm going to get 90 % off.

try to buy this. anyway, those are scams that affect real people. Ad fraud doesn't involve real people, right? It's basically websites that have no human visitors, so they basically have to buy all their traffic. And therefore, 100 % of the traffic is bot traffic, even if the legacy fraud detection companies couldn't detect any of it as bots.

Rio (36:26)
It was interesting, Dr. Fou I actually was doing CPA buys for a client. Funny I was tweeting about this this morning. the client wanted CPA. And we did some incentive deals with a bunch of these networks. And there was credit card fraud. People were stealing credit cards, placing fake orders, taking their money. And then the affiliate would do this, pump through 1,000 orders, and just close right away. And another one would appear. Sometimes I think the two can actually mingle together, which is even more messed up. Yeah.

Augustine Fou (36:52)
Yeah, exactly. Yes.

Brett (36:53)
Yeah,

it's fraud is an overarching concept that has real world implications and then to your point, just these bot-based sites.

Augustine Fou (37:01)
Yeah, some of it's literally

just bots and whatever. I was just, ⁓ one of my tweets from today talked about how you don't actually need to run websites anymore. You don't need to run ads anymore because just by faking the bid request, you can already be successful and make money. Okay, so let me reiterate. know, 15 years ago, bad guys had to set up WordPress sites and then put ad units on it and then buy bot traffic to load the pages, right? That's how they generated inventory to sell. Then they

became more clever, they said, why waste the bandwidth on loading web pages for the bots to load web pages? Why not just do naked ad calls, which means just load the ad impression? Because we can just save the bandwidth without loading the page. Now, especially in CTV, we're seeing tens of billions of fake CTV bid requests per day, just manufactured. Use a Python script. ⁓ Any kind of code on the server can assemble a fake bid request,

send that into the endpoints, there's not enough checks and balances. A lot of the fraud detection will just miss that that's completely fake and let it through. So then they can already make money. There's no ad that ever ran on a CTV. The apps don't ever exist. Correct.

Brett (38:13)
Yeah, that's an interesting point. So the ad never actually shows. They're just

making the call, you know, getting the cash register to go, Ching, right? And then in making money off of that. That's really interesting.

Augustine Fou (38:21)
Yes. Yes.

Rio (38:23)
Well, wasn't CTV

supposed to be like a cleaner, quote, I'm putting, quotes here, cleaner channel, right? Where, I mean, that was one of the pitches, Is it or is it not?

Augustine Fou (38:28)
It is. is. me, yeah,

here's a conundrum. So let me give you two statements. There's no fraud in CTV because it's impossible. Okay, that's one statement and that's true. And the other statement is there's 100 % fraud in CTV. That's also a true statement. How can both of those statements.

Rio (38:50)
Is it a problem with

the type of CTV inventory? in-stream versus out-stream, I've seen a lot of mislabeling.

Augustine Fou (38:54)
Yeah, you got it. You're just too smart. Yeah, real

you're too smart So yes, so there's no fraud in CTV because it's impossible that applies to when you're buying from a Home and Garden TV, a History channel, an A&E, a Disney whatever okay that actually sells the inventory team because Disney or Amazon Prime, they're not actually trying to rip you off okay, so when you buy inventory from them, you can actually see your ad show up on a real TV somewhere in someone's living room Okay, those are real and there's

Rio (39:09)
in Amazon Prime, something like that, right, yeah.

Augustine Fou (39:24)
fraud in CTV is impossible. But when you're buying CTV through programmatic exchanges, whatever DSP you happen to be using, then that creates the possibility of the bad guys basically manufacturing bid requests that pretend to be ESPN, that pretend to be Roku, that pretend to be anything they want. And no one has the ability to check that because we can't run JavaScript. We can't even verify that the ad ran anywhere.

So this really boils down to, this is like the most extreme case of incentives. If the advertiser doesn't care and they just want to buy a whole bunch of low cost CTV, that's the stuff they'll buy. It doesn't actually exist. The ad never ran. And then if on the flip side you look at ESPN, you'll notice periodically there'll be the red logo on a black background and elevator music. That's an unsold ad slot.

the mainstream CTV sellers have unsold inventory. They didn't sell out the stuff that they have. Whereas the big advertisers are buying tens of billions of CTV impressions from somewhere else. It's not real.

Brett (40:36)
Yep.

Yeah. And we've estimated at MediaRadar to be 35 billion across the AVOD channels, which are like streaming catalog content, live streaming, AVOD like Tubi or Pluto, VMB, MVPDs. That's a big, a big market. And you're saying that there's a danger that it falls into a similar trap that programmatic advertising fell into?

Augustine Fou (40:47)
Yeah. Yeah.

Rio (40:55)
So a

lot of it's basically crummy, outstream, bogus. It doesn't even get served up, right? I'm assuming there's no device ID. Obviously, there's no cookies in CTV. it shouldn't be that hard to figure this out, right?

Brett (41:05)
There's an IP. ⁓

Augustine Fou (41:08)
Yeah, it shouldn't be. In fact, two teams just notified me within the last two weeks saying that interns found the problem. They could see a particular device ID or a particular IP address generate tens of thousands of bid requests, CTV bid requests per hour. OK, that doesn't even make common sense, right? When you stream it, exactly, it is so obvious that you could just block it yourself if you wanted to. You don't need any.

Rio (41:29)
That should just be blocked immediately, right? I mean, like, why would you even accept that?

Augustine Fou (41:37)
fraud detection company, you don't need my tech, just look at the damn numbers and look at the bid requests coming in and you'll be able to tell.

Rio (41:43)
You don't even need

AI, you can just have an intern looking every day at all the logs, right? And you say, wow, this one... I mean, you really... You can argue that's just criminal malfeasance, right? The fact that no one cares, they're just letting it rip. ⁓

Augustine Fou (41:45)
S. Yeah.

It is so obvious and so large.

Correct. Correct.

Brett (41:56)
Yeah.

Augustine Fou (41:56)
Yeah.

Brett (41:56)
And you don't need a data analytics or data scientist to figure this out. Right.

Augustine Fou (41:59)
No. I mean, literally, in an Excel spreadsheet, can say, OK, well, this device is generating 10,000 bid requests per minute. OK, and that's not even possible for a real device.

Brett (42:08)
Yeah. Well, and has a ton of downward pressure on performance results. So when your data science or data analytics team reports up to your CFO or your agency reports to a brand, I mean, how do they explain all of this traffic that's not performing?

Rio (42:24)
Well, Brett, think it's

the opposite. think they're saying, look, we spent all the money. We got all these great impressions. Put me on stage, again . It's the opposite.

Brett (42:30)
So

they're just putting reach metrics out there? They're not talking about outcomes and engagement and...

Augustine Fou (42:34)
Yeah, yeah,

part of the problem is that they've been using, you know, reach clicks, you know, traffic to the website and all that. All of those are just fungible metrics that bots can easily fake. But Brett, you bring up another important point here, and I think it's important for us to cover it on this podcast. It's the problem of attribution. So there are some advertisers that do programmatic and the ROAS that they're seeing in the reports that they're given are so awesome.

They're saying, wow, I need to take all my money from every other channel and put it into digital because the ROAS here is 10 times better than any other form of advertising. You might be able to guess where I'm going with this, which is the reporting, the attribution reporting is basically taking credit for sales that had already occurred. The digital ads did not cause those sales. It's simply a figment of the way the attribution is done.

So a very easy way to understand this is the concept of view through conversions. So many years ago on Google's platform DoubleClick, they said, oh, it's so unfair that search gets all the credit for the sale. Display ads should get some credit because they must have helped cause the sale in some way. yeah, awareness and whatever. And people may not click on it, but later they're going to come buy it. So display ads should get some credit.

Brett (43:45)
Yeah, the last click attribution problem, yep.

Brand awareness, right?

Augustine Fou (44:02)
So then they made up this thing called view through conversions, which said, ⁓ well, if a device were exposed to the ad, we will count a window of 30, 60, or 90 days. If any sales occurred in that window, the view through conversion window, we're going to attribute it back to the display ad. Now, obviously, us on this call will be saying, OK, well, you can't attribute a sale to a single display ad, because one single display ad did not cause that sale. But yet, that's how the whole model was built.

Then exactly, right? And Google was the one selling that. So they also pushed this concept of view through conversions. So then a lot of people

Rio (44:32)
Well, you can say that if you're the one telling your display out. You get it, you will, right?

Brett (44:35)
Yeah. Yeah.

Which

doesn't seem like it's any better than last click attribution. And I came from the MTA attribution world and there are so many exogenous factors, ⁓ exposure factors, engagement factors that drive. And what we found is that more often than not, you're over-crediting a channel, you're over-crediting it ⁓ always.

Augustine Fou (44:44)
It isn't, it's just another way of shifting.

Yeah.

Yes, always, always over-crediting.

Because there's no exact math to that. There's no exact science to that. So you're always over-crediting. So then what's happening is that now we have ⁓ this attribution model that's over-crediting display ads or other types of advertising, ⁓ the sales to those forms of ads. Now on top of that, once you have this model that

⁓ the bad guys get involved. So what they're now doing, and here's an example. Let me use this story. In 2016 or so, there were companies that were called location-based companies, platforms. They would say, ⁓ you don't have to pay for the ad impression. You don't even have to pay for the click. You only have to pay us when we get someone to walk into a Dunkin' Donuts for you.

Brett (45:51)
with that

dirty GPS data, right? There's a lot of mobile GPS data players in the ad tech ecosystem, right?

Augustine Fou (45:57)
Yeah,

I'm not going to name names, but it's those location-based attribution companies. So long story short is, there were too few of the people who actually saw the ad and then walked into a Dunkin' Donuts. So what they ended up doing was cheating. So the form of fraud that has to do with attribution is where they would literally buy one cent CPM ads.

Brett (46:02)
Yep.

Augustine Fou (46:21)
and cause all 380 million devices in the US, literally every single mobile device, to be marked as exposed. So when any one of those mobile phones walks into a Dunkin' Donuts, they can say, yeah, we can claim credit for that because they were exposed. That's essentially cookie stuffing, which is parallel to affiliate fraud and whatever, whatever. But now if 380 million devices had a cookie stuffed, meaning it was marked as exposed,

Brett (46:30)
Hmm.

Yeah, that's like cookie stuffing in the programmatic ecosystem.

Augustine Fou (46:51)
then anyone, including the person that walks into Dunkin Donuts every morning anyway, would get attributed to that location-based company. that's how they... Yeah, awesome, right? We got so many people that walk into Taco Bell, Dunkin Donuts, whatever.

Brett (46:59)
Yeah, terrific, terrific campaign performance. Off the, off the...

Rio (47:07)
Even

though not a single new person, no incremental lift at all, the same people who have gone in anyway, just people who have anyone with a cell phone, it's wild.

Augustine Fou (47:09)
Correct. Correct. Yeah.

Brett (47:12)
Yeah, which is the

exogenous factors of like I'm already I already have high propensity to buy Dunkin Donuts, right? That's the explaining factor, right?

Augustine Fou (47:13)
Yes.

Yes, exactly, So

for years and years...

Rio (47:21)
That is a tax or a toll,

right? Because you're just basically inserting yourselves it hasn't pulled any foot traffic, it's wild.

Augustine Fou (47:25)
Yeah, yeah,

right. So you didn't have to pay for that, but yet the reporting said your campaigns were looking so awesome. Let's go stick more money into this because the ROAS was so awesome.

Rio (47:36)
Well, this is a big criticism of retail media, right? There's a lot of people who would have bought anyway, right? And we're just, you know, should you actually be doing this?

Augustine Fou (47:42)
Yeah, I mean, we're

gonna bring this full circle and basically say that the marketers are too young to actually understand anything that we just said. So basically they'll just take the report that Google gives them or whatever platform they're buying from gives them and say, look, your ROAS is 100x or 500x. Cool, bigger numbers are better because I'm a video game player and so higher scores means I'm doing better. That's all they understand.

Brett (48:09)
Yep.

Augustine Fou (48:10)
I'm gonna get a lot of hate mail after this, you know, that's that's really the problem I Do always all the time death threats and everything.

Brett (48:12)
You

Rio (48:13)
I think you probably always do, Dr. Fou.

Brett (48:15)
Yeah.

and Rio, that's a great, great point. I won't say what our CEO says about retail media, but that's a great point about retail media because the on-premise, on-site experience is what, yeah, that's what retailers do offline as well as online. But once you start going off-site and you're targeting people via the exchanges across the open web, it's the same buy you could do with The Trade Desk or DV360 or otherwise, right? So you're saying, well, what are you really offering?

Augustine Fou (48:28)
They're ready in the store.

Brett (48:45)
other than the claims of point of sale attribution when somebody goes back and buys something in your store, right?

Augustine Fou (48:51)
Yeah, and if more

advertisers started to ask about incrementality, the entire house of cards of ad tech and digital advertising will fall apart, right? Because there's just not enough.

Rio (49:00)
That was like a big

word in the last year, early this year. I even did a podcast on incrementality when I was talking about it and everyone rolled out these incrementality reports, right? You could push a button, boom, right? I don't know, then it kind of disappeared. But I guess it's interesting as we talk about this, it seems like a lot of people benefit, right? Certainly it makes marketers look good, agencies like it. I'm not saying they like the fraud, but there's no reason not to like it, right? As long as it's not caught, right?

Brett (49:08)
yeah. Yeah. yeah.

Rio (49:29)
All the platforms make money off it. It's not just everyone is like participating in this. They're not, but a lot of people seem like they're not, if not

They're saying, well, we're identifying IVT and non-human traffic. And

our clients aren't paying for it. I mean, but they're certainly paying whoever they bought the media from. It really depends on is there a clawback in the contract. So I think a lot of people realize the agencies are buying media on behalf of the brands. even if you find fraud, it's hard to recoup it. I've gone through this with clients where you approach the agency, so we found out 20 % fraud. Is there a clawback? Do you get a make good? It's not in the contract.

Brett (49:57)
At volume, at scale.

Yeah, yeah and-

Augustine Fou (50:11)
They don't want

to give it back, right? Yeah.

Brett (50:13)
Yeah,

and the chief investment officers at the big agencies will go and buy, I know one that will buy $18, $19 billion worth of ad inventory, largely digital. And because you get this volume based discount and then they can add margin based on what they're selling it at at a CPM rate. And so ⁓ it helps them keep the margin in a business that's very tight on margins. But to your point, you're disincentivized to sort of eliminate half of that as

Rio (50:22)
or principal buying.

Brett (50:42)
invalid traffic or

Rio (50:43)
Well, you

Brett, when you mentioned principal buys, I mean, just think that, like, so most like, if you are the principal, I mean, like, how can you be an agency representing a brand? I mean, I get it, like, that's the whole, like, part of the pitch of agencies are buying such volume, we get good deals, and it's true. If you want to buy, let's say try traditional media globally, you probably should work with a big hold, right, because they have the relationships and they know how to, they know, they're the only ones who can actually execute it, but I'm principal media buying just strikes me as being just fundamentally like a flawed model. I don't know if Dr. Fou, what do you think?

Brett (50:58)
What?

Augustine Fou (51:12)
The way

I think about it is the principal media stuff where the big agencies said, OK, we have so many big clients. We're pooling our buying power to get a better deal from NBC, from television networks. That worked years ago for television. But in digital, in programmatic specifically, when the theory is that you're going to bid for every single ad impression, you don't need that. In theory, you're supposed to get the best price per impression.

There is no need for an agent. Yeah, it's an auction. So by doing principal buying for programmatic media, they've kind of defeated the entire premise or the purpose of programmatic media. So now what I've seen is just vast abuses where the hold codes, like you said, Brett, they're buying billions of dollars and then they're basically selling it to their own customers for an undisclosed markup. And those markups could be like

Rio (51:42)
It's an auction, right?

Brett (51:43)
Yeah, exactly.

Augustine Fou (52:09)
90 % of the CPM, right? It is just

Rio (52:12)
Well, let's get into

markups. I think that's an interesting one. So I've run media audits a couple of times, and then I found it very difficult, surprisingly, to actually find what the markups were. Now, you look at the contract as saying we're going to take 15%, right, or something around that. But then you find out, and I remember one time we were performing an audit. This agency, they wouldn't send us the markups. They wouldn't give us access to the DSP. Finally, was wild. Someone accidentally sent us a spreadsheet one day that had all these columns hidden. We unhid them and all the markups there.

Augustine Fou (52:38)
Yep. Yep.

Yep.

Rio (52:42)
They were not 50%. Some of them had 400 % markups on certain buys. And then we asked them, what is this? This is crazy. Why are you charging a client four times what you actually pay Google while it's targeting? I mean, it's bullshit, right? Because that media didn't take that many additional work to do that. I don't know what you've seen.

Augustine Fou (52:45)
Yeah, 100%. Yep. Yeah.

Brett (52:46)
Wow.

Augustine Fou (52:51)
Yeah.

Yeah. So

the very basic thing that I tell my clients to ask their media agency is show us the CPM that you ultimately paid for the media you just bought us. OK. They will find every excuse under the sun to not show you that because they are hiding the 400 percent markup.

Rio (53:16)
Is that why they should maybe go pay per direct with the platforms? Does that solve it?

Augustine Fou (53:19)
Yes,

yes. So there's no advantage for a big company to have their media agency buy programmatic media because programmatic media is supposed to be an auction on a per impression basis. So for the media buying, do it your damn self, right? Get a DSP where you can buy it yourself so you can see what CPM you actually paid and then pay your media agency for doing the creative and whatever else you don't want to do for you.

Don't pay them to buy your media for you because they're going to be hiding margin in there and you're going to be paying way, way more than what you're getting. So if, for example, you're paying a $15 CPM for display ads, you'd be lucky if the agency were paying $1 for those ads, CPM for those ads.

Brett (53:50)
Yeah.

Yeah, which is like the price of remnant inventory, basically.

Augustine Fou (54:06)
Yeah,

because they need to keep the margin. So all of that is hidden. You don't see any of that. They won't tell you what CPM they paid. And in some cases, they don't even know what CPM they paid. So I don't hold the individuals who work at the media agencies responsible. I hold the holdcos responsible, because that's their business model to do that. So for advertisers, and I have some good ones who have taken parts of this in-house, like the buying part itself.

Just take that part in-house. Use a DSP that you can control yourself. If DV360 is too big and the minimums are too large, use something else. There's other DSPs that you can use now for yourself. And that way, you actually get to see what you paid. So if you're paying $15 and you don't actually know what's being paid out on your behalf, if you take over the DSP and you buy yourself and you still want to pay $15, then most of that's going to actually go to the publisher for working media, for showing your ads.

but for the most part, you could probably get it for way cheaper than the $15 yourself.

Rio (55:08)
It's interesting, I mean, those are all really good points. I've helped a bunch of clients build in-house teams, and I think it's rare that big brands will actually buy the media themselves, but I think just having the platform licenses themselves, having access themselves, being able to log in and not fight with an agency about understanding what the markups are, I think that's huge. But it's interesting though, I remember every time we've gone through this, the contracting is very, very difficult, explaining to, let's say, know, the, you know, the...

Augustine Fou (55:19)
Yes, to look at it, yes.

Yes, correct.

Correct.

Rio (55:33)
the brand, you're going to have to pay Google and Meta Net 30. That's non-negotiable. They will shut you off. It's not going to be Net 90, right? I mean, I think explaining to them sequential liability, there a lot of things that go into these that takes a lot of learning, but there's a huge benefit, right?

Augustine Fou (55:47)
Yeah. And I think, you know, I've seen that happen with other things. So I've seen over time, the brands have brought their verification in-house because, you know, years ago there would be a case where you'll say like, show us the verification vendors data. And the agency will say no, because they hold the contract and the advertiser has no rights to see the data. Right. So then I saw a few cases where the advertiser brought the verification vendor contract in-house.

So they could actually log into the dashboard and run reports themselves. So that was a step for, know, that was better step than what they were doing before. So I'm hoping that some of that will happen on the buying itself. Now there are companies that I work with, a Goodway Group where they will kind of provide that transition plan. So if the advertisers are not ready to bring it all in house, cause they simply haven't hired the right people yet, ⁓ big company like Goodway Group can actually help them ⁓ kind of make a plan, a transition plan.

Where they would do the managed service for them, buy it for them, and then use that process to educate. And also while the advertiser hires a person or two to actually do that in-house. So I think that's a rational way of doing that. And like, OK, let's use programmatic the way it was meant to be used. It's an auction system. You're supposed to get the best price for every single ad impression you buy. But all of that has not happened in the last 15 years because of all of the misaligned incentives at

every step of the stack. So I think it's time to get better. And once the advertisers actually buy their own stuff, they'll actually be buying better stuff. They're not going to be buying the remnant crap because they don't know that they're buying the remnant crap right now.

Rio (57:28)
What do you think is a solution. Do you think we need regulations? I guess everyone's been saying industry should regulate itself, but you could probably argue that doesn't work so well, but I'm not a big believer in regulations. I'm just curious.

Augustine Fou (57:35)
No, it's like, know, we've, yeah, it doesn't

work. I mean, because you can have all the regulations you want. It adds burden on the good guys and the bad guys don't follow it anyway. Literally doesn't help, doesn't deter the bad guys, okay? Yeah, and we have...

Rio (57:52)
So innovation.

Brett (57:52)
Well,

you need enforcement with your regulation and can you trust the industry to regulate itself, know, through the IAB and...

Augustine Fou (57:54)
Yeah, yeah.

No, and here's

the problem with enforcement in digital advertising. ⁓ the people who are supposed to enforce, so whether it's the MRC, the trade bodies, whatever, they don't have the actual tech tools to go do the audits. You guys have done audits. I've done audits. They don't have those tools. So when you say, us your log-level data or give us the reports, it's going to be whatever the platforms tell you. It's going to be 1 % fraud because that's what they do.

told you, right? So they're not going to be able to figure out unless they really probe and all that kind of stuff. at best, you're going to get something like the ISBA reports from years ago, where it's like 50 % of your spend went to somewhere else because we couldn't account for it, right? Only 50 % went through the publisher. You're going to get a generality like that that you can't act on. So regulations, laws, whatever, that's not really going to cut it. It's going to take years and years. Advertisers need to care about their own dollars.

Brett (58:28)
Yep.

Augustine Fou (58:55)
If they don't care, no one else is going to care for them. So I'm seeing a turnaround, but it's a slow turnaround. It's literally one or two or literally single advertisers will say, fuck that. We got to look at what we're buying. We can't just trust the reports that we're given. Once that happens, a lot of good stuff happens. it has to start with the brand. Yeah.

Brett (59:17)
It'll change behavior. Yeah. Yeah.

Rio (59:17)
So it has to start with the brands. has to really go back and they have to say like, not putting up with this. Like we don't, not going

to pay for 25 or 50 or whatever percent fraudulent traffic or whatever. Like, okay.

Augustine Fou (59:26)
Yeah. And

it really, we can't solve it for the industry, but we can solve it one brand at a time. And we will help those brands who actually want to be helped.

Brett (59:37)
Yeah. Yeah. then, and then there's a domino effect, you hope, right? You know, because they realize the benefits from, from their brand competitors, from incumbents.

Augustine Fou (59:44)
Yeah, mean their peers are

going to tell them, look, we just brought this in-house. We just started using better analytics rather than fraud detection. And we could see that, oh, well, they've been telling us it's been 1 % for the last eight years. OK, when we first put better analytics in there, I'm not even going to name my platform. But it's like, OK, when we have analytics, we can see it's like 90 % red. OK? And with that, you can actually see which ones are the bad sites and bad apps that you need to take out first, right? Which are the most egregious bad guys?

Rio (1:00:12)
Yeah, it's interesting. I think a lot of DSPs are starting to expose log files, which, amazingly, many of them didn't until recently. So I think that's good. if you bring these in-house, you'll have access to that data. You'll need teams to analyze it. But I mean, that's part of the in-housing work, is you need to build that muscle. I think that's interesting.

Augustine Fou (1:00:19)
Exactly.

some of it, but

let me help to ⁓ lower the fear for the advertisers. When you tell them you have terabytes of log-level data to process, they don't know how to do that. They don't even know who to hire to do that. I'm going to tell them it's not necessary. And the reason it's not necessary is that the log-level data also contains errors. The log-level data comes from recording all the bid requests. So in the bid request, there's going to be the domain.

the CPM, the this, this, and this, or viewability, ⁓ user agent, whatever. Bad guys will lie on all of it. ⁓ So if you have a fake site, fakesite123.com, you're never going to put your own domain in there because you're never going to get any bids. The fake sites, the fraudsters will always have to lie, and they'll put espn.com, marthastewart.com foodnetwork.com. They will always lie. That's what gets recorded in the log level data. So you won't see.

the fake sites in there. That's the problem with log-level data. Now, the reason a lot of people have been talking about log-level data is because of the analytics reports. They process log-level data. And yes, they can see some problems. But those are tiny problems. would characterize those. Yeah. I would say that it's the script kiddies who messed up and forgot to disguise their site. That's what they're catching.

Brett (1:01:38)
Yep. The CSAM reports and the MFA reports. Yep.

Yeah. And that's

a small percentage of the larger problem is what you're saying.

Augustine Fou (1:01:51)
That's a small percentage,

right? Because organized crime, well-practiced criminals have, yeah, they're going to have done this for 10 years. So they're going to lie better on their bid request. ⁓

Brett (1:01:57)
They're not gonna be that obvious in terms of... ⁓

Rio (1:02:04)
Yeah, it's interesting you

talk about the 10 year, going back 10 years because I think like about 10 years ago, maybe a little more, mean, like media agencies really split off from creative. Remember, that was a big decision everyone made, right? I actually think that was a strategic error, right? Because I think that forced people on the media side to overlook a lot of these things, right? You have to make your money. So if you're to make your money, then you're going to be marking up stuff and not telling your client. You're going to be not caring if there's a lot of fraud because you don't want to because I actually think that

Augustine Fou (1:02:30)
Yeah

Rio (1:02:33)
some of these trends you're talking about, in-housing things, I think it'll really be good for agencies. They'll be able to focus on what they're really good at. yes, brands need them. That's not gonna change.

Augustine Fou (1:02:39)
And they'll get paid for their expertise.

Brett (1:02:41)
Yeah,

Augustine Fou (1:02:44)
So I

Brett (1:02:44)
strategic expertise, MMM, creative expertise, all that.

Augustine Fou (1:02:44)
think it's really going to be, yeah. And the ones that are redundant, that are not actually adding any value, will die and go away. And that can't happen soon enough. So I think it's all going to be a good thing, but we're going to go through a period of pain where all these transitions are hard because people like to do what they're used to doing. Change is hard.

And business model changes even harder, right? So yeah, you're gonna have to figure out what you're actually good at and what you can actually charge for and what the clients are willing to pay you for, not just try to mark up stuff and not tell them.

Brett (1:03:19)
Yeah, and programmatic media buying at scale is probably not one of those places that they need to be playing in.

Augustine Fou (1:03:24)
Yeah, the whole

point of programmatic media buying is to automate that. So let the algorithms do their job. So don't hide all this stuff with all these shenanigans and tricks and basically financial fraud.

Brett (1:03:31)
Yeah.

Rio (1:03:36)
So five years from now, do you think this problem solved? Do you think it's partially solved? What do think we end up?

Augustine Fou (1:03:41)
We have a few more big advertisers who will actually be doing awesome work. I have one and I also work with Goodway Group and they're doing great across many clients. And that's because they have the right tools, they have their incentives in the right place. So they're actually looking at the data and that kind of stuff. So I'm hopeful given that, right? I've seen what it can be, right? How good it can be. So I'm hopeful that more advertisers will come around. But like you said, there's a problem of ⁓ expertise or experience or, you know,

hiring the right people if you need those people in-house. A lot of brands don't have that, because they've outsourced most of these things for so long. They just don't have the expertise. But I think with a lot of the layoffs that we've seen, there's real talented people who are actually available. Hire those people. You only need a few gems. That's all you need. Because then they can help you set up the campaign, look at it periodically. They don't have to look at it daily, because you need the algorithms to work. Look at it once a week.

Rio (1:04:40)
It's not like the

agencies look daily, don't worry about it.

Augustine Fou (1:04:42)
Yeah, exactly. ⁓

But you know, like check it once a week, make small tweaks, and everything's going to get better.

Rio (1:04:49)
And I think too, like I agree with you, think brands will say, it's hard to find the talent, which I think historically was true. Not a lot of people out there with the right talent, or even if you hired them, they don't like to stick around. A lot of, I think the most talented people, I mean, you're a working agency, you get to jump between big brands, people jump agencies a lot. It's kind of, there's a certain lifestyle there. I think working for an in-house team is very different, right? So, but I think that's a good point. There's a lot of talent out there that can help you get more value from what you're spending. And I think work better with your agencies too.

Augustine Fou (1:04:56)
Yeah, that's true. Yeah.

Brett (1:05:19)
Yeah. So, so Dr. Fou, if you had to mandate three immediate changes in the ecosystem tomorrow, what would they be? Like if we had to summarize your key recommendations, which have come out in a number of ways through this conversation, what do think they would?

Augustine Fou (1:05:32)
Yeah,

this is kind of un-pre-meditated, so it's just literally off the top of my head. One of the key things that I've been looking at recently, and I was talking with John Watts at CIMM, is about CTV. Define CTV more clearly, because right now, because it's such a fungible definition, you're getting a whole bunch of ⁓ low-cost crap that's being mixed in. So the CTV ads are running on bad websites and bad mobile apps. That's not CTV.

If you just define CTV more strictly, as we do in my platform, Fou Analytics, if it didn't run on a big screen TV, you can't sell it as CTV. You can sell it as OLV, online video, but don't charge me CTV prices for stuff that didn't run on the TV. Just that basic change, just tightening up the definition, will start to lead to better behaviors. If you see your vendor is actually selling you the crap that's running on websites and

Brett (1:06:18)
Yep. Yep.

Augustine Fou (1:06:32)
mobile apps for CTV prices, cut that vendor, buy from someone else who's not doing that. So that's what I mean by such a simple change will make things better because CTV is where everyone is piling investment in and it's the highest CPM. So you need to be more careful with your dollars there.

Brett (1:06:48)
Yeah,

it's gotta have the value that television, know, sound, sight and motion delivered on a giant screen in somebody's living room and an app.

Augustine Fou (1:06:51)
Yes. Yeah.

Yeah, there's no cheap CTV. Let me just put it that way, OK?

Rio (1:06:59)
rates have gone down. I think Netflix artificially inflated the market when there was $50 CPM or something like that. But they've crashed down to earth. But still good CPM, like in-stream inventory is still like $25, $30, which is...

Augustine Fou (1:07:13)
Yeah, should be,

right? If you're seeing $5 CTV, it's not CTV, okay?

Brett (1:07:17)
Yeah, yeah. Well, and advertising against streaming catalog, like a catalog of shows and content that these providers have, is very different than live streaming, for example, right? Where your CPMs are going to be television quality or television level.

Augustine Fou (1:07:29)
Yeah, I think both of you are

still talking about buying from real media sellers, real to see TV. Right. I mean, what I see is like, okay, these are CTV campaigns, but why are they running on these apps and nobody's ever heard of before? Right. So it is, know, so I'm seeing a whole bunch of other stuff that's not even running on the real CTV. Yeah. Yeah. And then the, yeah, the second one would be really challenge your legacy verification vendors. Right. So very basically.

Brett (1:07:35)
Yeah.

Yeah.

Yeah, on the Hulus or the Netflixs or yeah. What else?

Augustine Fou (1:07:59)
They've been telling you fraud is 1 % for the last eight years, right, through their press releases and whatever. Everyone knows. It's common knowledge that fraud is not 1%. So ask them, how the heck did they get that number? OK, now part of what will start to come out is the fact that they only measured one in 100 impressions. If they didn't measure the other 99. Now, obviously, this kind of sampling doesn't happen all the time. But you should actually ask, do they do sampling?

what percentage of your ads were actually measured with a JavaScript tag where possible. Because if they didn't measure it with a JavaScript tag, they're not going to catch the bots. Exactly. If 99 of the 100 ads were not measured, you should not assume that they were fraud-free. You should assume they just didn't measure it. That's all.

Brett (1:08:32)
Yeah. Yeah.

Yeah, extrapolation is not going to work here, right? It's not like a census level prediction of audience, right?

Augustine Fou (1:08:51)
So I think as more more advertisers ask their legacy verification vendors where they got their numbers, they'll start to understand, like I said before, the 1 % is not all the fraud there is. The 1 % is all the fraud they could catch. That's a big difference. And so it gives them the opportunity to upgrade their tools so that they can see better. And when they can see better, they can do better.

Brett (1:09:16)
Yeah, and we talked about that topic with Stephanie

Layser at AWS in our first episode. And we all kind of came to the conclusion that there needs to be some competition in that space to force innovation and to force change. So what's your third recommendation?

Augustine Fou (1:09:26)
Yeah. Yeah.

It'll kind of be what we talked about earlier, which is like focus on attribution and figure out the incrementality and don't just take the attribution reports that you're given because in most cases it's going to be over claiming credit for sales that had already occurred. So for advertising the care, you want to know what actually drove incremental sales, which were the sales that would not have occurred in the absence of the advertising, right? If the sales occurred,

Brett (1:09:36)
Yep.

Augustine Fou (1:09:59)
without the advertising, you didn't need to pay for the advertising, right? Because those are people that would have bought from you anyway. it's, yeah, a little bit, yeah.

Brett (1:10:06)
Yeah. And that takes a little data science muster and intelligence. Yep.

Rio (1:10:09)
It takes testing too, right? mean, again,

incrementality need an approach. Not like you press a button, boom, you're done, right? In fact, any platform that offers that, wouldn't necessarily trust it. would say, in criminality means...

Augustine Fou (1:10:13)
Correct.

Yeah, you have to do your own

testing. And I think just to simplify for advertisers listening in on this, it would be to run geo holdouts. So this goes back to ⁓ eBay, right? So they turned off the West Coast, and the sales kept happening. They turned off search ads on the West Coast, and the sales kept happening. then they could say, well, that's because the sales were going to happen anyway. So I think it's a.

Rio (1:10:26)
That's his key,

Brett (1:10:41)
Yeah, an unexposed

audience that's fully verifiable, right?

Augustine Fou (1:10:44)
Yeah, exactly. So I

think for advertisers, don't be afraid because they'll say, we need a vendor to do the incrementality. OK, yes, there are vendors that can do that. But there are simpler ways that you can do yourself. Like turn off a state, turn off a city, see if the sales are actually still occurring. If they are, then turn off another one, just check. So there are simple experiments that you can ⁓ do to see if the sales actually were caused by the ads.

If the sales continue, that means it wasn't caused by the ads.

Rio (1:11:18)
That takes actually thinking at work. I mean, would argue there are some marketers out there who probably just manage vendors, But if you're a true marketer, you should love this shit, right? Because that's the fun, right? You're coming up with an incrementality test, right? You run the test, you see what it's like. I I think that's the fun part of marketing, not just like getting stuff from your agency or vendor every month and showing your boss how well you're doing, right? I mean, that's whatever.

Augustine Fou (1:11:22)
man, we're being too mean to marketers. Yeah. Yes. Yes.

Brett (1:11:24)
Yeah.

Yeah. Well, and much easier-

Augustine Fou (1:11:31)
Yes.

Brett (1:11:40)
Yeah. Yeah. And it's applying

the scientific method to this, right? You're saying, how do you actually prove that someone who's, it's harder when you're doing a national television buy or a national radio buy, because then everybody has the likelihood of being exposed, whether it's trackable or not. When you do it digitally, that's when you can start getting to turn this off. Yeah. Similar demographics, similar audiences on and off, compare and contrast, you know, build your algorithm to properly predict this

Augustine Fou (1:11:43)
That's right.

Yes.

You can start to control these things, yeah. Yeah.

Yeah.

Rio (1:12:06)
had to say something bad about advertisers.

Augustine Fou (1:12:06)
we brought it full circle in exactly an hour.

going to get hate mail from every party in the entire ecosystem after this one.

Brett (1:12:13)
Well, Dr. Fou, you lived up to the hype. This was phenomenal, just having this conversation and great hearing your perspective, your hands-on experience, the research you've done. I'm sure a lot of us in the ad tech ecosystem that have been following you for a while are aware of some of these things, but I think this is gonna be good new information for a lot of our listeners. So this is terrific.

Augustine Fou (1:12:18)
Thank you, thank you.

Rio (1:12:19)
That's awesome.

Augustine Fou (1:12:36)
Thank you. Yeah. Great. Great dialogue. Thank you.