Good morning,

I am excited to welcome Eric Seufert back for a Stratechery Interview about the ongoing transformation in digital advertising.

Seufert is the creator of Mobile Dev Memo, where he writes about digital advertising; Seufert also invests in startups in the mobile space through his fund Heracles Capital. Seufert previously worked for several mobile gaming startups, including as the Vice President of User Acquisition for Rovio, and was a consultant for multiple companies exploring the mobile space.

My most recent discussion with Seufert was last August where we talked about the impact of Apple’s App Tracking Transparency changes, and how the advertising world would look going forward; today we have a clearer view, thanks in part to Meta’s earnings yesterday, which had the interesting combination of declining revenue and a sky-rocketing stock price. We discussed where Meta’s strength will come from going forward, the lingering effects of the ATT recession, why EU regulations are potentially a bigger problem than ATT, and finish with an extended discussion of the DOJ’s lawsuit against Google and their web advertising product.

To listen to this interview as a podcast, click the link at the top of this email to add Stratechery to your podcast player.

On to the interview:

An Interview with Eric Seufert About Meta’s Earnings and the Google-DOJ Case

This interview is lightly edited for clarity.

Meta Earnings

Eric, it has been six months since you’ve been on Stratechery, which feels like a very long time and it’s been a very momentous six months, so it feels like the amount of topics we have to cover today is a little overwhelming. You are skeptical we are going to get to all of them, but we will see how we do.

Now, I usually try to conduct these interviews earlier in the week so we can edit the transcript and recording. However, I have delayed this interview until Thursday morning my time for publishing Thursday night, because we wanted to wait to see Meta’s earnings. And needless to say, that’s great because now I get to take a victory lap for writing Meta Myths when the company’s stock price was under $90. It’s poised to open up tomorrow up over a hundred percent since then. One of the myths was my contention that Meta engagement was actually growing, including in the core Facebook app, and that appears to be true. What were your takeaways above and beyond that?

Eric Seufert: Yeah, so first of all, thank you very much for having me. It’s always a pleasure to speak with you. It’s been a very eventful six months, it was a very eventful day to be honest. I was sort of glued to multiple readouts. We add the FOMC announcement today, which jolted markets, and then we sort of segued immediately into Facebook earnings. So okay, Facebook. Headline numbers, revenue down 4%, add impressions were up 23%.

Yeah.

ES: Average price per add down 22%. Right? So let me unpack that a little bit. So basically you see higher engagement. That’s Reels, that’s exactly the outcome that Facebook, that they are trying to engineer with Reels and it seems to be working. Now average price per ad down, well, they’ve said that Reels monetizes less effectively than their other placements, that they feel like they can get it to parity over time.

But my sense is, you start this process by just getting the feature to be adopted by users and that is demonstrably happening. People are using Reels, they’re spending more time in the app. And my sense is that’s the strategy. Now, they have many levers to pull in order to drive the monetization up over time. But I think the hardest part, the sort of more strategic victory there was just that people are adopting Reels, they’re engaging with it, they’re spending more time in the apps. DAU (Daily Active Users) is growing, they crossed 2 billion DAU on the Facebook app. And so I think in a lot of ways they’ve accomplished, or at least they’ve shown that they’re moving along the trajectory that they described for themselves a couple quarters ago when they outlined this transition.

I completely agree with you that the important number is the increase in impressions and the corresponding increase in price, because the price is also going to go down just because there’s more inventory available. And so that’s just the law of physics as far as this stuff goes. And a point I’ve written about for a long time is that I think this is the most misinterpreted number for digital ad companies and it’s basically been misinterpreted since Google’s IPO. For these large companies, growing impressions and the corresponding decrease in price is a very bullish indicator because they can increase the price over time as they improve their targeting and it becomes more valuable. But it’s not just that they have more room to grow now, but also, this crowds out competitors. Because when your Facebook opportunity is much more accessible and easier to get into because the price is lower, why would you not just go to Facebook instead of going somewhere else?

And again, we saw this with Stories, the exact same thing happened. We had this big increase in impressions, we had a decrease in price, Wall Street freaked out and you look back and it’s like, “What were you freaking out about? This was actually a big opportunity.” I mean, again, I sort of wanted to celebrate writing that piece, but writing that piece was super easy. Because basically it came down to this is just Stories all over again. And if anything, it’s even more compelling given how just engaging the short form video format is.

ES: So I think there’s a couple things. Okay, they’ve managed to increase engagement. Now, if you can increase engagement but keep ad load stable, then you are increasing the number of impressions that you’re serving. What you wouldn’t want to necessarily do, or probably the option that you would prefer the least, in terms of growing your ad revenue would be to increase ad load, right?

That’s a good clarification.

ES: Right. So I wrote a piece back on July 8th, and I proposed that there are four ways to increase ad revenue for an ad platform.

The first is to increase ad load. So that means given some session length, more of the time in that fixed session length is dedicated to the user interacting with ads relative to organic content. So you take that block of a session, let’s call it a minute of session time, and if ad load was 10 seconds of that minute, if you increased ad load, now 15 seconds of that minute are ads. Now you’re sort of saturating that session with more ads. And why is that not preferable? Because of churn. People are seeing more ads in the same fixed amount of session time and they’re saying, “Okay, there’s too many ads I’m going to churn.” And that will inevitably happen. If you increase ad load, you will inevitably see churn increase. Now to what degree? We don’t know, you’d have to test it, but that will happen. So that’s why it’s probably the least preferable.

The second is to increase reach, get more DAU, you get more users. So if you have more users and the ad load stays stable, session length stay stable, then you’re just showing more ads. Well, Facebook probably can’t do that. At 3 billion DAU across Family of Apps, it’s probably human constrained there, it’s humanity constrained.

Right, they’re growing a little bit, but they’re doing it in the least profitable markets. Almost all the actual DAU growth is in Asia and rest of the world, which just doesn’t monetize that well. That’s the case for everyone.

ES: Exactly. And that’s across the board; they’re just lower ARPU regions.

The third option is to increase the value generated by ads. Now, that was Facebook’s modus operandi prior to ATT. They had a steady drumbeat of products that utilized their data in increasingly innovative and sophisticated ways to just reach the right people with the right messaging and to sort of unlock that value just through better ad targeting. Well, that’s probably handicapped at this point, that’s probably not a path where they can extract much value from just given the restrictions of ATT.

And then the last is to just increase time spent on site. So I talked about, well, if you had a fixed session length and you insert more ads into that fixed session length, you’re increasing ad load. Well, if you increase the session length, you increase engagement, you increase time spent on site, you can put more ads there while keeping the ad load constant.

Now, that’s what they’ve done. They’ve increased engagement: people are watching Reels and they’re scrolling and they’re spending more time in Facebook. And therefore, there are more opportunities to show ads that keeps ad load stable. Now, they’ve also increased the value generated by ads. Let me get to that in a second. But then they also have increased ad load. They’ve thrown the whole kitchen sink at this situation, and my sense is it’s working.

So how have they increased ad load? Well, they had an advertiser new product showcase in October and they introduced a bunch of new ad placements: they have multi advertiser ad carousels on Instagram, where they actually have a carousel and there’s going to be multiple advertisers placing ads in the carousel. That’s new. They introduced post loop ads in Facebook Reels. So you watch the reel, you want to watch it again, they’ll show you an ad before you watch it again. They introduced ad carousels at the bottom of Facebook Reels. That’s new. Just more ads. They’re just pushing more ads into the product. And then in profile ads on Instagram: you go to someone’s Instagram profile and in some cases there might be an ad there. So they found more ways to apply ads to the existing service area.

So not only have the increased engagement, which just by definition increases the number of ads that people see, even if you keep ad load stable. But they’ve also increased ad load. Now, I said that they also have increased the value generated by ads.

AI & Black Box Advertising

This is what I get to, because I think the increased ad load is a good point, it’s a good call, particularly the stuff that’s happening outside of Reels. But there’s also an aspect where Reels had no ads. And so it’s totally justifiable to sort of increase ads there. But yeah, I’m very curious on this point. They talked about a 20% better conversion on their earnings call; that’s a number that was dropped a couple of times. Where is that coming from, from your perception?

ES: Well, there’s a couple things. So one is they’ve created new ad formats that my sense is, have the ability to convert better than other ad formats. So click-to-messaging…

Which is a huge business. $10 billion run rate, that’s very impressive.

ES: That’s an immediate connection with a retailer or just some advertiser. So you’re cutting out a lot of the intermediary steps that impact conversion. Because you think about conversion like a funnel. The sort of previous paradigm or the prevailing paradigm is someone clicks to go to a landing page, fill out a form, they click a button. Or on the app store, they go to the app store, they click install after seeing the screenshots, they install the app. There’s a bunch of points that could lead to drop off. Click-to-messaging puts you immediately in contact with the retailer.

But where I think they’ve made the most product progress here in terms of extracting more value from the ad system, that third point that I made, is in the application of AI. So we think about AI and their headline showcase of AI is really Reels. It’s classifying this content based on your interaction with video content and then finding more video content across this open graph, not constrained by your friend graph, but across all of the content available on the platform, which is a much bigger pool. And then finding the stuff that you would like. Now, they used to do that with ads. Now they do that with just the first party content, the UGC [user-generated content], because that’s first party to them and they have a much easier ability to do that than ads now as a result of ATT. But that’s immediately what most people’s mind travels to when you talk about Facebook’s utilization of AI to drive their business forward.

But there’s a hidden application of AI, or at least an application that’s not seen by the consumer, which is the application to advertiser tools. Now, what they’ve launched in I think the last six months of last year was a new tool called Advantage Plus. Now, Advantage Plus is like a black box automation suite. It’s where you just dump a bunch of assets into the tool and it’ll just go and test stuff. It’ll go and test lots of permutations of this audience that they’ve dynamically generated with this creative, and this permutation of different elements of a creative to find the right mix there. And I mean, you could say that’s AI. You could say it’s just machine learning, you could say it’s automation, whatever descriptor you want.

The things we can’t do are AI. And then once we can do them, they’re machine learning and just straightforward. Yes.

ES: Machine learning. Yeah, And then once it’s in a spreadsheet, just linear regression or whatever.

That’s right.

ES: So that’s something that’s not whatever the mainstream media’s going to report on because it’s boring, it’s on the backend, it’s the commercial element of their business. But that has actually been very successful. Because what does that do? Well, it uses all those tools, it brings all those tools to bear, all that sort of machine learning, AI capabilities. And it just takes all of the human error and human inefficiency out of the process of managing campaigns.

So whereas, me as a media buyer, I would log in a Facebook Ads Manager every day, I would take a look at the campaigns. I’d say, “This creative’s not doing so well. I’m going to stop using it. I’m going to kill that campaign. I’m going to kill those ads”, or “I’m going to kill that creative from the adset and then I’m going to reallocate some budget elsewhere.” Well, Facebook now is doing all of that on the fly in real time using Advantage Plus, using the system. Now that’s much more efficient. And now bringing all of those capabilities to bear on just the management piece. I mean, you could see where that would eliminate a lot of inefficiency and allow the budget to go farther and allow people to spend more.

Well, the other thing I think that is under-appreciated about this point is, number one, obviously this is the long-term solution to ATT. It’s like, “Look, you’re just going to have to trust us. We’re going to figure it out. We have all this knowledge and data and it’s incorporated in these models, and we will figure it out to the extent that you can’t as any individual developer.” But number two, by doing that, the more that it is a black box, the more it becomes a higher margin product for Facebook in the long run. Because you, as a buyer, you’re like, “Look, I’m getting lift here and that lift is worth it.” I don’t know, maybe that upside is 10 points and Facebook is saying, “Okay, we used to take five, now we’re going to take six points of margin.” But from you as an advertiser, it’s like, “Look, I’m getting conversions that I would not have gotten otherwise so I guess I have to accept it.”

I mean, this is the Facebook story in a nutshell. The reason why Facebook’s advertising is so valuable is because they deliver customers that you as any individual seller could never, ever find on your own. And the more that it’s a black box, the more that number one, they’ll continue to differentiate, particularly relative to their competitors, because I expect them to be better at this than anybody else. And then number two, the more margin they’ll be able to take on all those ads, because it’s a black box and you don’t know how it works other than that it does.

ES: Right. And there’s a couple other pieces to that too. So the charitable take on that approach — and by the way, Google does the same thing, they have PMax, it’s called Performance Max, and it’s the exact same concept just applied on Google. Now, I will point out that both of those tools are not new. They are rebranded, and they’ve been rebranded because they’ve been expanded to web campaigns, which again, were particularly impaired by ATT. But they existed prior to ATT just for app campaigns. So the Google version of it was called UAC: Universal App Campaigns. That was the black box total automation system. You just dump a bunch of creative assets in there and it goes and finds the right audience and creates whatever, all the different permutations of the possible ad creative. And on Facebook it was called AAA, Automated App Ads. So these were just for app advertisers. And now what they’ve done is they’ve expanded these tools to web ads, where the destination’s a website, not an app landing page.

And as an advertiser, you’re not choosing advertise on the web or advertise on an app, you’re just saying, “Get me customers.”

ES: Right.

And all that is sort of abstracted on the backend.

ES: Right. But the thing is, all of the dormant arguments that surfaced when these things were initially introduced for app advertisers have resurfaced. So now the charitable take on this, and what these platforms would say is, “Look, this takes the human out of the equation and humans are slow and humans are prone to error, and humans have a bunch of preconceived notions about what’s going to work. Whereas, the machine looks at this entire sort of opportunity space differently. They don’t have the preconceived notions on what kind of audience is right for your product.” You might say, “Well, you know what? I need to target women age 45 plus because that’s the right audience.” “Well, actually, we can target on every single sort of dimension across which you would construct an audience, and we’ll find actually the right people. Don’t worry about that. Don’t let your preconceived notions pollute the efficiency of your campaign.” And that’s what these platforms would say. “You’re a human, you’re slow, you’re judgmental, you’re emotional, you’re prone to error, and you go on vacation or you take the weekend off and we don’t.”

Now, the cynical take would be like, “Hey, wait a second. What we’re doing here is I’m setting some very, very top line campaign parameters. I’m telling you how much total budget to spend, and I’m telling you roughly what the unit economics should be, and I’m letting you figure all the rest out.” Well, how do I know that these products don’t go and they acquire a bunch of users that are really profitable? And then they offset those very, very high margin customers with very, very low margin customers that I never would’ve acquired on purpose. And then when you blend that all together into a cohort, it averages out to meeting the standards of my unit economics. But there’s a bunch of bloat in there, there’s specific users that I never would’ve tried to acquire because I would’ve seen that they’re low margin or I would’ve seen that they’re unprofitable. Now, in that case, what the cynical argument would be is, “Look, you’re just trying to maximize my budget spend. You’re not trying to maximize the profitability of every single-

Yeah. My efficiency spend. Right.

ES: Exactly. But the thing is, you could say, well, both of those things are true, and on net I’m still better off. And I think that connects to the conversation we’ll probably have later about the Department of Justice’s Google case…

I was going to make that connection. You jumped in and stole it! We are on the same wavelength with that, because that’s exactly some of the stuff the DOJ was talking about with Google.

Before we get to that though, I do want to talk about this black box sort of bit. You talked about how you can go in, you dump in all these assets. You also made the point that the application of AI, everyone looks at the consumer side, but it’s actually much more compelling in many respects from the advertising side.

ES: Right.

I’m very curious. One of my theses here is this is where generative AI is actually pretty compelling. So generative AI, everyone thinks about the consumer applications, you’re producing images, you’re producing text, whatever it might be. But why would we not get to the point where — I mean, what are the challenges with Reels monetization, for example? Part of it is that it is more difficult to create a video than it is to create an image, just as it’s more difficult to create an image than it was to create text. Now, of course, humans have come along and gotten better at all those things, and they could utilize generative AI on the production side.

But what seems super compelling to me, and not possible yet, but will be soon, is where the black box also includes asset creation. And all you’re doing is just saying, “Here’s an app, here’s the goal”, or “Here’s a an e-commerce site, here’s the output.” The AI actually goes in, it generates ads, it runs tons and tons of AB testing to find the ones that work. Then it scales them up. I mean, is this fanciful? It seems like the next step down this road here.

ES: That’s the next step that will happen in the next three to five years. I wrote a piece about this a couple months ago called Abandoning Intuition, and that’s the exact point I made: “Look, just creating ad assets as a human endeavor involves a lot of preconceived notions around what’s going to work and what resonates.” And even just coming up with an ad concept ex nihilo is actually really, really difficult. If you just sat and said, “Okay, I want to come up with an ad. It’s got to be like a 15 second video.” It’s hard without any sort of external prompt. This is really difficult to do, right? So normally what you do is you say, “Well, what’s worked in the past?” And then, “Let’s make a variation on that. Let’s change it a little bit, because we know that works,” right? But that’s generally the process.

Now, a lot of times creative teams will say, “Okay, no, we’re going to reserve some time every week to just come up with totally new concepts.” But subconsciously, those are informed by ideas, right? Now, what you can do with Generative AI, and that was the point of my piece, is that you avoid all of those preconceptions, and you also avoid some of these sort of specious associations that you might come up with as a human being. It’s like, “Well, you know what? This video had a dog in it. I bet people like dogs in these ad videos, so we put more dogs in.” And actually, it was because the color emitted by the sun on the lake in the background was a certain hue of yellow, and that would capture people’s attention. That never would’ve occurred to me. But the thing is with Generative AI, they can just sort of interpret that. These systems can interpret more dimensions to the asset than probably we can consciously recognize. Now, maybe not subconsciously, because obviously, we subconsciously recognize them, because that makes people click, right?

Yeah.

ES: But how can you on the input side, make those associations? It’s really difficult, right? But the machine can do it because they’re looking at things that are not just based on these sort of immediate visual cues, right? So my point in that piece was that has to be the next step, because there’s an immense amount of commercial value to be extracted from disintermediating human intuition and asset creation. Just putting a machine there, letting it interpret what worked and what didn’t, and coming up with totally new concepts ex nihilo, and then just pushing them out into the ether and determining what works.

Right. So the way I interpreted your piece is how Generative AI can be a tool for advertisers in sort of creating the assets in the first place. That makes sense as the next step, but isn’t the end game here that just Facebook will do it for you?

ES: Yeah, and then integrating those things. That’s the end game.

Yeah, that makes sense.

The Duopoly

Well, so this ties into one of your other theses is you wrote last year which is that the duopoly is over. I’m curious how much of that is a function of the Google-Facebook duopoly actually being over in sort of the space that it competed in, versus brand advertising in particular moving onto the web from TV in a truly meaningful way?

The one company worth pointing out here is Amazon, just in part because ATT was so impactful on eCommerce sales in particular. If you can go to Amazon, it’s all first party data, then that’s an obvious thing to do. But by and large, and maybe I’m wrong here, but my general thesis is when it comes to the sort of direct response space that Google and Facebook, on which that duopoly that was predicated, they had a dominant share of advertising because most of the advertisers was direct response and they were the best at it. It would seem to me that yes, there’s way more ads online. Everyone’s a first party ad platform because ATT dictates that. But when it comes to exactly what works on Facebook and what has always worked on Facebook and what works on Google, is some of that dispersing, or is there just more ads online in general?

ES: Well, both. I think some of that is dispersing, right? So Amazon, a lot of that ad spend is direct response, right?

Right. Amazon is the one big exception. And by big, I mean absolutely massive and gargantuan. So there’s a very large sort of hole in that argument.

ES: Right. Well, the thing is their ads business has been growing at an impressive clip for years. I mean, I called the sort of combined Google, Facebook, Amazon advertising segment the triopoly in 2019 or something, because it was so obvious that Amazon was going to grow into a sort of bonafide competitor to those companies.

And ATT’s the best thing that could have happened to them.

ES: An accelerant, right? But the same with Apple. I mean, that’s not brand spend that they’re capturing, that’s direct response, right?

Right.

ES: So I think what I would say is that, look, on the retail media side, I think it’s really difficult to parse out what is brand and what is DR. Especially for a lot of those platforms, it’s all kind of mixed together. Now, if you’re using that space, if you’re using those impressions in a performance manner, meaning I have a measurement apparatus set up, I have some kind of interpretation mechanism that allows me to guess or estimate what my performance is when I spend that. You don’t really need to differentiate the platforms between brand and direct response. Obviously, direct response is a very specific type of ad strategy, but you can use them in a performance way.

So I could be doing performance ads, right? And I hope people are spending their dollars this way. But I could be running ads in a way where I am measuring the impact, I am trying to estimate the lift, I am using tools like media mix modeling and incrementality measurement to determine what my dollar is actually buying me. And then there’s no real differentiation. It’s not just me plowing money into an ad system to yield reach or awareness. That’s what you typically think of when you think of brand advertising.

But even the big advertisers like P&G and Coca-Cola or whatever, they have teams of data scientists that are trying to measure the impact of that spend. They’re just doing it on a much sort of slower feedback loop. It’s usually quarterly, where they’re looking, “We spent this much money in Southeast United States, and these are the retail sales numbers, so now let’s try to sort of interpret which channels had the most impact and provided the most revenue.”

But I would say that a lot of the retail media stuff probably is capturing the direct response budget that fled from Facebook. That’s why I think they’re spinning these up. I think when you look at just the laundry list of companies that are spinning up these ad platforms, they have to be doing that in response to ATT because that money is looking for a new home. Now, we’ll see if Facebook’s able to recapture some of those dollars, and I think they will.

Right, that’s my big question, because particularly when you go back to this idea of they have this increasing inventory, by and large thanks to Reels, prices are down, their conversion is improving because they’re rejiggering their ad engine, and they’re, in my estimation, the best place to do that well going forward, then isn’t the duopoly thesis sort of still in place, and maybe a lot of these first party ad efforts are maybe not going to work out as well as the folks putting them up hope they would?

ES: Right. But just to answer your first question, is the duopoly over because the sort of market has expanded into new types of advertising that they’re just not participating in, and therefore, they have no share in it, and therefore, their share of this sort of newly defined market smaller than it was before? Yeah, that’s probably part of it. But I think also some of the dollars that they were attracting prior to ATT, they’re no longer attracting, right? They’re finding homes elsewhere. Yeah, I don’t want to overstate that data. I mean, Insider Intelligence published that data. They’re a very credible organization, but-

Which data?

ES: The data that showed that combined, Google and Facebook have less than 50% share.

Right.

ES: But I don’t know. I would want to see a couple of reads that say the same thing before I sort of start chiseling the gravestone. But I would just say the bull thesis for the sort of incumbent platforms would be that they have a lot more levers to pull to improve efficiency and to just improve engagement and the sort of ads experience on their platforms. There are things that they didn’t really care about before because the primary vector of growing revenue was just that value extraction. It was the sort of deep machine learning for audience segmentation and targeting. And now it’s like, “Okay, well, actually, we have to focus on the product to get people spending more time in it,” which is probably net better for consumers. Just as a product sensibility. And then on the back end, we just have to take more of the human element away, and do more modeling and do more sort of estimation and probabilistic forecasting and that kind of stuff.

What that affords you is a moat. The ability to do that is a moat. Now, if you can do that really well, you get to a point where you say, “Look, a lot of these outcomes are modeled. I can’t tell you exactly where this dollar came from, because I’m modeling it,” right? Now, if you combine my platform with another platform that’s similarly doing modeling, there’s probably going to be a lot of overlap there. You’re probably going to be doing a lot of double counting. How much are you spending on my platform and how much are you spending on their platform? And is it worth it to keep both, knowing that there’s going to be some wastage on the double counting? Or would you rather just spend all your money with me? Because if you spend all your money with me, you know there’s no double counting, right?

Now, to that point, prior to the holiday season, and this is why I say that Facebook threw everything at last quarter, Meta reintroduced 28 day click attribution reporting, which is modeled. They had this prior to ATT, but it was deterministic.

They knew for sure that you clicked.

ES: Right. Well, they got rid of it. They got rid of that post-ATT because, well, we can’t know 28 day. The restrictions on web ads based on ATT require us to only give us the opportunity to observe conversions within seven days. Now, that’s an Apple restriction. That applies to Safari, not on the apps side. The app measurement framework is SKAdNetwork. On the Safari, it’s Private Click Measurement. Now, what Facebook did in order to comply with ATT, because you might remember, when Facebook first responded to the announcement of ATT, they said, “Okay, well, we’re just not going to show the prompt. We won’t collect the IDFA. That’s all we need to do.” And then they came back and said, “Okay, well, Apple told us we have to show the prompt.” And also, this applies to web ads, not just app ads, and therefore, the IDFA is kind of irrelevant in that sphere. So what they did was, my understanding is, they said, “Okay, look, if we build a framework for measurement of web campaigns that’s modeled on Private Click Measurement, can we deploy that for web campaigns?” And Apple said, “Yes, that would be ATT compliant.” That’s my understanding. I’ve heard that kind of secondhand. Maybe it didn’t happen exactly like that. Maybe they just guessed.

So they reintroduced 28-day attribution windows in late October. Which means, okay, if I’m an eComm advertiser, Private Click Measurement only lets you observe conversions for seven days. So if I’m an eComm advertiser, prior to ATT, the gold standard for the attribution window for my eComm campaigns was 28 days. That means I’m going to deploy a dollar, I’m going to say, “Facebook,” or whatever ad platform, “anything that this person does within 28 days from the click, I want you to attribute that to the campaign, to the campaign that I reached them with,” right? That’s going to be my sort of true understanding of the attributed value of that campaign.

Now, when you say, “Okay, well, we can’t offer that anymore. We can only offer seven days,” that’s one-fourth of the timeline. Now, I don’t know if it’s one-fourth of the value. The value curve might look logarithmic. But it’s probably substantially less. You’re counting substantially less attributed value on the seven days than you are the 28. So if the person that comes and they buy something eight days later, well, you have no way of observing that. You have no way of attributing that to the click that landed them at your website. So Facebook reintroducing that, even if it’s just a modeled reporting number, it’s not actually an attribution setting, but it’s just reported and it’s modeled, automatically gives you some sense of how much money you’re getting after that seven days, which is pretty much the basis on which you modeled your monetization curve. And it just gives you more confidence to spend more money, because it’s showing, “Hey, actually, with the seven day attribution window, you’re getting 2X ROAS. You’re getting 1.1X. And now with the 28 day, we’re attributing more revenue. All the revenue that happens from day 8 to 28. And now it’s 3X ROAS.” So what are you going to do? You’re going to spend more money.

It’s also probably a little bit of playing with fire from the Facebook perspective, because you can understand there’s probably an incentive to maybe get a little bit less conservative in your modeling and to say, “Look, we know historically we were delivering this return on spend. Our estimates are probably a little bit too low. You could tune that up a little bit.” Now, this is speculation. I don’t know if this is happening. But you can definitely see there being an incentive to be a little bit more aggressive in the modeling. And in the long run, it will shake out because people, at the end of the day, actually measure how much stuff they sold, right?

ES: Yeah.

But it will be interesting. I’m curious to what extent that ended up playing a role in this throwing everything against the wall sort of approach as well.

ES: One more thing and then we can close the book on Facebook earnings. SKAdNetwork 4.0, right? That rolled out last year. That rolled out with iOS 16. But there was a bug in 4.0. So SKAdNetwork 4.0 is the upgraded version of SKADNetwork, which is the measurement framework for app campaigns, but there was a bug, so it actually wasn’t functional immediately when 16 launched, iOS 16. It only really became functional kind of end of the year.

Now we’re just starting to see the SKAdNetwork 4.0 postbacks flowing through the advertising ether. But who can do more with that? That’s more signal. SKAdNetwork 4.0 is more signal, right? It’s more data to latch onto. It’s more data to feed into the machine. And who’s got the best machine? I think that could be another bull case to be made for the biggest platforms. In fact, when SKAdNetwork first came out with ATT, Google said, “We’re not going to look at the conversion values and the postbacks. We’ve got our own system for doing things. We’ll use that for install reporting, but we’re not going to look at the conversion values.” Now they’re saying, “We will. We’re going to ingest those conversion values and use that for ads optimization.” So if you think about, okay, more signal, that’s great. We kind of had a level playing field or a little bit more of a level playing field when all the signal was taken away. Who benefits most when you add a little bit signal back?

Right. Well, just on the Apple point, one of my big questions around ATT the whole time is isn’t there some extent to which Apple’s shooting themselves in the foot here? Because at the end of the day, particularly when it comes to app installs, why would you want to get an app installed? So that you can drive in-app purchase. Who benefits from in-app purchase? Apple does, right?

You wrote in an article a few weeks ago that a maybe counterintuitive way you’re a little bullish on mobile gaming, despite the fact it’s gone through this very difficult transition because of ATT, because Apple is going to start feeling the pain and is going to start delivering better measurements and better signal because it hurts their bottom line. And I guess SKAdNetwork 4 was maybe the first version of that, but it sounds like you expect more signal coming from Apple, not less.

ES: My sense is they want to provide tools that are functional and that contribute to the prosperity of their ecosystem. They don’t want to hurt the developers. There’s a cynical take and there’s a generous take, and I think the reality is probably somewhere in the middle. I’ve articulated the cynical take enough, so I’ll spare your listeners.

But I would say the generous take could be, look, this sort of rapacious behavior in the ad ecosystem had to stop, and there’s been no progress in national legislation to rein it in. So we had to do something. Now, yeah, we benefited from that. We’re a corporation. We do things that benefit us. Sue me. Or well, don’t sue me. But of course I would do something that benefited me. But we believe it benefits the users too.

And maybe they were a little bit overzealous with that, or maybe they were a little ideological with that and they didn’t think through even the first order sort of damage would look like, but especially not the second order. And now I think they’ve sort of awoken to that. My sense is there has been a paradigm shift or a mentality shift within Apple around what they need to do to make developers successful.

Well, I mean, it’s one of those things too where you have to be wary because I want to believe this is true because it does fit my long-term thesis, which is that look, ATT was a tremendous blow, but the shift from deterministic to probabilistic is going to favor the companies with the most data and the most capability of understanding that data, and Meta is sort of top of the list there.

I mean, just one more aside before we move on, because we have a lot of stuff we wanted to get to. One thing I thought was really interesting in the call, to your point about product improvement, is they basically said, “Yeah, we could grow Reels more, but we’re going to actually slow down for a while because we want to improve our monetization before we do,” because it actually is quite costly to have people in Reels. The longer they’re in Reels, the less they’re in the feed, the less they’re in stories and they’re seeing less valuable ads. But I just thought that was interesting that they feel like they have such a handle on it now. Those dials are in place. They know what to turn up and down.

The other bit of throwaway line that I thought was super interesting is the other thing that we’ve talked about in the response to ATT is the importance of moving conversions onto Facebook itself, because then it’s first party data and then you can retain the deterministic measurement. And I can’t remember who it is, but said, I think, in answer to an answer that we found product market fit with shop ads, and it was a small test, but we feel very confident that’s going to grow. To the extent that’s true, that’s a big deal, because eCommerce was still down. They called that out as one of their most important categories that is still decreasing. But if they can actually crack that nut of purchases happening within Facebook so that they can do deterministic sort of measurement, I mean, that’s a big part of the bull thesis also.

ES: Yeah. Well, I think it was Javier Olivan, the COO. Yeah, so that was my content fortress thesis, that you’re just going to try to pull more of these conversions into your own first party environment, where you can measure them to your heart’s delight. And that’s not in violation of ATT. We could talk about why that maybe isn’t true if we get to the European regulatory landscape.

Digital Ads and the EU

Let’s get to Europe. So there have been three recent decisions in Europe that seem like a big deal for ad targeting. One was Meta can’t use a contractual basis to get user agreement for ads, even if its first party data for targeted ads. They have to ask explicitly. There was a real conflict between regulators, where Meta’s regulator argued that it was fine. After all, that’s the price of getting the service, which seems reasonable to me. But obviously, other countries disagreed and they ended up losing, they’re appealing.

Then there is WhatsApp’s use of contract to get permission for first party data for general analytics and security. This one blew my mind. It’s not even about ads. You can’t even get data about whether or not the app is crashing, or whatever it might be.

And then the French regulator, which is an important point here, this was not the EU, but it highlights the complexity in Europe, said that Voodoo Games cannot use the IDFV, which is the unique identifier for vendors. Because one of the assumed ways that you get around ATT is owning lots of apps: both of us predicted this would drive a lot of mergers, because if you have lots of games that are owned by the same company, you do get to share data across those games using the IDFV, and now they’re saying you cannot do that.

So, I mean, it seems pretty apparent that all these are problematic, but whatever. It seems like this is now the new reality, at least in Europe. What’s your high level take about this and the implications going forward?

ES: Well, first a caveat, I’m not a lawyer. I’m not a lawyer who specializes in Europe.

We’re going to talk about the DOJ too, so let’s put that caveat out there. These are not our lawyerly takes, not case takes, just the implication takes.

ES: Yeah. So let’s start with the Facebook Instagram case with the Irish DPC. So the Irish DPC originally, there’s six legal bases for processing user data under the GDPR. And so what Facebook was saying was, “Look, we put in the terms of service, if you agree to the terms of service that means we get to use this data.”

And to be clear, this is not third party data.

ES: Not third party data.

It was first party. I watched this video on Reels and that contributed to me seeing a particular ad.

ES: Exactly. And data, we’re going to utilize data artifacts that are the result of you engaging with our product for ads targeting. And we’re allowed to do that under the contractual basis because people agree to the terms of service and it says that in there. And the Irish DPC said, “Yeah, that’s probably fine.” And the thing is, with the GDPR, if there’s dissent from the other DPAs [Data Protection Authorities], the other EU DPAs, then it goes to the big board, the board that was formed to arbitrate disputes under GDPR and also to enforce GDPR. It’s the European Data Protection Board, EDPB. And they said, “No, we don’t agree with you, Irish DPC, and here’s a binding decision that says you have to fine Meta and they cannot use a contractual basis anymore for ads personalization using first party data.”

So then the Irish DPC said, “Okay, we’re going to sanction Meta, but we disagree with you, EDPB. And, by the way, we might sue you because we think this is an overreach of your powers granted by the GDPR.” Facebook’s going to appeal. Irish DPC is threatening some legal action, some pushback. We’ll see where that lands.

Just big picture, I mean, it just drives me up the wall. Why are you entitled to use Meta services? Again, there’s a reasonable debate to be had about third party data. Heaven knows we’ve had the debate plenty of times over the last two years. It’s just conceptually mind blowing that, look, the price of using our service is that your data is used so we can show you ads that are more compelling to you. It’s weird. Basically, the long and short of it is you have to deliver your service on terms that you don’t want to, which is, I don’t know, it strikes me as a problem, but that’s neither here nor there, I guess.

ES: Yeah. So there’s a couple things. So, with ATT, we saw, and I wrote this long piece, The ATT Recession, and I think you’re seeing that play out still this quarter. I mean, Facebook’s revenue is down 4%. I mean, that’s pretty significant.

I was going to ask you, did Facebook’s results validate your thesis or dis-validate it? It sounds like from your point of view it validates it. Just to summarize, the ATT recession is the argument that, look, everyone sees these digital companies that are down, but actually it’s some combination of COVID overhang and the fact that ATT was actually a bigger deal than people realized. And most of advertising’s doing fine. It’s just the ATT-related stuff that’s down. I think that’s right. I agree with you, but are you still on board post these earnings?

ES: Yeah, I think the SNAP results validate it. I think the Meta results validate it. And down 4% is a lot, right?

Yep.

ES: Now, it’s up 2% on a constant currency basis.

But that’s also not very much historically speaking.

ES: Right. And even then, well, the dollar was down, so would they have gotten that? But you can make the argument both ways. But nonetheless, I think the results validated my thesis.

What markets were reacting to was, hey, Advantage Plus and Reels actually are more effective than we thought because we had priced in a 6.5% decline in revenues. So they beat on a very low bar, basically. So I think, and what makes me nervous now is, you see these companies making progress with the remediations against ATT and you say, “Okay, maybe we’re past the low point here. Maybe we’ve passed the nadir and there’s a bright horizon ahead of us.” And then the European regulatory system is saying, “Hold my beer. No, we’re going to go after first party data.” And the thing is, in the Meta case with the Irish DPC for personalized advertising, a lot of people are saying, “Well, there’s another legal basis they could use, which is legitimate interest. They can say it’s in the user’s legitimate interest that we target ads to them as a stipulation of using the service.”

The thing is, TikTok tried that. They tried to go down that route. They were going to change their terms of service and their privacy policy to utilize legitimate interest. And the Italian DPA stepped in and said, “No. That’s against GDPR.” And then TikTok consulted with the Irish DPC, which is their one-stop shop regulator under the GDPR because they’re headquartered in Ireland, because Ireland tends to have a business friendly environment and privacy friendly I guess too. And the Irish DPC was like, “No. You shouldn’t do this. I don’t think that this is defensible.” So I think when a lot of people have looked at the Facebook case and said, “Well, they’re just going to change the legal basis,” I’m not so sure they have any other legal basis to use.

It’s not just that, but the European Union is not going to tell you whether or not it’s okay. They’re going to wait for you to do it and then they’re going to fine you and it’s retroactive to when you did it. So there’s a lot of risk attached to even finding out if this is possible.

It seems apparent to me that, basically, there’s just going to be this carve out of Europe. You should probably be pricing in decreased monetization there. I have a hard time seeing regulations of this type coming in the U.S. where we’re going to say, “No. You’re compelled to offer your service at terms you don’t want to.” That’s very easy for the EU to say when they have no skin in the game as far as these companies are concerned. I have a hard time seeing that happen in the U.S. And so maybe you’re just going to see this real big bifurcation in terms of revenue per user in Europe versus the U.S.

My question, though, is how does this play in with the AI stuff we’ve talked about with this data? I mean, does that just cut it off completely or you’re just going to take contextual advertising to these crazy places? I mean, how are they going to work around this?

ES: Well, my concern is that this conversation just becomes deranged when you introduce AI to it. Because who owns the inputs to the AI model?

Exactly.

ES: And those are the arguments you’re already starting to see. So my fear here is that the Internet is facing the nuclear moment where you have activists and special interests that prevent the right thing from being done. The right thing is let’s have the economic point of view and let’s have this philosophical social point of view, and let’s merge those two point of views or at least have them consult one another, and let’s forge a path that finds some workable, passable middle ground where we feel like people’s data is not being exploited, their privacy’s not being exploited, and yet we’re able to cultivate a productive economy around this. And my sense is the activists and the special interests can win out here and this prevents any of these technologies from even taking root in the first place.

Yeah, I agree. I know our listeners get annoyed because you and I are on the same page on this in that we both see tremendous value and broad gains from these easily accessible ad units that, all along, one of my big theses and things that I want to see on the Internet is small and medium sized businesses that have the whole world as their market, such that they can sell something unique, and that’s great for consumers and it’s great for the economy, it’s great for the businesses, but they have to be able to find the customers. And Facebook in particular has always been best at that. And I don’t know, it’s probably not productive to go back and forth because I think we feel the same way. I mean, okay, the third party debate I can get. I think the damage is overstated, but you in particular always been very good at saying, “Look, there were real abuses happening here.” And you’re right about that. To go after first party data, I mean, it boggles the mind.

The DOJ’s Google Case

Let’s finish with the DOJ Google case here. You talked about the concerns about the black box, which is that Facebook might be getting you a bunch of super high value customers and a bunch of crappy customers and averaging it out. That’s explicitly one of the things that Google was accused of doing, along with a whole host of other things.

ES: Right.

You wrote an excellent article summarizing the case. I thought the complaint was pretty good, all things considered, but your report is much more digestible. However, you said at the top, you’re not expressing your opinion. You’re just summarizing it. So here is your opportunity. I want some opinions.

ES: Right. I think my sense is that if you think about what the DOJ’S thesis is, it’s that Google used its end-to-end ownership on the publisher side, on the demand side, on the advertiser side, and with its exchange, to suppress competition, to prevent other companies from being able to compete. And the things that it was able to do, if you look at Project Bernanke, Project Poirot, Project Bell, all of those things, it was only really able to do them because it had this end-to-end ownership of what the DOJ calls the ad tech stack. I think that’s probably demonstrably true. I think, at least, it’s not a prima facie absurd argument. I think you could argue either side of that, but I think the argument in favor of that is probably pretty strong. And we’ll see how it ends up.

Where I think the DOJ is going to run into trouble is on two points. The first is that it employed a little bit of sleight of hand in framing the problem. So what the DOJ did, and it did this at the outset when it was talking about the DoubleClick acquisition, it portrayed supply as chasing demand.

Just for the people listening, demand being the advertisers and supply being the inventory where the ads are placed.

ES: Exactly right. And so the point that they made was they had all this demand and they used the demand to capture the supply, to lock in the supply. But that’s just not how ad markets work. So the application of an advertising model in the first place, the ability to use an advertising model presupposes the existence of supply. I can’t sell ads if I don’t have some surface area to put ads on. Now, what the DOJ did is it ignored the fact that, first of all, just that law of nature of ad markets, which is that you aggregate supply and demand follows. You don’t aggregate demand and supply follows. You need supply to have any ads product in the first place. Now they’re saying, “Well, what Google did was it aggregated demand, and by aggregating the demand, it was able to lock these publishers into contracts, into using their tools in ways that worked against their best interest.” But that’s not true.

What Google did was it aggregated supply first and the aggregated supply attracted the demand. It’s just that their aggregated supply was Search. It had the supply which was search results. That was the product and that brought the demand over. And so what the DOJ is proposing is that it unfairly aggregated demand because that demand was the result of abusing monopoly power by locking publishers into its publishing tools. But it’s ignoring what happened before that, which is that it aggregated the supply, which brought the demand to them.

That’s exactly right. I think this is the problem you see with so many of these antitrust arguments, is they get the cause and effect backwards.

ES: Right.

Because the entire premise of antitrust laws as they were passed 120 years ago was about controlling supply. There’s scarcity in some asset and everyone wants access to it and you control how to get it, then you can charge extra high prices. But the reality is, the reason why publishers all want access to Google Ads is because Google has all these advertisers. But to your point, you have to follow that back to Google Search being the original and still best digital advertising product. And there’s this weird lock-in chain that ignores the fact the source of Google’s power in this value chain is the spot where there’s the least lock-in of all. Advertisers are not locked into Google Ads at all. They use it because it’s convenient and it works. And then Google will argue, maybe this is your point too, “But we’re not compelled to deliver that to anybody else. If our advertisers want to go to another exchange, they can go to another exchange,” which is absolutely true. And if that’s the case, then what are we doing here?

ES: Right. So I think it’s just that they start the complaint off with an argument that I think is fundamentally specious, because the demand was acquired fair and square because they had this product, Search, which worked really well, and they built an advertising business on top of it and that brought the demand. The demand was acquired fair and square. And then the third party inventory, which by the way, and I put that out in the piece, something that the DOJ probably should have said in the complaint was that this part of their business has been shrinking since 2008. Since the company went public, the network side of the business has been shrinking to where now it’s 15%. We can talk about that when we talk about the remedies in a second. But let me show you, the second problem I think the DOJ is going to have is it’s going to be really hard to prove harm to customers.

So the complaint makes really clear that the publishers, they saw increased fill rates and increased CPM payouts. And people are going to push back on this and get mad that I’m saying it, but that’s what the complaint says. That’s what the DOJ says. Now, what the DOJ says is that while the only reason the publishers were getting these outsized outcomes is because actually they were being subsidized by hostile terms on the advertiser side. Now what you say is, “Well, let’s talk about those terms. What does an advertiser do when they come to a platform that sells ads?” These ads don’t have price stickers attached to them. The advertiser says, “Here’s my bid. Here’s the maximum amount of money that I’m willing to pay. Now, if you can find me customers for that amount of money or impressions or clicks or whatever, I’m getting some margin out of that. I could potentially be indifferent to what you do behind the scenes. Just here’s the max I’m willing to pay and here’s my max per user, per conversion at the unit level, and then here’s my max budget. You fulfill that however you want. But every single time you garner a conversion for me that is under the max price that I told you, or even if it’s at the max price I told you, I’m making some incremental money.”

And so I think what people are trying to portray this as is that the counterfactual is, well, if Google would’ve not manipulated these margins, the advertisers would’ve been getting more conversions for their budget. But that’s not the counterfactual. The counterfactual is, what if Google just didn’t make this available at all and the advertisers are starved from those incremental conversions? I think if you get a performance marketer on the stand in this jury trial, they would say, “Okay, what would you have done if you’d gotten more margin on your ad spend?” You know what they’re going to say? They’re not going to say, “Oh, I would’ve passed that along in savings to the consumer.” They would’ve said, “I would’ve spent more money on that ad campaign until my margin decreased to my standard. That’s what I would’ve done.”

It’s right for you to highlight this bit, because when I was reading the DOJ complaint, I swear I had to read this paragraph 10 times because I could barely believe they admitted it right up top, which is their first and most prominent demonstration of harm is that publishers made more money than they should have. But that’s the only part in the stack where there is arguably lock-in. Because, yes, you’re not going to have 47 ad servers on your site. Your site’s not going to function otherwise. And so this idea that the advertisers are being harmed and the publishers are getting paid too much when it’s the advertisers that have the least lock-in and can go anywhere they want to, it encapsulates this bit, which is, if you get the cause-and-effect backwards, your chief harm demonstration says the exact opposite of what it should, if your case was correct.

ES: Right. And I think that’s why they approached the complaint from the perspective of the publisher. Because if they did it from the perspective of the advertiser, it just wouldn’t resonate, because the advertisers are saying, “Well, yeah, I’m buying Google inventory.” And then I’ll click the button that says, “Also, open this up to third party network traffic.” I’m already at Google. I’m at Google because I’m buying their owned-and-operated inventory. And then I’ll just click this button. Well, actually, you can’t. Now it’s all black box so you don’t actually get to decide what inventory you want to buy. But you know that, well, okay, I guess you can sell me some of your network inventory. But what I came here to do was buy search ads, was to buy YouTube ads, was to buy ads in the Google Play Store.

Right now, yeah sure, throw in some network traffic and if it works, then you meet my profitability standards. And that’s great too. That’s just even more reach for me, and that’s even more conversions. But what I came here to do was to buy Google owned-and-operated inventory because this is the only place I can get it. And you see that come up a couple times.

And that’s not a crime. The fact that you have to use Google tools to buy ads on Google properties, of course, you do. Although, the EU is actually attacking that as well. But at least in any sane world, if you want to buy ads on Google, you use Google’s tools. That seems reasonable.

ES: Right. So Google actually, on the EU side, they’ve proposed a compromise where they basically said, “What if we made some of the YouTube inventory available to external demands partners?” So they could buy YouTube Ads in YouTube if we open that up. Because that is sort of an avenue of inquiry in the European Union. That hasn’t been resolved yet, but I feel like that that’s also been interrogated. But if you think about approaching this from the perspective of the publisher, it’s kind of what they had to do because if you’d approach it from the side of the advertiser, people say, “What are you talking about? They’re there to buy the Google stuff.” And Google’s making this other stuff available. And you see that happen. You see that surface in the complaint a couple times because they spend a lot of time talking about Project Bernanke and Project Global Bernanke. But actually in the sort of 10 accusations that they make, it’s Project Bell, which they highlight as being the sort of problematic abuse of power. And Project Bell is actually where they manipulated the publisher margins.

Yes.

ES: Project Bernanke was where they manipulated the advertiser margins. And the comparison I’ve made here is if I go to the used car lot or whatever, and I say, “I want this car. Here’s my max amount I’m willing to spend.” And the sales guy says, “Look, the car’s price is this, and then if I add my margin on top of that, my commission, it’s going to be too much. So I’ll just cut my commission down and that’ll get you within budget.” That’s kind of what Bernanke and Global Bernanke were doing. But on the publisher side, it’s basically taking something out of what they would’ve received. But again, if you take it from the demand side, people might have said, “Well, that happens elsewhere in the economy. That’s not necessarily this damning indictment of abusive power.” Now what I’m not saying is, I’m not saying that they didn’t do that. And I’ll let the lawyers make their case.

When I wrote about it, that’s exactly what I said. I’m like, “Look, you can take every single thing the DOJ said and assume it’s true, and there are these fundamental holes in the case.” Which gets to your point about basically looking at it backwards. This is my whole Aggregation Theory thesis which is if you don’t understand where the actual power comes from, which is controlling the demand, then you’re going to get it backwards.

ES: Right. Yeah, exactly.

Well, what you sort of hinted at this, the remedy was basically splitting off the exchange and the publisher tool, the DoubleClick for Publishers, which by the way, I actually think highlights the weakness in the case itself because number one, that’s exactly where they would have to split it because Google Ads are first and foremost for Google Properties. But if the foundation of their power in the market is Google Ads, then even this remedy seems like in the full run of time, why wouldn’t Google just rebuild their power?

ES: And they might. And just kind of circling back to my point about the sort of direction of dependency, just even look at the chronology of the acquisitions. YouTube was first. They had to get the supply first. The demand materialized because they had the supply. That’s how these two-sided marketplaces work. That’s how ad marketplaces work.

You have to have the customers. You have to have the viewers and then the advertisers will flow in.

ES: Yeah, they’ll just come because that’s just potentially incremental revenue.

Well, we just talk about with Facebook. What is Facebook doing? Number one, they’re building up all the customer engagement so that there’s a reason for advertisers to flow in.

ES: Yes. And they’re actually pumping the brakes. They’re saying, “Look, we’ve got all these people using Reels now. That’s great. We’re going to slow the roll so we can build the monetization stuff. We want to make sure we get it right. The supplies there, we’re going to ratchet up the demand until they hit an equilibrium. We’re not just going to sort of let it free run because that might actually overload it and break the economics of it.” But okay, so what happens? Google divests Google Ad Manager and AdX. Google Ad Manager, the publisher ad server, it’s been renamed from DFP. Well, I think the consensus among most of the people in advertising is just that. Well then Google will say, “Okay, well we’re not making the margins on AdX anymore because we don’t own it. So why would we push any of our demand through the open web? Let’s just push it all through our owned-and-operated sources.”

And if that just goes away, publishers are screwed.

ES: And then the publisher side, well then there’s sort of more concentrated demand or less concentrated demand. And so the rates go down and there’s more demand going through Google’s owned-and-operated. So their rates go up. And so maybe, again, it’s 15% of the business, but it’s very low margin. Where the owned-and-operated, it’s effectively 100% margin. So maybe they just say, “Look, it’s just the juice isn’t worth the squeeze anymore.” And by the way, that’s where all of the sort of privacy risk is. Right?

Yep. So maybe they just cut off all the third parties, exiting the web. The reason why they have power over the publishers is because the publishers, the nature of the web is basically infinite inventory. And so there are lots of spots to put ads. And if you don’t have access to the supply coming from Google, and to your point, this pricing is not set in a smokey room, it’s set by an auction. If there are fewer advertisers that want your inventory, your prices are going to go down. That’s exactly what will happen if Google’s gone.

ES: I do want to make sure that the argument is aired here about Google’s behavior being unscrupulous. They have pricing influence and they exercise that with their control of the end-to-end stack. And then they did that through Project Bernanke. They did that with Project Bell.

Which goes back to the black box idea, where they were putting in more valuable bids and then less valuable bids, and they would say, we’ll deliver an average instead of saying, we’re going to deliver explicitly 20% or whatever it might be.

ES: Exactly, if we didn’t do that on your behalf, we just said, “No, we’re not taking your money,” then the advertiser would have earned less revenue. They would’ve earned less profit. The problem here is that have comparisons being made to brick and mortar retail and hamburger outlets or something, or a lemonade stand. Well, hey, if the price of lemons went down, I’d probably pass that along to the consumer. That model just doesn’t apply here. The auction breaks a lot of that model. And the idea of incremental revenue breaks a lot of that model. And the idea of the advertiser actually saying, “Well, here’s the max price I’m willing to pay,” breaks that model. And so I think the arguments here need to be more sophisticated than just borrowing these sort of antiquated perceptions about what monopolies look like and what monopoly power really is.

I just want to jump on this because I think you just made a really important point that brings this full circle, which is the reality is the price is what it is. The argument in this space, and this goes back to the Facebook point earlier, is about the margin that the company in the middle takes from it. That’s the part that sort of shifts up and down.

ES: Right. So if I was Google, here’s what my argument would be. I would say, “Hey, Department of Justice, you said we’re behaving as if Citibank operated the New York Stock Exchange.” Well, that’s nonsensical. That literally makes no sense. I don’t know what they were trying to invoke with that comparison, but it’s nonsensical. I don’t know what you’d even be saying with that. What we, Google, are is like a market-maker. So we’ll go out and price this traffic for you, especially this long-tail traffic, because our data is so good and our models are so good that we can price that. A lot of people can’t. If there’s not a lot of people buying that long-tail traffic, then other parties can’t price it. But we can because we’ve got all this data that was acquired legitimately through our first party owned-and-operated business.

We’re going to take that data, we’re going to bring it to bear on pricing these long-tail impressions, and we’re going to get you a price and we’re going to sell it and we’re going to allow you to buy it. And in such a way that it delivers incremental profit to you. Now when we do that, oh, I’m going to rip your face off. I’m going to charge you a margin that’s totally unconventional, but you know what? You’re going to get some incremental profit out of that deal. What do you say? Are you going to do that? Now that’s how I would argue that side of this suit. I would say we’re not the combination of Citibank in the New York Stock Exchange. What does that even mean? We’re like a market maker. We’re making markets for this inventory that would go unsold if we didn’t.

Yep. Yeah. Well I think this stock exchange analogy is alluring because you get this idea of them being all sides of the market, but it presupposes a level of scarcity and network effects that don’t exist in the ad ecosystem. That’s why it falls apart. There’s a ton of exchanges, there’s a ton of ad managers and there’s a ton of advertising products. And just because they’re delivering an integrated one, it’s like Apple delivering an integrated smartphone where you can get modular ones elsewhere. It’s a natural plane of competition.

One last thing. You sort of mentioned this, especially with all the privacy stuff going on. Is there an aspect where Google might not mind dumping the whole thing? I mean, why not? Or is it a matter of look, having maximum inventory, having maximum control, that actually protects our core business in the long run. The more moats you build outside, the less likely even to get to the core moat. What’s your take on that?

ES: Well, so in this suit, they mentioned the kind of agreement they had with Facebook, but that agreement played a much more prominent role in a previous suit.

With Texas and Texas basically lost on that point.

ES: Yeah. So I don’t remember the specifics of it, but it was Facebook Audience Network, which is their third party product. It’s their sort of product for serving Ads in third party properties on behalf of Facebook advertisers that they had some kind of agreement where they didn’t apply Project Bernanke to that or something.

Yeah, I think it was that Facebook would get privileged access to Google’s header bidding alternative. And I thought an actually interesting point from the complaint was that Google would basically not look at that data. Because I think Facebook’s worry is that to the extent we’re on Google’s platforms we will get leakage and Google will be learning more about our customers. And one of our moats is that we have our own data and Google’s like, “Okay, we won’t do that.” And then Facebook’s like, “Okay, then we’ll be on your platform,” which was win-win for them.

ES: The bigger point is Facebook shut down Facebook Audience Network on the web. They said, “Eh, this isn’t worth the squeeze.” Now the question is — and here’s what I might argue if I was in the DOJ seat is, “Look, there’s a reason you haven’t just abandoned this. What are you getting out of this beside the revenue? Is it the data? Is the data actually worth more than the sort of net revenue that you accrue from this?” And if that’s true…

And that might be true, particularly the pricing data and knowing what everyone else is bidding. You own the exchange. You see a lot of stuff.

ES: Exactly. And so they would say, “What’s your motivation to even keep this alive? Because it’s 15% of your overall business, it’s a headache, and that’s on gross. So when you net this out, what are you really getting here?” And it might just be the data. The data actually flows back into their ecosystem and then they can apply it to their owned-and-operated inventory and get better outcomes there, right?

Yep.

ES: Now I say that because that’s exactly what’s happening in the mobile app ecosystem right now. You saw all the consolidation on the SSP side, and that was my hypothesis there too. They get much better price discovery if they own the supply side because it’s a much broader scope of data than if you’re just operating on the demand side. Demand side, I send a bid out and I find out if it won or not, and I find out which bid did win if I didn’t.

On the supply side, you can see every bid and you get to see who bid it. It’s just a much broader scope of data. Now the other thing here though is that the regulators might not want them to divest this stuff because guess what? If they do and Google just says, “Well, I’m going to just drive everything to owned-and-operated now,” that’s probably worse for advertisers and publishers. Right now, keep in mind that when Google said, “Look, we’re going to abandon the third party cookie in Chrome,” the [UK] CMA [Competition and Markets Authority] said, “Not so fast.” They came to an agreement, which they published last year in February. The CMA said, “Look, we’ve come to an agreement with Google that’s legally binding. They’re going to conduct the privacy sandbox experiments in the open with us. They’re going to be very transparent and they’re not going to abandon the cookie until they’ve replaced it with something better.”

Because if they did that, that would just drive all that revenue through Google’s owned-and-operated properties and it would starve the rest of the ecosystem in the open web of that revenue and of that advertising spend. And so there might be some kind of situation where they say, “Look, we’re not going to allow you to just dump this because that’s probably going to be worse off for the ecosystem. We’re going to sort of just force very specific rules on you for operating because we know that you’re getting something out of this that’s not just the sort of monetary remuneration.”

Well, this is the big picture thing, which goes back to the privacy question. The stricter and more rigid the privacy regulations, the greater the benefit accrues to the larger companies because they just have more signal, they have more ways to figure it out. We’re seeing that play out. This would accentuate it. To your point, I think the data point is very important. There’s a point I was going to make earlier, which is people who advertise on Facebook — you’ll hear this if you talk e-commerce folks — it’s a complicated relationship because one you both simultaneously know that Facebook is extracting so much margin out of you, but two, you would never in a million years find those customers without Facebook.

I think this ties into your point about this better data where you see all the bids and you see everything that was available. And if you’re at the publisher’s level, you see it not just from advertisers, but across exchanges. You can imagine how that makes that black box that much more effective. And this is the sort of pushback on the point that advertisers choose Google willingly. They do, but they do it because they also have no choice. And how does that actually manifest into anti-competitive outcomes or not when it is beneficial, it just kind of feels bad. That’s maybe a summary of digital advertising in general: it’s beneficial, but it feels kind of crappy and this is an articulation of that.

ES: That’s the perfect characterization of ad tech. That’s exactly spot on. And it’s just that situation where it’s like, look, reading through the complaint, you can’t help but feel that this is unscrupulous behavior. This is somewhat hostile, but even if that was transparent, maybe they should have just been transparent about it. Would the advertisers have done it anyway? So there’s a setting on Facebook. I’m blanking on the name for it, but they basically do something like that where they’ll get you some conversions that are more than what you set as the price cap and then some that are less and just so long as it averages out or that’s just sort of the commitment that they make to you that in aggregate, on average, the conversion price will be at or below your cap, but not for every single conversion. And that’s just a bid strategy. You select that. But maybe if they just made it transparent like that they wouldn’t be in this position.

Which they kind of have started to do recently. That’s what was kind of striking. Google, it feels like they’ve gotten a little more sort of brass knuckles about things recently. And honestly they were brass knuckles all along. And to your point, had they been more transparent and honest about that reality, maybe people would’ve felt crappier earlier, but also they wouldn’t be in as much trouble because there wouldn’t be any question about what was going on.

ES: Right.

All right, well I think we did pretty good. We went a little bit long, but we had a long list of stuff. I think we covered all of it. It is good to have you back on and of course I look forward to talking again in the future.

ES: Yeah, cheers, Ben. It’s always a pleasure. Thanks so much for having me.

---

This Daily Update Interview is also available as a podcast. To receive it in your podcast player, visit Stratechery.

The Daily Update is intended for a single recipient, but occasional forwarding is totally fine! If you would like to order multiple subscriptions for your team with a group discount (minimum 5), please contact me directly.

Thanks for being a supporter, and have a great day!