Revenge of the SSP

0:00 Feels and impressions like a verbal sprint Breaking it down with an ad tech

0:04 tin Ad tech, ad talk It's a

0:06 live Q and A Buy side, sell side clearing the way CEO spitting truth, no delay Chalice and Gamera owning the day. Programmatic flows in the scene real tight RTB dreams lighting up the night Adam says pivot, Garrett says scale together. Hello, young folk, what you shaking?

0:38 I gotta try that Single naked Liberace

0:43 turning on who say remember and go

0:49 deuce today while you play a piano just look up and say feeling groovy

0:58 wow.

1:00 Like they dosed you was spry.

1:06 Hello, Friday again, we are back.

1:12 Welcome everybody back to Ad Tech Ad Talk. I'm Gareth, I represent the sell side.

1:16 I'm Adam. I represent the buy side. It's August. We're into the. What's happening? It's August. It's. It's.

1:26 No one saw that.

1:27 We're into summer. It's earnings season. It's a very busy week. You know, you always want summer to be kind of slow in advertising, but it's actually usually very eventful, at least until the end of August. And this year is no exception.

1:40 No, definitely not. Like, there are people on vacation. Yes, but things are moving, man. Ad tech is moving and shaking, like, and you can see it everywhere on the socials. Everyone's like, whoa, like, a lot is happening right now. It's really nice. It's really fun. I love it when. I much prefer it like this to where things are changing, to when, you know, it's just the same people parroting the same talking points over and over again.

2:04 I mean, that's the. That's the life of a disruptive startup. Right? Like, the more change, the more better, the more shaking. But it's, It's. There's a lot of people invested in the status quo and all this activity makes them very nervous. And it might be good for us. It's. It's discomfort for most people.

2:19 Yeah, I mean, it makes sense. It's also a really disruptive time in the best way. So I've got something I want to talk about and I. I won't say who I. Who I got a meal with to talk about this with, but it's. It's a perspective on the future battleground. And I think it's a really interesting. Because they. They put it to me like this. My. My come. I'm betting my company on this, basically. Right. Like, this is.

2:45 This is our future. And it's in the vein of fabric. Right in the vein of us Fabric.

2:52 Okay.

2:53 Where the notion is the future of the scaled ad tech provider, of a certain breed of scaled ad tech provider, is being nothing more than cloud services that allow other vendors to, to participate in the ecosystem seamlessly with one another. And you charge technology company prices to facilitate the integrations between these companies.

3:22 Well, okay, wait, wait, wait, wait. So this person you're talking to and you agree believes that value accrues in that infrastructure layer, not in the application layer or in both?

3:32 Well, I've got to believe it accrues in both. Right, because the infrastructure play doesn't work if there's not value in the application layer. Because the application layer companies need to exist. They need to, to make money. They need to sell.

3:43 I guess the question is, which is more competitive? Right? Like we could build our applications so that they work in AWS and Index cloud and Snowflake Cloud. Right. We could build our applications, but so can anyone else. So that will be very competitive and it will be hard to charge a lot for applications. But the same thing applies to those cloud providers where if they try to charge a lot, then we can, you know, the user can switch and say, oh, I'll just use camera and index if AWS is charging too much.

4:12 So I would argue. I'm not sure value accrues in either because they both seem to end up very competitive.

4:18 Well, I think it comes down to network effects at the end of the day, right? It is. Are there any differentiating features that you need to be a participant in this ecosystem to take advantage of? And I don't think that's shaken out yet. Right. I don't think the interoperability question has come up yet. And I certainly think the landed gentry of ad tech have something there where I think of AWS and tam. Right. And like the, the a linchpin of demand in TAM is Amazon dsp.

4:49 Amazon DSP buys like crazy through tam. It loves tam.

4:52 And tell everyone what TAM is.

4:54 TAM is. Yeah, yeah, yeah. Throw the rest of the market. So TAM refers to Amazon's publisher product. It's part of aps, Amazon Publisher Services, and TAM is Transparent Ad Marketplace. So it really is just there, like header bidding. God, I don't know.

5:12 Pseudo ssp, right?

5:13 It's not an S. Yeah, well, it's, it's not technically. It's kind of an SSP. It's a pseudo SSP because it calls as mostly SSPs.

5:21 More importantly is the actual action of the bid request occurs in. Right. The publisher. Some publisher tech is sitting on us, right? Yeah.

5:31 Oh, correct. So this was they were really smart when they did this because they realized they had the hefty. Because everyone wanted Amazon buying on their websites and they did not originally go through prebid. So the distinguishing feature of APS and TAM is it is built entirely independently of Prebid or was for a very long time. There's a prebid adapter now, but you still can't just. To my knowledge, and someone correct me if I'm wrong on this, to my knowledge, you can't just do APs and Amazon's Prebid integration.

6:00 You still have to put their library on the page. And by the way, this was inspiration for open ads from the Trade Desk. That is the trade desk saw this and they were like, we should have something like this. Amazon has this.

6:15 What does it mean to put their library on the. On the publisher page?

6:18 Yeah. So when a. You are working with a publisher, you and you go through prebid. Prebid is a framework that has a bunch of rules. So when your code is implemented with a publisher via pre bid, it has to play by Prebid's rules. That means there are certain things it's not allowed to do on the page, there are certain ways that it's allowed to interact with the page, certain timing, etc, etc, Prebid is the boss and Prebid will control what your code is allowed to do.

6:48 And this is managed by an open source community who's reviewing everyone's code. They're looking to see that it's behaving. This is the wonderful nature of transparent open source code and infrastructure is that everything has to run through Prebid and Pat McCann and Pat McCann who is the. He was. He's the head of research at Raptive. I worked with Pat at Prebid. I love Pat and he, he will tell you this is bulldinky. Your adapter can't do this.

7:16 Like this is not allowed in prebid. Whereas if the library is directly on the page, it can be loaded before prebid. In fact, APS often has to load before Prebid does and it can do other things. Things. Now what those other things are I don't know. And like I'm not sure that I understand fully the value here, but there is ostensibly some value and having your code on page outside of pre bid for a buyer.

7:42 Does the code call to Amazon servers and it does. Yeah, that's, that's. So that's it. And that's probably what Trade Desk wanted because then they have some signal in advance of the bid request. Right.

7:52 So. So that the original advantage was actually that Amazon got called before pre bid and wasn't subject to pre bids timeouts. So you would have like a, a pseudo meta auction. I remember we had to code this together for Amazon specifically where it's like hey, we need something to manage the Amazon adapter and pre bid at the same time. But because of this Amazon's bids got into the ad server sometimes a little bit more effectively and they.

8:17 And that was an advantage. Right, right, that was the thing.

8:20 All right, I should wait for you to be done but I just want to throw out there like for this thesis to be true and value accruing to Amazon because of all the hooks they have in to media companies. Right. Like most publishers run on them, most DSPs run on them, a lot of SSPs run on them. So that's a lot of advantage. But it's hard for me to imagine Google with all their ambitions from cloud sitting back and letting value accrued to aws.

8:48 And I often say I truly believe that Microsoft has not made their move in this space yet. But as the leader in enterprise technology with a history of waiting for markets to mature before making their move, I think that be an Azure play against AWS for this business as well.

9:08 You know my only reservation about that is because of the network effect at play here. This might be a winner take all market. Like yeah, the person who gets, who gets critical mass first where if you're starting an ad tech company and you integrate and build your module on top of their hosting, you can now sell it to everyone way more effectively. You will prioritize integrations with those and But I do think you're right.

9:39 I think it can be very, very competitive because for smaller companies it's easy to host in multiple places. Right. It's easy to have infrastructure on AWS and on GCP and each of those. For larger companies it's harder because it's more expensive. But it certainly could shake out where if Amazon nails it they, it's. It just makes life so easy that the integration discussions and all those things just disappear and it's like oh, we're in fabric, just click, just click the test button and turn it on.

10:11 Yeah, we'll see. A really big factor is CDPs and data warehouses, right? Because I think there's a huge number of e commerce sites that are sitting on GA4 and then the Google cloud integration comes in right behind it and they start doing all their CRM and email out of that. It's all Google and audiences straight into Google. There's a lot of E commerce sites that are operating that way. Full Google stack and the only one compete.

10:39 Yeah, the ones competing with them are Salesforce and Snowflake and neither of them have like the whole application workflow that Google has yet. So I see AWS seems to have the leg up in enterprise and Google seems to have a leg up in the bigger market of, of long tail E commerce.

10:55 I completely agree. I've also in my experience Google doesn't seem to really like playing with ad tech companies. Like they don't necessarily don't, don't like it. It's just like it's. That's like that scene from Mad Men where someone is like, you know, I think about beating you all the time or whatever and then he's like yeah, well I never think about you at all. Like I feel like, like Google doesn't think about ad tech companies. Like it's not even on the radar.

11:21 We should hear from a booth.

11:22 I mean that's not how Google cloud providers think of it. That's how I think when someone says they're on Google Cloud. I think oh, that's not a real cloud. There's really no interoperability. Google Cloud is just Google's infrastructure for combining GA4 and YouTube. But to be a real cloud there has to be an application and co working environment.

11:43 Let's see now there does. The original definition of cloud was just.

11:48 Yeah, it's just servers. Servers you pay for as you go. Yeah.

11:52 But the expectation has changed though. I think you're right. And that's because of people like Amazon pushing it forward. And I do think that there's a, there's a nice tack here for, for legacy ad tech companies that gets into our thesis for some of the earnings later I think.

12:08 Yeah. Something I realized from my team that I think a lot of people don't realize is that Amazon is, is far ahead if you want to use multiple LLMs. You know, if you want to have a mix of frontier models and, and open source models are they called or off the shelf or. It's not called open source, open weight models and everything. Amazon's far ahead. There's a product called Amazon Bedrock where you have access to all those libraries and if you want to build modularly and switch that up as we do, it's looking like it's best by far.

12:43 Oh wow. Yeah. Don't sleep on Amazon, man. Don't sleep on them. They're amazing.

12:50 They're.

12:50 They're making moves. And that Amazon DSP products works that thing.

12:58 It works. Yeah, they have a lot of Share huge.

13:02 They have real data. They have great data.

13:06 Yeah. I don't know, it feels like a mess in there. I don't understand Amazon. Do you understand them as a company?

13:12 Not at all. Yeah, I, my experience with some of the teams is you can have a call with them and if they don't already do what you need them to do. Right. If they need to build something, the answer is, well, the product roadmap for the next 365 days settled. Like there's no moving it. It is done. I think they strategically plan in that way where their product cycles move in years.

13:42 The thing Steph Laser said when she was on our show, she of course works at Amazon and the one who got them on prebid said that the culture of listening to the customer and building based on what the customer wants is imbued throughout the culture. So if there is a long term roadmap, according to Steph, it's deeply rooted in what customers say they want for the future market, but sometimes invisible to us.

14:11 Yeah, I mean I, they talk to so many people in so many places. It's, yeah, it's one of those things where it just looks cryptic. But yeah, I, I'm, I, I think fabric's a really interesting play. I really, I love all of these interoperable operability plays because my life is so painfully consumed with integrations and I just like want that to be over. I just, I want the integration to not be a discussion. I just want people to be able to test and use things.

14:41 You know what I mean?

14:42 Yeah. Oh my God. The amount that ARTF and the, and the shared compute environments and SSPs are underreported, it's just so insane. Like, you know, we got coverage when we launched with Index and Index Cloud, got coverage and Open Environment. I forgot what it's called now. Is got a little coverage. We got good coverage at the Hyundai case study. But now Chalice is live with the exact same containers we use in index and OpenX with pneumatic and Pumatic has a whole bunch of potential add ons connecting to agentic hosts, their agentic DSP and service.

15:19 And now Magnite just launched and now both public SSPs are adopting versions of ARTF and we have three more SSPs lined up. But to have Index, OpenX, Magnite and Pneumatic already adopting within the first couple of years. This is the most successful standard since header bidding and all that you hear about in the news is the other agentic framework.

15:43 Yeah, you know, I don't think people get it. But I think the markets get it. That's how it seems to me.

15:49 What do you mean markets? Earnings. Yeah. You think those markets, those. I mean, they understand revenue, I think they understand revenue.

15:57 And I think that these. This creates revenue. Right. Like I. In my mind, there is no way that index being early on, the chalice integration didn't generate lots of revenue for index. Like it.

16:11 Yeah.

16:11 That it meant that budgets flowed to them because of their architecture. And like, when can you say that? That's I guess, early open rtb, Right. Like early open rtb. If you were an early open RT adopter and you got in there quickly and you got integrations that other people didn't, you got spend that other people didn't. I GU is like the closest corollary. But I think we're there now. And I think that the reason that the SSPs are doing so well right now is they're just.

16:40 They're the first to open architecture. That's all that it is.

16:44 Yeah. There's something very simple like if, if your infra makes the media cheaper, then there's more room for margin and some, you know, value is going to accrue somewhere. Let's get to earnings. We have a lovely earnings graphic that we haven't used. And there we go. Earning seasons, folks. Bunch reported yesterday and the day before. So this news is very fresh. The one I was just reading was Pubmatic. They did well and their stock shot up, Right?

17:11 Yeah. Which is great for Pubmatic. They needed it.

17:16 I thought, I thought it was kind of funny about Pubmatic is like they're accrediting all these things that I, I believe are very small. It's like, oh, we have thousands of agentic campaigns. I was like, yeah, I don't think that's it. And then they're like, oh, our CTV is up a certain percent. And I was like, yeah, I don't think that's what drove the better earnings. And then they're like, oh, and our infra costs are down on 30%.

17:36 I was like, that might be it.

17:38 Yeah, they cross earnings per share.

17:43 Yeah. And the stock really pops. Because while she didn't expect that like the, the market had priced in, you know, declines to match those of other sectors of the market. So when they go up instead, the stock really pops. But it's hard to tell what drove the revenue growth. The market may believe that it's because of agentic campaigns. I don't believe that. But there's something very real and they're improving infrastructure costs. It's not just a quarterly thing, but a result of probably years of redevelopment and moving things to more efficient and containerized workflows.

18:19 Yeah. I also think that the containerization movement has moved so much of margin management, and not just containerization, but curation in general has moved so much of margin management out of DSPs into SSPs for agencies. So like really big agencies have realized there's a, there's now a centralization point where they can push all of their localized teams to buy through the global curated market thing and put it on a different P and L and do all of this.

18:51 And SSPs are the direct beneficiaries of that because a lot of SSPs charge more for curation than they do for opening open web. It is, it is incremental revenue for them. So even just curation succeeding, right, and, and selling a curation deal to an agency holding company where it's like, oh, implement all your stuff right here and you know, implement all your margin controls. I, I can see that being a big gooser of it's a real cost reduction.

19:16 Right. So the agency deal is just the agency trying to get some of that value to accrue to them, which is totally rational. But I think what people don't realize is like one of the biggest problems in Programmatic, one of the biggest costs has to do with the unevenness of bid requests. Right? There's a big news day, there's a surge in bid requests, there's live sports, there's a surge in CTV bid requests, and any sort of infrastructure flow, the best ones, the most manageable ones, are inhuman.

19:50 Even running a power plant is tough because there's surges in power demand. But media is much worse into the surges, the peaks and valleys of it. And there's a lot of costs associated with trying to manage these peaks and valleys. And when you see SSPS saving money, it's essentially new technology that makes it easier to manage these peaks and valleys. And containerization is one of the easiest examples to explain. It's like if you split a server in two, that's what the containerization technology comes from, the way servers are built, that if you partition the server, then the other half one that you partition is like another computer is essentially like another server.

20:28 So the more you containerize, the more you have the ability to turn on other servers as you need them or to only use servers when you need them. So that's a lot what's going on on the SSP side. And it's been going on for years as they get better at managing the peaks and valleys of programmatic demand. Their costs go down. Right. And that value is going to be captured somewhere.

20:48 Yeah. I mean, I think they're crushing it. ARTF was a great thing for all of them to adopt. I would push all of them to adopt it faster and to adopt it robustly.

21:05 Yeah. I'd say. There's a Magnite piece by their cto, David Buenos Aires, he wrote about their hybrid infrastructure, saying that they've found the right split between Cloud and On Prem to optimize the peaks and valleys. And Magnet has one of the biggest problems because there do so much premium video like Magnat has a bunch of businesses. They're one of the more complicated SSP businesses. Right.

21:28 So he wrote three SSPs in magnite. Three different ones. Two on cloud, one on prem.

21:34 Yeah. He wrote a white paper on how this hybrid infrastructure is best for them.

21:38 Yeah, it makes sense.

21:40 Yeah.

21:41 I remember when I was there, we did a lot of cost modeling on this because there was a big initiative. Managing On Prem is very expensive. You're literally buying and amortizing actual computers. Magnite had a global team flying around managing these boxes. These were not licenses, literally. They had to go into the data center and look at them and say, oh, this one's about to explode. It's real old. But it's all business.

22:10 In predicting when they're going to fail,

22:12 oh, 100%, you have to know you use them, of course, up until the last moment that you possibly can, but eventually they fail and then you have to have cutover and redundancy and all these things. But when I was there, there was a huge motion inside of Magnite to migrate to Cloud. And the On Prem team was like, no, no, no. There's tremendous cost savings and actual cost of infrastructure here, which is true. But there are just all of these other costs.

22:40 Right. That don't show up in your build. They are elsewhere. They are development costs, they're maintenance costs, they're all that stuff. But it certainly is very, very, very cheap for certain scale. And I think that there are in betweens, like we've talked about before, but the cloud is really, really good for spinning up new things quickly.

23:04 Yeah.

23:05 When you. When you don't know what your capacity is going to be. I remember every quarter there were all of these discussions, guys, how much of volume are we going to have this quarter? How much volume will there be? Like, because we need to know, like, and we need to plan if it's going to jump up By X percent. We need to know that because we need to provision infrastructure for this. We have to provision it well.

23:25 That's why those agency deals have so much value for both parties. Right. If they, you can get the agency to commit on running a certain number of bids through there and. Right. And then they, they know that they have that capacity, they can set their infrastructure accordingly and reduce cost a lot and then realize a bunch of money. Right. The agency can make money, the SSP can make money and even the end advertiser can make money through one of those deals.

23:50 No, I mean the agency holding companies get real wild with this. Like there are plenty of agency holding companies who have their own exchanges like already extant white labels of Magnite and things like that. Yeah, it was, it was curation before curation was we want our own separate instance Zaxis. Right. Zaxis and have Nexus did this. So there's precedent for it. But yeah, I, I think when this the the governing thesis here, right. If you look at Pomatic's earnings, you look at Magnite's earnings, the revenge of the SSP is complete.

24:22 Like happened fast. It happened fast. This was a moribund category that could nowhere to go but down.

24:28 No differentiation. Right. Commodity race to the bottom. Like their fees are just going to keep going down. And they realized, wait, we don't have to be SSPs, we can just be ad tech companies.

24:43 And that's like what we're gonna, we're

24:45 gonna be in, we're gonna be ad tech infrast. That's what we're gonna be. And you know there's a, there's a through line for this which is everyone did all of this bellyaching about spo, all of this bellyaching about it. And to the layman it just looked like, well, just buy from one exchange and tell them you only want their direct publishers. Yeah, it's not that complicated. Just do that. And no one could make that happen. Right.

25:15 That wasn't the thing. So nobody, nobody ever just implemented all we're just going to buy from like two exchanges to spoil. But so the exchanges were like, okay, well then we're going to get you here by thinking, by selling you something else. We're going to sell you curation. Right. We're going to sell you some other sort of product. We're going to sell you moving your decisioning, we're going to sell you these things. And then lo and behold, you've accidentally spo'd.

25:41 Right? You didn't mean to, but when you're using Chalice containerized on index, you're buying direct and like you listen to all the QPs and it just works better. Like what a surprise what every publisher was saying for a decade. Just buy from one of them.

26:00 Yeah, the real. Right. I think we just covered what was really behind spo. The real goal of SPO would be to capture some of the value that's created through more efficient infra. Right. And picking an SSP doesn't do that. But obviously Jounce's message is like you're definitely on the opposite extreme. If you're buying a lot of resold inventory, you're on the very wrong end of the spectrum because that's the least efficient infrapath and there's no way to realize any savings that way.

26:30 Totally. The servers themselves have to be paid for and then the martinis have to be paid for. And between those two things, you're not going to have a lot of working media if you buy through tons of resold paths. And working media is what leads to outcomes. And you know, the buyers get some credit for this. In order for this to work, buyers have to actually care about the performance of their campaigns. And there are still plenty who I don't think do, but I think that there are a lot who do and they're the ones who are driving.

27:00 This infer is a tough topic. Let's. Let's turn to dsps. Touch on trade desk earnings. Because what do you, you know, like if SSPs are becoming more valuable because they're infra, then the DSPs copy them and also become infra. Or do you see another path?

27:17 It's a really good question because I think that the desire for infra is new. If you were to talk to a DSP 10 years ago about all this interoperability and all this stuff, they would be like, yeah, our clients don't care about that. Our clients just want us to make their life easy and send them nice PowerPoints of successful campaigns. That's what they want. So I think that there's a little bit of a shift here in what the market is expecting of ad tech companies where it used to be make it easy and make it work well enough and now for a certain percentage of the enterprise it is help me bring my assets to bear in a way that gives me solid results.

28:03 And I think that those are just really different markets. Right? They're really different markets. And the trade desk came up in the first one and was the best in the game at the first one. They were the kings of we make our agency clients happy and the market has just shifted to, well, I need you to tell me how you're going to help me activate this as, as efficiently as possible in the enterprise. And the agencies have been forced to do that because they have to keep up with each other and with brand in housing.

28:35 So the market has just moved and what worked then and what went like, what, what solved things 10 years ago doesn't solve them today. So this is the, I think the trade desk's earnings and what the trade desk has been going through is just, they're just going through this shift where that what their business was 10 years ago is different than it needs to be today. And I think Jeff has like, for all of the, the misery in the stock price, he's been very consistent in what he thinks the future is.

29:08 And he is like, I'm building towards this future and I'm playing the long game in this. Yeah, he's always, he's always said that he's, he's, he's been a broken record about it for a while and I, as long as he builds to the openness that those buyers will require or maybe he like builds what they need to bring their assets to bear in this way. Like, I think about Snowflake a lot, right. You did a lot of press with, with Snowflake early on and I think that that's, it's a really interesting play.

29:42 It's a really lovely play where Snowflake comes in and talks about how, oh, nice, nice Snowflake mug. But they talk about, hey, hey, we're going to meet your data where it is and help you to activate it from there and not have to integrate in a bunch of different places. And I think that's the sort of message that needs to get out there. It's like, hey, we have all of these different vendors. We'll do this all in one place.

30:05 It's best in class for you. And we're going to deliver this enterprise solution the way that it is. And, and SSPS just got there first because they created all of this super interoperable tech. There were no integration queues. Like, my gosh, Index has integrated some astounding number of vendors via ARTF and can do more because it's a standard, right? Like, it's, this is a scalable business for them. So they, they can go to a buyer and be like, oh, you have assets here, here, here and here.

30:33 Great. Here are all of the vendors that we work with that integrate natively with those that will Build an advertising stack for you right here. Trade desk needs a version of that where when they're in that meeting they can say, okay, you work with this, you work with this, you work with this. This is the stack we're going to give you. And. But it needs to be interoperable. Right. People need to. You're going to compete on the quality of the vendors in your ecosystem, not your tech itself.

30:58 A lot of the time.

31:00 Yeah. You know where the rubber meets the road on that. I was thinking as you're talking, is log files. Like a really big difference working with SSPs compared to DSPs for me is the ease of getting log files. They're like, yeah, the logs. Like, since we're infra and we want you to build applications on top of the infra, we want you to use your logs, we want you to get the logs and use the logs and we'll make sure you have permissions and that we're papered over about how you use them and have all the privacy restrictions and all.

31:28 But we're not, we're not in the business of getting, of being afraid of what you do with logs. And this always got me, yeah, all

31:36 the DSPs are nervous about their. Not all, but many of the DSPs were always cagey about logs. Always cagey and so privacy.

31:45 But it's coming from a belief that this is our data and if we let anyone else use it, they're going to play in our sandbox. And I think that's the kind of the pivot they have to make if you want to be infrared. Empower enterprise advertisers to bring their alpha to market through your tech. Like, you can't tell them your logs are my data.

32:07 I completely agree. And just think about integrating. And I know this is true not just of the trade desk, but of lots of big DSPs. Right. When they integrate with a data provider, let's say it's a real time data provider who wants to synchronize data in some way, or let's say it's a custom algorithm provider. They're not built using RTF. They're like, oh, we'll give you these 10 fields that you can use to make your decisions, not the entire bid request.

32:28 And that's it. They're hamstringing the performance of these vendors. So if these vendors have successfully sold into the enterprise, which they normally have, because their pitches are normally very good, they naturally will gravitate towards the places where their tech works the best. And you just, I can't imagine A world where like we, we in ad tech, right? The royal. We look at each other as infrastructure, as applications built on the infrastructure providers, where we look at each other with suspicion or like we should all be looking at each other and being like, okay, how do we deliver the best possible outcome for this campaign that's coming through?

33:09 And the world of, the world of DSPS and working with DSPS, the 10 year ago world was I really don't want to integrate with you. Really. Like this advertiser's making me do this,

33:23 but I'm not share revenue, right? And share revenue like integrate and share revenue chalice is like a Johnny Appleseed going around to data companies saying, hey, there's this new way where we share data, share revenue. And they're like, we get a variety of responses.

33:39 I mean it should be joy.

33:42 It's not joy. I mean the more invested they are in the status quo, the worse it sounds. Right. Which I think would bring us to the M and A activity this week. Like the, the company who's I think that that pitch falls flattest to has been Nielsen and the companies that Nielsen owns. And of course they bought Double Verify.

34:02 Oh, this is a good one from. From John d'. Souza. The day Amazon DSP gives their logo logs away outside of AMC aggregation is the day it gets torn apart.

34:11 But yeah. What gets torn apart, John?

34:14 Yeah. You mean that, you mean the DSP will die because people will realize what users they're buying? I'm curious. We'll let him respond.

34:24 Yeah, I mean I don't care about Amazon logs. AMC is a big barrier to things I want to do in aws.

34:34 What's amc? Amazon Marketing Cloud.

34:38 Yeah, it's their clean room. Their clean room is an S3 bucket. Is a specific S3 bucket with weird permissions where you can see and do some operations with logs, but very minor operations with logs. It's useless. But the problem, the problem is not that it's useless because we don't really need those logs. The problem is that anything you want to do in aws, the service team tries to steer you into this clean room that somebody at Amazon has championed as the future of their integrations.

35:07 And they're just wrong. Like it's just like a very limited S3 bucket. They should just let it go.

35:13 I mean it's the worst S3 bucket.

35:15 It's for Amazon sellers, right? For Amazon customers like this data is really useful for them to make better Amazon audiences. But when I work with a commerce, if I'm working with Hershey Or Bayer. They're not trying to just drive Amazon sales. Right. They, of course they care about Amazon sales. They sell product through there. But the majority of their sales are off Amazon. So these logs are not the be all, end all.

35:38 Of course they're. In fact, they're used probably in large part to measure like incrementality, I would think, like, if you, if we want our Amazon sales on TV elsewhere, like, yeah, like we just, we just want to develop a complete picture. This is not to reverse engineer.

35:56 Yeah. So they can't get out of their way. They can't get out. They can't break with their legacy as a giant e commerce company and actually build for cpg, Pharma and auto the way we've been asking them. But they'll get there eventually. But, and we were talking about Amazon, I was thinking a little bit about Microsoft. In the last era when Microsoft was an ad tech company, you could talk solutions to them and you'd get like brilliant people who wanted to be really helpful, but nothing ever got done.

36:26 And I was, I'm afraid that Amazon is going to be like that.

36:31 Oh, that's funny.

36:33 It's like so great working with them. It's just too bad they never do any. Yeah, yeah.

36:38 I mean, that was a story that was with someone yesterday was talking about working with Xander as the ship was going down and they were like, man, the product could have done so much and then just. It was a ghost ship. It was just floating on the ocean with no one in it.

36:54 Yeah, Microsoft.

36:57 Rough scallions.

36:58 Let's get to Nielsen's acquisition of Double Verify. A 30% premium over the stock price, but still 13 where it was at 45, I think in 2024.

37:10 So, yeah, that's really interesting. So I, so I looked at this. I've. I haven't been following their stock price. I was just following their acquisitions and we were chatting about this before the show. I like their acquisitions. Like, they make sense to me. Like from a, from together perspective. It made sense to me. But I guess they didn't make much more money with them like they did.

37:32 They went from like 400 million a year to like 700 million a year. Like added 100 million every year since those acquisitions.

37:38 So they bought cybids in 2023 for 180 million and they put up an additional. They kept growing like more than 10% a year. And the market didn't like that. Why?

37:51 They expected more. Yeah, they expected more growth, I think, you know, as money moved to the walled gardens, their Growth got behind Meadows, right?

38:02 Yeah. They weren't measured absolutely. They were measured against how much are you out? How much are you keeping up with Google and Meta.

38:10 Yeah, that's what drove their stock market.

38:13 They're just such distinct markets. Gosh. People like, measure them in this way. They're just so distinct. They're selling to different groups of people.

38:23 Very hot right now. If you like the narrative of double verified buying cybids and Rocker box, then you must love the narrative of Nielsen buying all three of these boxes and putting them inside the Nielsen house. If aggregation works, then this should too.

38:39 Well, here's the thing there. There's interoperability between the systems. That makes sense to me. Not necessarily the Nielsen one, because I still don't understand Nielsen's core business the way that I don't understand, like, a lot of those measurement companies, core businesses that were, that were the kings of linear. Even though you've explained it to me like multiple times, like, they're just still somewhat cryptic to me. They're still somewhat cryptic to me. But the notion that you have doubleverify, that has tags and advertiser campaigns blocking things or measuring viewability, and then you have Rockerbox, which is an mmm, and an attribution system, and then you have side bids, which in theory optimizes to the outputs of the viewability tags and the MMM attribution system.

39:25 Those three things in concert make a lot of sense to me. They make sense to me. That's a closed loop now. It's not an independent loop. Right. It's not a, like, like, not a, A neutral third party, which I think is what a lot of people were saying. The reserve. I've seen two sets of reservations about this deal. One is that Nielsen is not an activation in any way, shape or form and they are a neutral third party and that's their job.

39:51 Well, that grace note. I don't know. Grace note.

39:53 Oh, yeah, Grace note. Yeah, yeah, Grace note. Like the, the lyrics and the, the grace note data is integrated for, like CTV and shows. Right? Like, for, for like.

40:05 I think they do some added service too, but they might be moving away from it.

40:09 Oh, interesting. Well, but people were concerned about that. They were like, if you're Nielsen, your objectivity is really important. And if all of a sudden you're not really objective anymore, if you're a media activation partner, you have every incentive in the world to make your activation pathways look more effective and make other activation pathways look less effective. And that's not desirable in an, in a, in a measurement partner, whatever that means. But the other one, the other reservation that I really thought was interesting was Jay's, and we were talking about it before the show, and he was like, this sounds good on paper.

40:45 Freedman of Choreograph.

40:46 AI Cartograph, sir. You're welcome, Jay, you're welcome. But, yeah, I think the other. But his point was they talk about synergies right. Between these two companies and that they're. They're selling into the same businesses. And he was like, at a lot of companies, the person that Nielsen is selling to and the person that Double Verify selling to is in a different building. And just because you're selling to the same company, the same organization doesn't mean that there will be sales synergies.

41:21 Because if they're selling to people who are in different enough sides of the house, it doesn't really help that much, I don't think.

41:27 Yeah, my. I see. But Jay is thinking of new sales. And the way I think about Nielsen and Double Verify is that their compliance standards, an old one and a new one, most of their money comes from existing sales and licenses that have been going on for years and years. And. Yeah. And the goal is not to lose them. So by getting someone in the other building, it makes them more likely to hang on a few more years because now they have two hooks into the.

41:58 Into this agency, you know. Into the. Yeah, into the agency. Yeah.

42:02 And that's awesome. And the other. And a lot of the headlines were like, this is how Nielsen gets into digital, because they weren't really in digital in a meaningful way. And so like, they, they reinvent themselves in digital. I'm still fascinated in side bids and all of this. I don't hear from sidebiz that much. I was on a panel with the side bids lady.

42:18 Yeah.

42:19 And she was lovely. But I. I don't hear about that much. And I don't know. I don't know if it's a. An IIS public a situation where it just kind of like, runs on its own, but, like, doesn't get a lot of love and just has, like, it has their current clients, but it doesn't really conquest much.

42:36 Most of their PR was the founders and the cmo, and I think they're all. They've all moved on.

42:43 Interesting.

42:45 So they did, they did an advertising campaign within Double Verify was called let's Let Cyvance Cook. Did you see that? Japanese.

42:53 No. That's funny.

42:54 Let Cyber Cook. It wasn't that good. And we haven't heard from them since. Yeah, we heard. Yeah. We used to run into the market all the time, like there were arrivals and people would be deciding between us and them. But it hasn't been happening lately. And you know, one of the reasons we changed from custom algorithm Chalice custom algorithms to Chalice AI was to get out of the same category they were in. Because we had the sense that we heard that they were, you know, doubleverify was giving it away, had kind of degraded the value of it by making it like a free add on.

43:29 If you sign on for more Double Verify, this is not the way we think about it.

43:35 Oh, that's interesting.

43:36 That is in. I got to give a plug to Chalice COO co founder and my wife Ally Manning on the Ad Exchanger Talks podcast. Please listen to that if you haven't. Besides the story about changing our name, which was a good question from host Allison Schiff, she tells a story about the beginning, the dawn of artf. That's. That's my story as related to her, but a little more dramatic than the way I laid it to her.

44:03 It's a very good telling of a sidebar between Andrew Casale and me a few years ago. That was one of the factors in the prototype for what became artf now the most adopted IAB Tech Lab standard since header bidding, if I may surmise.

44:22 Yeah, I think it is

44:26 not great from Jonathan Savage is not great. From my experience, they've been trying to pivot into social YouTube instead, which makes sense based on what you said. If the market thought Double Verify should be getting money from social and YouTube, they would just. It makes sense they would assign that to Cybits, but it's a tall order. They don't want to be the wall Gardens do not want to be third party optimized. So you. There's a path, but it's against their will.

44:52 Oh, I. It's even possible how.

44:55 Yeah, like we do location optimization. You could send audiences. You know, there are. There are scraped data sets you can work with on both to try to improve things. But you know, they're all, all those paths are, you know, have mixed success rates.

45:11 Sure.

45:14 Our success on the things we do in the walled gardens is high now it's like 80%. So we feel good about it, but it's limited compared to what can be done in programmatic.

45:21 Yeah. Interesting. Do you do a lot of YouTube buying?

45:25 We don't do any YouTube buying, but we're doing. We do a good amount of YouTube location optimization.

45:32 Interesting.

45:33 So we use the structured data file to split a YouTube buy into 10,000 locations and then allocate across them.

45:38 Yeah, got it. And so you'll allocate inside of DB360. So when I say. When you say you are doing buying, it's just because DB360 is doing the buying. But you're doing constraint optimization.

45:47 We have an allocation algorithm on top of it and it's usually mixed with meta and a programmatic dsp. So you know, it's optimizing the total budget by location. So it's nothing any platform could do on their own.

46:01 Wait a minute, I didn't know you guys did this. You just described them like the meta bidder that we've been talking about for a while.

46:09 Yeah, just for location. Yeah, read, read. On my blog, the one that dropped this week.

46:15 I am halfway through it. I'm just at the beginning.

46:17 Location. Yeah, yeah, that location optimization algorithms, they work for share of search, they work for retail, sell through, they work for awareness lift. You know, location is so poorly optimized by the platforms, all of them, that if there's any. If you're a big, an enterprise customer with any sort of physical distribution or regional footprint, you can, you can gain measurable value by doing location optimization.

46:46 Oh, that's so cool. But you're saying you guys do it algorithmically across all three at the same time. And you centralize performance measurement. So, you know, oh, we should shift more into YouTube, we should shift more into Facebook and we should optimize internally by these locations. That's sick.

47:02 We've been doing just the locations, but if someone asks what you do between them too. Yeah, but just by location. Just a successful case study of doing it in England, you know, so it's especially powerful in Europe where there's no IDs anyway, so if you, as long as you could optimize something, you have a good play for, for Europe. But anyway, that's a Hershey case study that we made a lot of hay with. And now part four of my blog series.

47:26 The two architectures, the location. I forgot it's called the location of something is about that. So we have a lot of public material and this. Yeah, this thing. I said 80%. But our allocation, regional allocation optimization has worked every time we've tried it. 60%. It works every time.

47:48 I really think that there's a future to dynamic allocation between YouTube, meta and programmatic. That's really hard to read, if you were to ask me.

47:59 It's really hard to read. Right. They're so secretive about what they're doing. If he knew what they were doing.

48:05 Yeah, no, because you're. You have to optimize to something else. That's optimizing. That's probably optimizing something else. It's the way their system works hard to make it move.

48:12 They always have to give you a mix of good stuff and garbage. That's how their systems work. Right. And if you had a way to find. Know when they're giving you good stuff and when they're giving you garbage, you could just move the money accordingly.

48:24 Oh, man. I had. I. So I didn't put it in. In our notes, but I had a really funny Twitter interaction yesterday with someone where I. I tweeted an anti packaging tweet because I am in general anti packaging.

48:39 I do not know what you're talking about. I thought you were talking about packaged goods. I was like, no, you like the bulk buy when you're like whole foods, when you just scoop the almonds into the bag instead of packaging them.

48:49 No, it was all a metaphor, Adam. It's all in one extended metaphor about programmatic versus direct sales. That's really what it's about. And direct sold, and they package. Right. Because publishers, if they can, will package. They want performance to just be good enough, and they want to throw some trash in there. Ooh. Jonathan says Smartly is trying to do it predictive budget allocation. Yeah. Going between Facebook meta or Facebook, YouTube and programmatic. That's. Yeah, that's really good.

49:19 But Smartly, the. Which one is Smart?

49:24 I don't know, man.

49:25 Smartly is the one that has the former head of Publicis at the helm. All right, never mind. I can never tell Smartly from a company with the same similar name.

49:35 Yeah, I mean, yeah, you know, they certainly have a website, but the packaging discussion, I'm not letting this go because some guy really argued with me about it. What I mean is, there are two ways to think about him. The guy was like, well, you know, like, I want to buy a package because I don't want to.

49:58 He.

49:58 He described programmatic as getting the carcass and buying direct and getting a package as eating fully mignon. I don't know who this person is they were going with.

50:11 You are. That's like you're arguing with a dumb person.

50:14 Hey, you know what? I appreciated that he. That I. I love all engagements, so please argue with me all you want. And my. My retort was that, well, actually, packaging is they sell you the entire cow and they charge by weight. Right? So you're getting a bunch of stuff you do not want to Eat. You're getting the intestines, you're getting like, like all a bunch of things. The hooves. Yeah, you don't want to eat these things.

50:41 And programmatic is a butcher going through the cow and bite by bite curating a meal for you. I called it an omakase. So programmatic is omakase where bite by bite you're picking what you want. Or someone who is really smart is picking what you want and packaging is. You're still getting a Wagyu account. Right? Like, it could be a very high quality. This is so old, but you're getting the food.

51:04 Jeff Green had a speech about this. It's like, would you rather give the kid $100 to buy whatever he wants from the supermarket or would you rather give him $80 and he keeps 20 and he buys exactly what you want from your list? And it's like, yeah, of course. $80 to get what I want. $100 to get whatever he wants is ridiculous.

51:23 It is. But this, this argument is renewed, man. This is a lot of the agentic discussions are extensions of this precise argument where there are.

51:32 It's all going to fail. It's just all going to fail.

51:36 No, I don't think so. I'll tell you why. I'll tell you why I don't think it's going to fail. It's going to fail against implementations like yours. It is going to succeed against lazy programmatic implementations where they end up like, remember how much of the market was mfa?

51:52 I know, but that's going to be a shift within Google. Right? A lot of lazy DV360 things are going to move to a lazy Google buyer direct thing. And everyone else who's trying to do direct is going to fail. That's what I ranked Way Cat.

52:05 That is certainly a possible outcome. There are lazy buyers outside of DB362 though. There are lazy buyers outside of D3360. But yes, I do tend to agree. Buyer direct is a terrifying creature for people who believe in packaging. It is, it is.

52:21 And I don't want to be impatient or dismissive, but the argument about this packaging argument is over. Because there's 2 trillion dollar companies in ad tech and they're the ones who sell impression by impression. So much more value is created impression by impression. Otherwise, you know, Disney and Paramount would be competitive with YouTube and Instagram and they're not in the same sport as those guys.

52:48 No, it's not the same sport. I completely agree. I think that's a really interesting way to describe it. Biddable media is a different sport from media sales. Like, it's a different sport. They're different things. Maybe selling to the same people, but they're different sports.

53:06 Yeah. And I mean, the argument is also settled in finance. Like the reason you buy funds is because they're safer. No one argues that funds make more money than good stock picking. Like, obviously not.

53:19 Well, they might.

53:21 They make more money than bad stock picking. Right. It's settled.

53:25 Yeah. I, I certainly think that the long term view will go towards things that work better. I really hope it's the case. I really hope. I think it would. I think it's a profoundly cynical view to think that if we go back to packaging, it will work better, period. I think that is a.

53:46 Nobody really thinks that.

53:48 Disturbingly cynical view. I don't.

53:49 Nobody thinks that.

53:50 I don't know, man. Some people do. They get me. I would agree. I do not think people who think that have any clue what they're talking about. But there is a whole class of media buyers who logged in to DV360 and hit go and compared that to how it performed against Facebook. And just think, oh, Programmatic's all fraud. And they're joined in their chorus of woe by all of the publishers who used to sell 90% direct and now sell 10% direct and go, oh, Programmatic killed my business.

54:26 And those are our naysayers. Those are the programmatic naysayers. The chorus of those two groups, I'm not going to call them deplorables again.

54:36 You could be safer. You could be safe and buy packages from reputable sellers and fire them if there's too much garbage in your package, that it's not going to go away. But I think the argument we're having now, like Agentic for guarantee or Agentic for real time, I think goes away. I think they meet, they come together. We'll be doing it in Pubmatic probably by the end of the year. So the end user who doesn't want to think about it and just buys a package agentically that might be real time optimization underneath it to make them happier.

55:10 Well, this is the necessary outcome of Agentic, in my opinion, is it just becomes agentic selection of strategy. It's not like, you know, we're automating direct sales and we're upending Programmatic. It's like, no, no, we're just going to let the lazy agency person say, hey, pick 10 strategies, go with the one that works the best. Agent and Programmatic in a DSP using XYZ settings is one of those strategies. I think that's a lovely outcome.

55:41 I think that's an amazing outcome.

55:43 That's quickly. I mean, because the agencies were all there. AI claims like, well, that's what they're building is simple allocation tools. A bunch of strategies will live and set.

55:53 And you know what? Agents are going to test new things at a scale that humans never, ever, ever.

56:01 I hope so.

56:01 Humans were corrupt. Humans. Oh, gosh. We're going to try this new vendor because they're taking me out to lunch.

56:08 That is a thing like.

56:10 That is how deals got done for decades. And agents will just be like, is it interesting? Does it have, like, is it credentialed? Is it compelling? Well, we're gonna put some budget into it, and if it works, we'll give it more. It's the best case scenario for programmatic. I got going on this and everything.

56:28 Snowflake for the Taking me to the US Open this year. I'm looking forward to that.

56:33 Make it rain.

56:39 You're not going. You're going to the US Spot.

56:41 That's an earned spot. Snowflake. That was. That was a good investment. No, I never get taken to anything.

56:48 The U.S. open is like half ad tech. The early rounds of the US Open, it looks like, can. Like, it's everyone.

56:53 That's. It makes sense. Yeah, yeah. It's New York is advertising, man. Like, advertising is evolving into ad tech. Eventually, they will be one. Right? We should take Garrett somewhere. I feel bad now.

57:08 No one takes Gareth anywhere. I'll take you somewhere. Let's take Gareth somewhere. Yeah, I'll get you. I'll help you out.

57:17 That's fine. It's dangerous for me outside. I don't know if you can see this, but it's sitting in stands with sun on my head. This is.

57:25 They have hats.

57:26 There's a lot of sunscreen. That's true.

57:28 Also, US Open is really nice to go at night. Hopefully I'll be going at night to Snowflake.

57:33 Oh, cool. Yeah. So, yeah, I'm. Oh, David says we have a postmortem. And I have a hard stop at 11 today, so let's. Let's hit it. Let's do the postmortem.

57:43 Okay, Here we are. And if you fact. If you look. Come to think of it, I love this interim.

57:50 It's so good.

57:51 I gotta look at this myself.

57:59 Wow.

58:00 Wow.

58:02 What a. What a view.

58:04 Such a salesman.

58:05 Such a salesman.

58:07 Post mortem, the segment on ad tech, Ad Talk, where we talk about the most notable social media ad Tech posts of the week. Oh, I'm glad we got Kakowski.

58:16 Yeah, I'm glad we only have two minutes for this one. I've been going at this on the Internet in a Laura O'Reilly post.

58:25 Yeah, the beginning of the story is.

58:28 Yeah, explain it. Yeah, explain it.

58:31 Our friend Jonah Goodhart, founder, legendary founder of moat, posted a case study about serving ads to AI agents. Gareth, you seem to understand it better than me. I have nothing but love for Jonah, but I confess to not understanding what the ads for agents play is. And then William Kakowski, a former Chalice machine learning engineer who I have other things in addition to love for now, Will's great. He's a machine learning engineer who's very extroverted and able to communicate.

59:03 He's a joy to read on LinkedIn wrote this very challenging post which I think Jonah replied to. So that's the.

59:13 I haven't read the reply.

59:14 You fill in the content. You can fill in the content about it.

59:17 Well, yeah, the thesis here is that publishers and buying systems detect the user region of an LLM and serve ads with content that they want the LLM to index. And in this way you can surface your contents to the LLM many, many times over the course of the Internet and over represent certain thoughts in the mind of the LLM by using ads to content inject it.

59:50 Okay, and what's your. And let's get your response and then turn to Will's.

59:57 So my response is that this sounds like black hat SEO, right? This, this sounds very similar to me to when people would put text all over their website that you couldn't read. Right? It was like grayed out or like the same color as the background. Try and you can still see it sometimes to try and trick Google's crawler, which is the same sort of thing that an LLM is to think that these words are true and are associated with this page.

60:26 It's very, very similar, right? We're trying to inject something into the LLM to modify its outputs. And I just cannot imagine a world where this works because it will just do nothing but start a war. If I'm an LLM, I'm like, these people are trying to manipulate me. Like, this is not the content of the website. This is dynamic content being served to me. I don't want this. And I think that Will's message here is, we already know Claude hates this.

60:55 Claude gets super pissed off when you do this to Claude and the other LLMs will do. I just, I don't so much have Strong feelings about it as I think that it's a nothing. Like, I think this is a, this is a nuisance to the LLMs. This is not a business. This is a nuisance.

61:13 They should be smart enough to just ignore it.

61:16 I think they will. I think. And then you're in a war of how do we outsmart how they ignore things.

61:21 It's just like black hat SEO.

61:23 It's very similar. And then the, the only argument that I've heard that's like half sound is that, well, publishers don't get paid by the LLMs, so like this is a way for them to recoup money. I just think the part I don't

61:36 understand is, I mean the LLM already has access to the information, so what are you adding? All right, I'm going to read Jonah's response.

61:44 Yep. And then I gotta go.

61:45 I think the test and the product are measuring different things. Your methodology tests whether off topic content survives retrieval. And that is not what we do. But we serve as brand verifiable facts. Sponsored or closed in the first line. Every claim cited to a source. Advertiser approved, publisher authorized. Placed on contextually appropriate pages. Agents neither ignore nor avoid ads and favor structured data. Effectiveness is testable, which is why we measure it. Our approach is able to lead to a 90% confidence.

62:14 Read on. Changes in brand visibility, though it's early days and you know, happy to take you through it if your mind isn't made up.

62:22 It's gaming, man. It's gaming. You're gaming the LLMs to increase brand visibility. That's one of these.

62:29 I go back and forth a few times. It's kind of interesting if you want to read it, but I will read it.

62:34 I only saw Will's first post. Just, I, I, there's no world. I, I will not be convinced that this is not gaming. And I just, I've, I've been in the SEO game, I've been in the black hat Internet game long enough to know gaming when I see it. And like that's what this is.

62:49 I just, yeah. Will is now challenged, openly challenged. Brian o' Kelly and Jonah Goodhart on tech.

62:57 So got a boy, but he's not getting any angel investment checks. But you know what, Will? I didn't get him either. So. Godspeed, buddy.

63:03 Yeah, he's a great up and comer to watch and I think he's working on a stealth startup. So we'll see. Yeah.

63:11 All right, guys, I gotta go. Thank you so much for joining.

63:23 Darn.

63:23 That's the crowd is back. The lights are dim, front rows buzzing. It's about to begin.