The Dan Rayburn Podcast

Executive Interview: Dr. Fou Details YouTube’s Ad Scandal and the Complexities of Fraud in the Online Video Advertising Market

July 17, 2023 Dan Rayburn
The Dan Rayburn Podcast
Executive Interview: Dr. Fou Details YouTube’s Ad Scandal and the Complexities of Fraud in the Online Video Advertising Market
Show Notes Transcript Chapter Markers

This episode offers a deep dive into the murky waters of ad fraud detailing the allegations against YouTube. Dr. Fou, an expert in ad fraud joins me as we head straight into the eye of the storm, exploring the latest report alleging Google's TrueView ads appear muted and on autoplay on third-party websites. Discover how these misrepresentations are impacting advertisers' trust, why YouTube's reply to the allegations ignored the issues, and what this could mean for YouTube's revenue.

We also detail challenges in the overall video advertising industry when it comes to detecting ad fraud, why advertisers don’t have much incentive to fix it, the lack of clear measurement, methodology, and transparency in reporting and the impact of AI on the market.

If you're interested in understanding the complexities of ad fraud in the online video advertising market and the challenges and trust issues that plague the industry, or if you're an advertiser wanting to ensure your campaigns are not misrepresented, this is an episode you won't want to miss.

Research report and data from Adalytics: https://adalytics.io/blog/invalid-google-video-partner-trueview-ads

Google's reply to the data: https://blog.google/products/ads-commerce/transparency-and-brand-safety-on-google-video-partners/

Podcast produced by Security Halt Media

Speaker 1:

Welcome to this week's edition of the Dan Rayburn podcast, the show that curates the streaming media industry news that matters most, unvarnished, unscripted and providing you with the factual data you need to know, without any of the hype, the pulse of the streaming media industry.

Speaker 2:

Welcome to the Dan Rayburn podcast. I am Dan Rayburn, and joining me today for a special executive interviews podcast is Dr Foo. Dr Foo, thanks for joining me today.

Speaker 3:

Thanks, dan, good to be here.

Speaker 2:

So what we're going to talk about today are a couple of different interesting things that have been taking place in the market over, I would say, more than the last couple of months, obviously, but what we're going to focus on today is recently, youtube has been having some accusations against it based on some data that has come out from analytics around some. Basically, what they're saying is fraud. So what we're going to talk about today is Dr Foo.

Speaker 2:

I think of you online as, even though your name's FooFOU, I think of you as, like Dr Ad Fraud. That's really sort of, in my head, the nickname I have for you because at Foo Analytics, you're pushing out so much information and working with advertisers on what is taking place with fraud on the internet when it comes specifically to advertising. So what I'd like listeners to know first and foremost is go into your background a little bit, because you cut really deep into analytics when it comes to advertising on the web, and what I like about that is there's a lot of people out there who are specialists, for instance, in something large like infrastructure, just video in general, or, say, encoding, but you cut really deep on one thing specifically. So talk about how you're tied into the YouTube ad scandal in terms of what you do every day, that's tied into it, and the type of work you've been providing and how long you've been doing it for.

Speaker 3:

Sure, so, at the high level, I've been doing digital marketing for 25 years, so dating back to mid-90s, like 95, 96, when we had to convince advertisers they needed a website. So a lot has changed, obviously, in the last 30 years, and these days we're buying and selling trillions of ad impressions using algorithms. So a lot of people have heard of AI and we've been doing something like that in programmatic so what we call real-time bidding for the last 10 years. So the last 10 years, I've actually focused on the problem of ad fraud, because, as we started automating the buying and selling of ads, it also allowed the bad guys to automate their fraud and scale their fraud.

Speaker 3:

So these days they're using algorithms to create thousands of websites by template, and they're also using bots, which are fake users, to repeatedly load the web pages or continuously play the mobile apps, and so that's the area that I study, and the way this intersects with the YouTube scandal of the last two weeks is that digital ad fraud is basically the misrepresentation of ads.

Speaker 3:

So advertisers think they're buying ads shown to people humans but instead they're actually buying ads on sites that no one has ever heard of before and so therefore there's no humans who are visiting and therefore all the traffic is bot traffic, so those ads are being misrepresented to the advertisers as ads that they're buying legitimately. So there's lots and lots of that. It's really proliferated in the last 10 years. So, basically, where I sit is right at the intersection between the technology and the marketers. So I have a platform that I've built over the last 10 years to use as a tool for auditing digital media, and that's where we can find things like bot traffic and fraud, but also other things that are not necessarily fraud but are suboptimal in their campaigns, for example, ads that run in the overnight hours when humans are sleeping and bots are out to play.

Speaker 3:

So those are kind of examples of things that I do primarily for advertisers. So advertisers are my clients. Foo Analytics is the analytics platform that I use to help them audit their digital media.

Speaker 2:

Very cool. Okay, so before we go into what's going on with YouTube, maybe from a high level, talk about why is ad fraud so underrepresented in the market. By that I mean you've talked about previously where certain companies will say, oh, it's only 1%, and we know that's not the case. Do advertisers not care? Or is it that vendors are just so good at selling these solutions and not having to provide real methodology? Because this is a big problem in the ad market there's no other way to say it and yet there doesn't seem to be that much exposure on the problem and how big it is, or really fixing it.

Speaker 3:

Yeah, that's exactly right. But I'm not going to blame the advertiser because they've had limited tools, right, the tools that they were using what I'm going to call collectively the legacy fraud verification companies, or those vendors. Those tools, like you said, just keep reporting 1% fraud, 1% IVT, which stands for invalid traffic, and they've been reporting that for at least the last six years. And the reason I say six is because the trade group called the Association of National Advertisers has been citing those numbers in press releases year after year for the last six years, saying that fraud is 1%. So if you're a big advertiser and you hear your trade group representing North American advertisers repeatedly saying that fraud is 1%, what are you to think? You're going to say? Oh well, they're saying it's 1%, so it's got to be 1%. But unfortunately, I think, as the YouTube scandal and other independent researchers have shown, fraud is far larger than the invalid traffic, the IVT number, that these legacy fraud verification companies have reported for years and years. So I think that's why the advertisers have not taken more aggressive action.

Speaker 3:

But even in the last two weeks, the ANA itself put out a preliminary report which they publicized it can. So they were talking about it at the Riviera, saying that MFA websites, so made for advertising websites, represent up to 21% of the impressions. Now you should think of that as OK. All websites that Kerry ads are made for advertising, so this is not anything new, but they made it sound like they discovered something new and now everyone's jumping on the bandwagon. But long story short is there's far more than 1%, and I think more and more the advertisers realize that. So they're starting to ask harder questions and demand more details, which they all should be doing to say, ok, well, let's take a closer look, because I don't think fraud is 1%. I think all of that's good for the entire industry.

Speaker 2:

It's interesting. It seems like it's the dirty little secret, and yet those in the industry, we all know this is happening and you're talking about a lot of money here at stake, and yet advertisers still don't seem to be doing enough, even if they're limited in the tools. Ok, build some of your own tools, but it sounds like it really comes down to what data is being shared with them and what they're allowed to see. Is that part of the problem as well?

Speaker 3:

Totally right. I mean, we've had years and years of transparency studies, so I think there's been a total of six. The latest one was the ANA doing that right, so they announced and preliminary results. But World Federation of Advertisers has done it in years past. Isba, which is the Society of British Advertisers, they've done it.

Speaker 3:

And all of these transparency studies have reported time and time again that there's not enough transparency. So what? Which means that the advertisers are basically getting Excel spreadsheets that say fraud is 1% and you've got this many ad impressions and this many clicks, things fine, don't worry about it, keep spending. So that's literally not enough information. And even when they buy fraud verification from these legacy verification vendors, the IVT number comes out to be 1%. So if you're an advertiser, it's like, oh well, if it's 1%, I really don't have to worry about it. It doesn't seem like there's a big problem.

Speaker 3:

The challenge is that the fraud verification companies are again missing a lot of different kinds of fraud. It's not just IVT or invalid traffic going to a website or bots. There's many other forms of fraud. I'll just name a few simple ones to understand Ad stacking, where instead of one ad in the ad slot, you're sticking 100. So you're not able to detect that whole stack of the 100, or these legacy fraud verification vendors are not necessarily looking for those kinds of things. Things like pixel stuffing or maybe the alarm clock app that's loading ads throughout the night when the app is not being used and when the mobile device is not being used. So all of those are ads that are not seen by humans. So it's a form of fraud, but they may not be caught and reported by the legacy fraud vendors. And one other problem that I've seen is that the bots are very good at blocking their detection tags. So when they see a double verify tag, they block it. When they see an irritable ad science tag, they block it, because obviously the bots want to avoid getting caught so they can keep doing their fraud.

Speaker 3:

And the sad part about this is that when these vendors have no data, they market as no fraud. So that's a very common sense question I ask of advertisers Do you think that no data means no fraud? A lot of them will say, oh, that doesn't make any sense. If you don't have any data, you would probably suspect it more. It's almost like guilty until proven innocent. You really want to have that data to make sure you know where your ads are going and also whether bots or humans loaded the ad. So when there's no data, you shouldn't just assume that it's no fraud. And that's really where the legacy fraud vendors go wrong, and those are the tools that have been available to advertisers in the past. So that's why now, when they're being more cautious with their ad spending, they're asking for my help to audit campaigns, because they've even had a gut feeling that something was wrong. They just couldn't see it in the reports, because they were just told repeatedly that fraud is 1%. Don't worry about it.

Speaker 2:

Fascinating. So many problems going on there that we could dive into. Let's go into. Appreciate you giving a high level overview there.

Speaker 2:

Now let's go into what we've been hearing about YouTube over the last few weeks and also just some of the things YouTube has also announced regarding ad blocking. So we're recording this on Tuesday, July 11th so about a week ago, a little more than that, Company Adlytics reported that Google's TrueView ads and, for those that don't know, that's a proprietary ad format that allows advertisers to pay only for actual views rather than impressions. So what the report was saying is that these skippable ads have been appearing in a muted autoplay format on independent websites. Mobile apps, they also claimed about 80% are not running on YouTube proper website, but what they're calling low quality third party sites Yep. It also said that they were seeing ads running on sites in sanctioned countries, including Iran and Russia Yep. So maybe break down for listeners what was collected from a high level. Do we know the methodology to be accurate? The report and the paper is fascinating because of how detailed it actually is 200 pages with code samples and screenshots.

Speaker 2:

Which is amazing, because we don't usually get that. So you also know the person personally who worked on this for multiple years. So give listeners an idea of what exactly was collected and why it's important.

Speaker 3:

Yeah, it's gotten a lot of pickup, but I would say that a lot of the reporting has been slightly inaccurate. They've been focused on whether the ads ran on YouTube or not on YouTube what they call GVP, which is Google Video Partners, and that's one of the issues, but it's not the main issue. So when an advertiser is buying TrueView audible, skippable ads I'll use the example that most humans are familiar with. So when you're watching YouTube, you would get an ad pre-roll. So before you get to watch your content video, they'll run an ad and on the lower right-hand side you'll see a button, a black button with white letters that says skip ad. So that's Google's proprietary ad format called TrueView, and just like their search ads where you don't have to pay until you get the click, in this case, the equivalent is you don't have to pay unless the person doesn't skip the ad and actually watches your ad. So that's a great idea and I think a lot of advertisers have bought into that and are buying TrueView ads.

Speaker 3:

The problem is when the ads actually run, not in that format. So when it's muted, it's obviously not audible. When the skip button is covered up, it's obviously not skippable. And when, basically, the ad runs without content. That's called Outstream instead of Instream. So Instream ads are the ones that run pre-roll, prior to the content video, mid-roll, which is during a longer content video, or post-roll after the video ends. So that's what advertisers thought they were buying. But when we see some of these ads running on third-party sites, collectively called GVP, google Video Partner sites, many, many of those sites are running those ads in a non-compliant way. So first of all, they're Outstream, so they're just running a whole bunch of those ads by themselves, without any content video. They're just showing a whole string of ads back to back.

Speaker 2:

So there's no content within the ads at all.

Speaker 3:

No, there's no content video. So when you're on YouTube, you're watching a content video from some kind of creator and then the ads run before, during and after. So I'm going to differentiate the video ad from the video content. The point is, a lot of these outside websites are cheating by loading just the video ad itself, with no content with no YouTube video, so those are called Outstream.

Speaker 3:

Ok, now, when those Outstream video ads are just being run continuously back to back without human initiation right, the human did not play press play, they played a YouTube video and then running muted and off-screen and all that kind of stuff, those were misrepresented as if they were TrueView, audible, skippable ads. Right, that is where the problem comes in. So the advertisers thought they were buying ads that were part of that TrueView proprietary standard. Right, they were only paying when the ad actually ran and was not skipped by the users. So that was the key issue, right, outstream versus in-stream. The secondary issue was that it was running on properties outside of YouTube.

Speaker 3:

Right, most humans are used to seeing these ads on YouTube and in some cases they do skip them because they want to get to the content sooner. Right, they don't want to wait for 30 seconds. So in those cases, again, the advertisers thought they were buying one thing, but then they were getting something else and all of it was reported as TrueView, audible, skippable ads, and that's where the misrepresentation comes in. Right, and by far. Yeah, that's the bigger problem.

Speaker 3:

That's the bigger problem. Now there is a correlation, right? So when it's running on GVP, on the video partner sites, that's where the majority of these problems are. So, yes, you could say yeah, if you're running on GVP, it's an entirely different thing than if you're running your ad on YouTube itself, Because the other thing you'll know is a whole bunch of humans do watch YouTube videos a whole lot, right? So if your ads are running on YouTube, it's usually less susceptible or immune to these kind of problems. The fact that so much of the ad inventory was running off of YouTube is where we now have all of these other problems, and that's what's documented by the research.

Speaker 2:

So talk about them to just the size of the research here, high level.

Speaker 3:

So Kristoff Frennichek, the researcher, he's not even from this industry. He's actually a computational biomedical researcher and that's his day job. He does this kind of stuff because he's a data analyst. He's a data enthusiast, if you will. He does all this on the side, for fun, and he's been collecting this data over the past three years. So there's data going all the way back to 2020 where it's continuously misrepresented. So you can even see in the network calls. All of these are declared to be TrueView ads. Clearly they're not. They're running off-site, they're running off-page, they're running muted, autoplayed and all that kind of stuff, and then literally ads back to back with no content at all. So this is the data and I think the total number was we've seen over 1,100 advertisers be affected by this in the data and multiple billions. So again, you can think of this as only a slice of the universe, because it's way, way larger than what his data was able to see.

Speaker 2:

Are you saying billions of ads served or billions in revenue?

Speaker 3:

Billions in ads served and because these are more expensive, because YouTube sells them as TrueView, skippable, audible ads. It's many, many millions of dollars involved, so billions of ad impressions and millions of dollars.

Speaker 2:

You'll cross over 1,000 advertisers at least.

Speaker 3:

Yeah, over three years, and really the point is tying this back to the fraud verification site. Okay, well, these fraud verification companies, it was their job to detect this, but they didn't detect any of it until an outsider Kristoff and his company Analytics, which is a one-person company actually detected this, and this has been the case over and over again. I'm kind of an outsider. I have my own tools. I've been detecting fraud for the last 10 years and I can show you side-by-side cases where the existing vendors are catching 1%. There's way way more fraud in there, because there's way more fraud than just bot traffic, Right, so you can almost think of that 1%. You know the 1% is not all the fraud there is. The 1% is all the fraud they can catch, and that's why it's time for the advertisers to really upgrade their tools so that they can actually see more details and then take more aggressive action against the fraud that's happening.

Speaker 2:

And why aren't they doing this? If these fraud vendors are not catching the fraud, that's the entire thing they're supposed to be doing. Obviously, whatever their methodology collection, whatever they're working on, it's not working. Advertisers must know that. Why are they continuing to pay them?

Speaker 3:

well, it gets to incentives. So I'll State something that that might be kind of hard to believe, but I've come across too many Advertisers who just have to spend their budget.

Speaker 3:

They're deathly afraid of not spending at all, right, so they keep telling me okay, well, if we detect the fraud, then you know it's gonna make it harder for us to spend all our money, and we don't want that. I mean, there's even been cases where we say, here's four million dollars refund and the advertiser doesn't want it because they won't know what to do with that money. So that's one kind of misaligned incentives. They have to spend it all. But the other thing is again going back to what we said earlier in the podcast.

Speaker 3:

The ANA Association of National Advertisers has kept telling them it's 1% fraud, don't worry about it, right? So again, it would be kind of counterintuitive for them to say, oh well, we actually think it's more than that, so we're gonna try to look deeper, right, if everything looks good in the reports, you're not gonna deliberately look harder and look closer and do more work than you have to, right? So it's kind of like counterintuitive. But I think by now you know I would say that the pandemic has actually rocked the boat, right? So prior to the pandemic, no one wanted to be the one to rock the boat, but because the pandemic already rocked the entire boat for everybody, I think advertisers are now, you know, knowing that money is gonna be tighter going forward, they're now taking a closer look, so I would.

Speaker 3:

I would say, since 2020, more and more advertisers. They've at least come to me, to you know, to ask for help to audit the campaigns, and that's where we're uncovering there's way more fraud than the 1% and, like we said, the tools were limited. So you know, we can't really go back and help them recover that, because those are the tools that they had in place before. But now that we have better data, we can actually help them prevent a lot of the money flowing to to the bad guys in the first place.

Speaker 2:

Interesting. Okay, so now let's go to the Google side. Here, google did put out Blog post rather short.

Speaker 2:

Yep and I'm gonna quote one sentence said quote a Recent report by a third party used unreliable sampling and proxy methodologies and made extremely inaccurate claims about the Google video partner network, and quote now what I find interesting here is they didn't say why it was unreliable sampling, they didn't say what they didn't like about the methodology and it only talked to the Google video partner network and they came out and said well, any advertiser can see if they're delivering ads on GVP, because it shows on that in their account. Yep. They also talked about Measurement from third-party companies Yep, you know it's validated. But I thought the most interesting thing here is that the data is being supplied by YouTube.

Speaker 3:

Yeah, there's so many things wrong with that. I wrote a whole article refuting their non-answers. I mean the couple sentences you read out. You know it's typical of these big ad tech platforms. Those are true statements, but they are also completely irrelevant to what we're talking about here. Okay, so they did say, you know, like Quoting, all campaigns run on YouTube. Okay, that is true. Right, maybe one or ten or a hundred ads ran on YouTube, but the other 99% ran on GVP, not on YouTube. You see how that's both a statement that's true but also misleading, right they're?

Speaker 2:

there's a word smithing.

Speaker 3:

Yeah, they're wordsmithing, they're misdirecting, they're just relying on advertisers being gullible enough or not asking further questions. So then just say, oh, oh, they just refuted it and so Google must be right in this outside Independent guy must be wrong, right? But I know for sure they haven't been in contact with Kristoff, they haven't asked him about his methodology, even though a lot of it is laid out in the research right. It's 200 pages of research with code samples, screenshots. I'd like to see Google actually address what's in the research right. And then some of those other examples.

Speaker 3:

They say, oh, we have third-party validation. Okay, well, google knows, because they set it up this way, that those third parties double verify Integral ad science and moat don't actually do measurements on YouTube. They're not allowed to run a Third-party JavaScript tag to do the detection. They are simply given data by YouTube, right, and they're performing calculations on the data and they're putting out the reports on the viewability, but it's still data that is supplied to them by YouTube. So, in essence, you know, as an outsider, I would say that's no different than YouTube continuing to grade their own homework, right?

Speaker 3:

they're just giving the data to some outside parties, letting them do some computations and then do the reporting on it. That doesn't mean it's independent validation. Okay, because none of them are actually measuring anything with a JavaScript tag on YouTube, right? They're simply given the data by YouTube.

Speaker 2:

Yeah, it's a charade. I mean, let's call it what it is.

Speaker 3:

Yeah.

Speaker 2:

They're controlling the amount of data that a company gets. It's not actual verification, yep. So they're trying to make YouTube look good and this is a problem We've seen in a lot of topics tied to anything around data Yep, which is, people will take a slice of the data. Yep you that is their argument without giving all of the data or allowing everyone to see all the data, and that's what I love about the report is the report is so detailed.

Speaker 3:

Yes, and basically Google's blog post did not address any of anything right.

Speaker 2:

It was glossed over using generic terms and basically PRBS, basically but now for this, because so many news outlets picked up on this, which I thought was great. A lot of times they don't. What is the next step here on Google side, because there's even some stories that some regulators are looking into this. Of the over a thousand advertisers that you mentioned were impacted, some of these are government agencies. So what is the fallout that potentially happens here? Obviously, that's your opinion of what potentially could happen, but what's going to basically come from this?

Speaker 3:

No, I actually believe this is happening. So the large advertisers they're now realizing okay, most of the ads running on GVP, the video partner sites, are crappy and they don't conform to the standards of a real TrueView ad. If you want a real TrueView ad, it's much better to limit your ads to running only on YouTube. And you do that by turning off GVP, the partners, right, and you can, like. Google said yes, you can see that in the data. So first thing is to go look up what proportion of your ads are not running on YouTube itself. Right, it might be 80%, might be 90%, could be 10%, right, so really, that's the first thing advertisers should do. But the other thing is Google also said a true statement advertisers could turn off GVP if they wanted to.

Speaker 3:

But herein lies another kind of misdirection. I set up YouTube campaigns right, and the option to turn off partner networks and to keep my ads only on YouTube doesn't appear during campaign setup. So there's no option for you to make that setting, to choose that setting Only after you set up the ad. You have to go back into each ad, click into the ad itself and then you see that option to select the radio button to say I only want my ads to run only on YouTube, right? So okay, that may not be fraud, that might be an oversight, but I'm kind of calling that a dark pattern, right? Because if the advertisers don't know to look for that and don't specifically want to do that, they won't even know that option at present.

Speaker 2:

Right, and you're not making it easy.

Speaker 3:

Yeah, so they're kind of making you're hiding it from most common people in their workflow, right, you have to do an extra step to go back in there and do that. So I think you know, like I said, the Google blog post contains a bunch of correct statements, but they're not necessarily. They didn't address the research and again, it's kind of misdirection because, yes, all of these things are true, they have third parties, right, they didn't say that the third party is simply given the data to compute the ability. They are not actually directly measuring things. And yes, advertisers have the option to turn off video partners, but you have to actually go back in there and find that setting manually and do that. It's not done for you by default, right? It's not obvious where it is.

Speaker 2:

Super interesting to see what the fallout is going to be. There's also if Google is going to address this at all.

Speaker 3:

Well, that's a lot of advertisers right, so the whole point is like whether they're going to offer any refunds for the misrepresented inventory, right? So in the campaign that was studied, that was cited in that research report, 84% of the ads were misrepresented as true view ads. Okay, so does that mean Google gives back 84% to that particular advertiser? And also, what's the rate of misrepresentation of ads in other campaigns? It obviously could be different, right? I'm not saying every campaign is 84% misrepresented, but obviously if the advertiser doesn't even know how to do the setting on YouTube versus not on YouTube, you would expect a large portion of those ads to be not running on YouTube, and that's where the problem is.

Speaker 2:

Right right.

Speaker 3:

So I think the practical step for advertisers is to first look up their ratio of ads running on YouTube versus on GVP and then going in and setting that setting because you're going to get good true view ads on YouTube itself, but you're probably not going to get that when advertising on GVP. So you might as well turn that off, and if more and more advertisers do that, that's going to significantly impact Google's revenue, which is intended effect. Right? You're wasting a lot of your money on these misrepresented ads that humans are not seeing anyway, so why waste that money? And let that go into Google's pockets.

Speaker 2:

Yeah, I mean it's going to have a direct impact on advertising revenue and I want to bring that up because, for listeners that don't know, youtube's ad revenue growth has now fallen year over year for the last three quarters in a row and, in addition, they have confirmed that. Youtube has confirmed that they're currently running a quote small experiment globally that urges viewers with ad blockers and able to allow ads on YouTube or try YouTube premium. So interesting, they're also now going after ad blockers, pushing consumers to sign up for YouTube premium at $12 a month, $120 annually, and, yes, you get other benefits like offline downloads, no ads, you get music, but I mean you got some interesting things all going on at the same time with YouTube's ad business.

Speaker 3:

Yeah, so that really tells me that they're desperate for revenue. And one of the key things, you know, when we talk about ad revenue growth, right, if those were based in reality, meaning human audiences, okay, think about every human in North America or maybe every human on earth, okay, and they're all watching YouTube literally hours upon hours a day. So the human population did not double in the last five years, right, and the amount of hours in a day did not increase beyond 24 hours a day, right. So there's a finite number of humans with a finite number of hours watching YouTube. So you know when all the investors in Wall Street laments the deceleration of ad revenue growth, of course the human audiences are finite. They don't grow that fast.

Speaker 3:

So revenue growth should not have been that rapid, right? So, yes, it can be rapid for a short period of time, but then it plateaus. It has to settle out somewhere. So in this case, yes, we're seeing the effects of reality, which is the human audiences. There's just not more hours of a day for humans to spend on YouTube, watching YouTube videos, so there's a limit to how many ads they can show, right, same?

Speaker 2:

thing has happened in TV. We've seen this with Espod, yeah Right. So in terms of the subs, do you think Netflix is going to get after the pandemic's over?

Speaker 3:

Yeah.

Speaker 3:

So all of those growth rates, I think there's been too much of a focus on growth rates and in fact I can tell you that's one of the key drivers of fraud because in the programmatic space, right, when the advertisers were kind of tapped out, right, they were buying from legitimate publishers like New York Times, Wall Street Journal, Washington Post.

Speaker 3:

All those human audiences have plateaued years ago, right, not only have they stagnated, but they're also decreasing because humans are spending time doing other things other than reading Wall Street Journal, right. So in those cases when those legit publishers were tapped out because of the insatiable demands from the advertisers for more impressions, that led them to buying more media from programmatic exchanges, because, magically, they could just create as many ad impressions as you wanted to buy, right, they could just use bot traffic to generate as many impressions as were demanded, and that's what led more and more of the dollars to flow to programmatic exchanges and away from legitimate publishers, right, and that's what we've witnessed for the last 10 years and that's why I said I've had to study the problem of fraud over the last 10 years. It all ties.

Speaker 2:

Wow, a lot going on. So, as we wrap up here, a couple additional questions for you. Would you say this comparison is similar to Facebook's previous inflation years ago of their ad metrics, because I've seen a lot in the media compare that, but I'm not sure it's quite accurate.

Speaker 3:

Yeah, it's not quite apples to apples. So what had happened was that the MRC, which is Media Ratings Council, is kind of the standard setting body for digital media. They had taken a number of years before they arrived at what was considered a viewable video. So it's basically 50% of the pixels of the ad in view for two seconds that's for a video ad and one second for a display ad. So prior to them arriving at consensus and publishing the standard for what was a viewable video ad, facebook had its own way of counting viewable videos. So it might be 100% of the pixels in view for a half a second or something. So, whatever the metric was, that was Facebook's own internal metric.

Speaker 3:

Once the official standard was published by the MRC I believe it was like 2017, 2018, then Facebook had to come into alignment with that in order to comply with the standard. So in that case, compared to the way they had counted viewable videos in the past, the past methodology counted many, many more videos as viewable. When applying the new standard meant far fewer would be considered viewable. So that was really the case. So it wasn't like Facebook deliberately falsified the metrics. It was simply the way they had counted was different than the new way of counting according to the standard. So when they came into alignment with the standard, all of these articles came out and said oh hey look, facebook had been falsifying their viewability metrics for years and years on their videos. You see how it was reported incorrectly. And Facebook, there was no malicious intent on their part to deliberately falsify, it was just the way they had counted was different than the way the standard said the video ad should be counted.

Speaker 2:

Okay, great, I appreciate that explanation for listeners. So the challenges we see in the space right now tied to fast A-VOD. Obviously A-VOD is interesting. With Netflix and others rolling out A-VOD services, we're going to see a lot going on there. Formats I don't think ad formats is the issue, even though some companies are coming up with new ones. Really, it's about the methodology and the metrics and analyzing what's being seen, what's being delivered to who. Do you see a point at some point where online video advertising will have the same trust standard that's really been long established in television? Is that possible?

Speaker 3:

I would say it could have been 10 years ago, but it's no longer possible because we've already kind of gone down the slippery slope of far, far too many ad impressions that can be explained by human media consumption, and the CPMs are already going to crap. I mean literally. We've seen this movie over and over again. Right, when we first started doing display ads, we saw fraudsters get in on it by setting up fake websites and using bot traffic to generate a whole bunch of display ads. Then everyone went on to video ads oh, that's much more popular, and all that kind of stuff. So then bad guys moved on to video ads. And then when we moved into mobile and mobile app, bad guys went there as well. They set up fake mobile apps and just cloned them by the dozens just to run as many ad impressions as possible.

Speaker 3:

Now we're talking about CTV ads or AVOD advertising, supported video on demand. Bad guys are moving there because the CPMs are 10 times higher than the CPMs for video ads. So bad guys go where the money is. So because we already gone down that path, I would recommend to advertisers you should not trust anything. It's almost like don't trust, always verify, Because if you're buying from these vendors. What are they going to tell you? They're selling you stuff. They're going to give you reports that basically say everything's fine, Don't worry about it.

Speaker 3:

So, just use this YouTube example. You've used the metrics supplied by YouTube for years and even if they say, oh, we have independent verification, it's still data supplied by YouTube, so it's not independent at all. So in this case, it's yet another example that, even if you have those third party verification, they still missed all of this. So the bottom line takeaway lesson is don't trust anything from the vendor that's selling you the wares, the goods, because essentially, you're relying on them to grade their own homework. And what do you think they're going to tell you? Right?

Speaker 2:

So it's not truly independent.

Speaker 3:

It's not independent, but I'll kind of leave a parting exercise or thought with the advertisers. I mean, I'm going to appeal the common sense. So this is an exercise I run in class with my marketing students Name 10 websites you visit every day, as fast as you can. You don't have to do that right now, but just think about that when we're done with the podcast. Or name 10 mobile apps that you use every day. And then finally name 10 CTV channels or providers that you've heard of. So every time my students try to do that exercise, I tell them write it down as quickly as possible. I give them one minute to do it. They can't even get past seven or eight, they just stop. They don't actually use or remember more than seven or eight sites that they use every single day or mobile apps that they use every single day, let alone CTV channels. Right, I mean, common sense will tell you most people spend their time on YouTube, netflix, amazon Prime and maybe Disney Plus right, A few more, but not more than 10.

Speaker 3:

So what the heck are these six million other Roku channels that are generating billions of CTV ad impressions? Common sense will tell you that doesn't make any sense at all. So that's why I'm telling you the advertisers really use some common sense to say, okay, this can't be real.

Speaker 2:

Yeah, it's a good way to think of it. So I don't know if you saw yesterday, the day before, a YouTube's Twitter account at Team YouTube. It was interesting. They posted a question. They said we're looking for feedback from YouTube premium users Just give us comments. And there's already hundreds of comments there. Interesting how many of the comments are tied to ads. I started reading through all them last night. It was why are you giving me two minutes of ads for less than a content that's less than one minute long?

Speaker 3:

Oh, interesting.

Speaker 2:

Why am I getting the same ads delivered over and over again? Now there were more issues around search and discoverability. Why can't I find content and why are you giving me the same content every time? But I thought it was interesting how many just average consumers are talking about the ad experience.

Speaker 3:

Yeah, that's on.

Speaker 3:

YouTube that's literally crowdsourced. So even if it's not a representative sample, you can say, okay, if there's smoke, that's where you should be looking for the fire. So I think a lot of that feedback from human consumers is very valuable. I mean, we've kind of seen the same thing with TV advertising, right? So out of a 30 minute sitcom, how many minutes do you think you can fit in ads? So the ad load just kept going up, where nine out of the 30 minutes were ads and 22. So two thirds of the 30 minutes were actually content, and it's actually worse today.

Speaker 3:

But that's something you can actually physically see, because in the physical world, like on TV ads, there's a finite amount of content. Right, it's a 30 minute episode of which you're taking up a portion of that for ads. But in digital, what I've said before is that we're dealing in just bits and bytes. So there has been virtually unlimited fake websites, unlimited fake mobile apps to generate an unlimited quantity of ad impressions for advertisers to buy. We are no longer limited by human audience size, right, or the hours in a day that they spend consuming media. So I think in digital that's why we've had a proliferation of the fraud problem that we really had not seen in any other channel, like TV, print radio out of home, things like that.

Speaker 2:

So final question for you here AI. I know some people listening and say, hey, well, okay, this is a problem, but don't worry, ai is going to fix it, because supposedly AI is a fix for everything in this world. Yeah, when people comment on this not even understanding how it works, how it scales. So does AI really short term or long term help with the ad fraud problem? No, make it worse.

Speaker 3:

No, and here's why. So for the past 10 years in my studying ad fraud, the bad guys have been using algorithms to set up websites, fake websites, right, you think, you know, think of algorithms. We didn't call it AI back then, right? 10 years ago, 15 years ago, but it was basically WordPress templates. They would scrape some content from various websites. You know the particularly hard hit ones are the celebrity, you know, paparazzi shots, the sports ones, the news ones. They would just scrape that or maybe even take an RSS feed from all those sites and just slightly remix the content. Then you have content for your fake site. Copy and paste some code on there and start running ads. Right, so that has been done for 10 years.

Speaker 3:

The fact that AI has come along I mean, the reason we hear a lot about this is really in the last year and into this year it's become publicly accessible to consumers, right, so previously it was just used by bad guys in academics or whatever, right, but now it's more public, more talked about. So that's why we think it's a new thing. But if you understand or have studied fraud as long as I have, the bad guys were doing this for years and years. So not only are they setting up fake websites? But the bots have become more and more advanced as well. So do you consider those algorithms or software programs to be AI? We didn't call them AI, but the bots are now mimicking human mouse movements, click patterns and that kind of stuff. So do you call that AI? So again, to me, that's why I said you can't get any higher than 99% fraud. It's already there. So AI is not going to make it significantly more and it's also not going to solve the fraud problem either, because those are algorithms.

Speaker 3:

And again, when you pit good guys algorithms against bad guys algorithms, good guys are always playing catch up. The bad guys have no constraints. They can just do whatever the heck they want. They don't follow any rules, they don't abide by any laws anyway, and so they can do whatever they want. And for the good guys, the challenge is knowing where to look and what to look for in the first place. So sometimes we might get lucky when the botnet goes down and we say, oh, how did the traffic just disappear? And then come back all of a sudden a few hours later. So when the bad guys mess up, that's usually when we can actually see the fraud happening. It's much, much harder to detect, and the bad guys are very good at staying hidden so that they can continue to make money while we're all thinking, oh, we've got all this traffic and all these ad impressions to buy.

Speaker 2:

And it's not real. Yeah Well, dr Fu, thank you. I mean this is awesome to get this much detailed information for listeners.

Speaker 2:

You know this is what you do all day, which I love. So for those listening, if you look in the notes section of the podcast, wherever you went, I put in the link to the actual report so you can go click on the link and see the report we're talking about, Dr Fu, best way for people to contact you. I know fuanalytics F-O-U analyticscom, but also on LinkedIn, Best way to reach out to you.

Speaker 3:

Yeah, linkedin. I'm on Twitter as well, so I constantly talk about the fraud issues, so feel free to reach out and connect on LinkedIn and Dan. I'll also send you my rebuttal to Google's blog post, because there's a lot of this stuff we did cover and I have screenshots in there to show they're kind of misdirecting the readers.

Speaker 2:

Good. What I'll do is I'll add that to the notes section as well, and I truly appreciate all the content you push up on LinkedIn. You know, for people who read my content or listen to me in my podcast, you know I'm a firm believer in facts over fiction. Yeah, I don't care if you have an opinion on something, right, but you know well, you're a doctor or scientist, right? That's not even used.

Speaker 3:

Yeah, ideal in facts. So if I don't have the data to back it up, I won't make a claim.

Speaker 2:

Right, and we need more of that when we're talking about anything tied to the video stack, ingestion through delivery, monetization, what's being delivered to whom, when on the right device at the right time to the right user and our industry as a whole shares far too much garbage data. It always regurgitates anything given to them without even double checking it. So I truly appreciate all the information you put online on LinkedIn, especially just you know, saying okay, here's the methodology, here's what's wrong with it, here's what you agree with Yep.

Speaker 2:

That's just. That's rare. We need more of it, so I certainly appreciate that. So thank you for joining me today. Thank you to the listeners. Many Questions reach out to me. Yeah, Go go see Dr Fu on LinkedIn if you want an introduction. Can't find them or not? Just reach out to me, Dan, at danrabercom. I'll be happy to make an introduction. Any other questions reach out to us anytime. Look forward to doing the next Executive Podcast Interview Series in another week. Thanks very much.

Speaker 1:

If you enjoyed the show, send it to a friend. Have questions for Dan or Mark? Click with them on LinkedIn at any time and be sure to check out Dan's blog at StreammediaBlogcom.

Dr. Fou’s background and work in the ad fraud market
Why ad fraud is purposely under-reported and is not 1%
What YouTube has been accused of and the data and methodology used
Impact is in the billions of ads across 1,000+ advertisers
Why Google's reply to the allegations ignored the issues
The potential fallout legally and financially for Google
Online Video Advertising Challenges and Trust
Parity ad measurement between linear TV and streaming isn't possible
Why AI won’t help reduce online video advertising fraud