Hands on the Contract, Eyes on the Bag

0:00 10 [music] tech talk it's a live [singing] Q&A buy side clearing the way CEO spitting truth no delayal >> 5 4 we've gone for main engine start we have main engine All right, >> ladies and gentlemen. Attempt.

0:34 [music] [music] by the sound. [music] [cheering] >> Yes. Yes. Folks, welcome to Adtech Ad Talk. [music] >> Happy Friday.

1:05 >> It is Friday, June 12th, 2026, noon Eastern time. I'm Adam and I represent [applause] the buy side. and I'm Gareth Glazer and I represent the cell side. Welcome everyone. Thank you much for for joining. Anyone who's here, please ask us questions. We're here to talk to you, not just to each other. Even though we can we can do plenty of that, I'm not worried about it. Um, but it's been a relatively quiet week until like 10 seconds ago.

1:32 >> 10 seconds ago and now we're amped. Now we're amped. >> Now we're amped. >> This is we're the first podcast to convey this news [cheering] in a southern rap song. David Maldo, please cue the song so everyone can hear what just happened in Planet Ads. >> Uhhuh. PUBLICIST. >> Come on. COME ON. COME ON. COME ON. COME ON. COME ON. THEY WAS OUT. THEN THEY'VE BEEN [music] AART. TOO MUCH PRIDE in the room. [singing] Too much heat in the dark. One side talking deals. One side talking trust. [music] Same old table.

2:07 [singing] Same old stain. Same old dust. Now the phones got quiet, then they lit right up. I SEEN THE TENSION DROP. When the money [music] called us, you can hear that room. When the eos crack, now they back in the lobby trying to take it back. Hands on the contract, EYES ON THE BAG. [music] No more side eye, baby. No more drag. Close that wound up. No, let it fade. When the beat go boom, WE MEET HALFWAY. PUBLICIST AND [music] TRADE desk back in the room. Back in the room.

2:36 My publicist and TRADE DESK. SHAKE OFF THE FUME. SHAKE, SHAKE OFF THE FUEL. We being [music] through the fire. NOW WE PASS THAT TEST. Publicist and trade desk. Back in the room. Back in the room. [music] Talking. Yeah. Words got sharp. Whole crew froze up when they split that chart. But the streets don't care about a wounded pride. They just want [singing] the winners on [music] the same damn side. So they roll back in with the chest out stiff. New handshake. [music] Firm grit.

3:06 [laughter] If you got a seat, then pull it near. Cuz the [music] loudest room is the one that's clear. Hands on the contract, eyes on the bag. No more side [music] eye, baby. No more drag. Close that round up. Let it fade. When the [music] beat go boom, we meet halfway. Publicist and trade desk. Back in the room. Back in the room. [music] Publicist and TRADE DESK. SHAKE OFF THE FUEL. SHAKE. Shake off the fuel. We be through the fire.

3:34 [music] NOW WE PASSED THAT TEST. My publicist and trade desk. Back in the room. Back in the [music] room. It wasn't pretty. It wasn't clean. But [singing] both them names kept the same machine. Let the [music] old smoke rise and let it go. If the bag stayed full, then you already know. Now the [music] whole flow shout out. Let it stand. [music] Two big players back hand in hand. [music] I [music] can trade this back in the room. Back in the room, I can trade this. Shake off the fuel.

4:15 [music] Shake. Shake off the fuel. We keep it through the fire. Now we pass that [music] publicist and trade desk. Back in the room. Back in the room. Hey, [music] black back black kick in the room [music] and trade desk. >> Wow, I love AI.

4:45 >> I mean, this the street doesn't care about their feud like >> Yeah, the street smoking. >> Yeah, I mean the [applause] other street didn't like their feud either. Like none of the streets appreciated their feud >> and now it's over. >> And now it's over. I mean, >> I hope somebody saw that. >> It was a It was a rough It was a rough feud. It was like a like a just a spree of like bad things happening at the trade desk and like now it's over. I feel like this is this is the beginning of an upswing. I'm I'm hopeful for them.

5:20 >> What's next? >> What's next? Will we see Brian Okelly and Jeff Green hanging out at Cam? >> No. >> Sharing some wine? >> No, I don't think so. I don't think that happen, >> right? >> I mean, things change. You know, >> things change. I think uh I've been thinking a lot how people underrate uh the chance of change or the certainty of change. like we're made to >> sort of uh assume that things stay the same and extrapolate off of uh the how with how things are familiarized and I'm thinking a lot about change.

5:56 >> Yeah, I think it's fair. [laughter] >> I think it's fair, man. >> Just because the way things were done and just because the companies that were like were how things were done the past 20 years doesn't mean it has to continue. You know, one of the thesis that failed a little while ago was DSPssp mergers, right? Like there was like a like a whole laundry list of failed DSPs mergers where it was just disastrous. I'm not sure it'd be disastrous in 2026.

6:27 >> Like >> what do you mean? You mean like um Magnite and P Magnite and Pomatic have DSP SSP combinations? You think now they're going to succeed with that? Oh, but those acquisitions well one which which Magnite DSP SSP combination the first one was Rubicon Chango they bought Chango for like $80 million something like that and it was like they wrote off the entire thing and shut it down. Um >> didn't go well >> and went terribly. clear line. They've built CTV has kind of like changed the landscape a little bit because I think CTV there are huge advantages to being really close to the supply on CTV because the uh the price landscape is so freaking ridiculous like frustratingly asymmetrical and confusing where like oh why can I get this thing for $4 here and $10 here and $50 here? I I can't wrap my head around it. But I think that's why Clear Line had success is because Clear Line was like, "Well, you want cheap inventory? We're we're pretty close to the source. We'll help you get cheap inventory." Um, but prior to that, they were there were failures. And then I guess Pomatic bought Martin. Um, but that was really quiet.

7:36 >> It wasn't a huge deal. >> It's called has a different name now, actually. >> Has a different name now. Um, and you're right. I think that that one has been like relatively successful. Like I think that people do use it. Um, but it's it didn't cause the the giant shift of oh, now Pubmatic is a a buyside player. Like I have I have yet to see Pubmatic mentioned in a a Forester wave DSP. >> But why are you saying why are you [clears throat] saying that strategy can succeed now? What has changed?

8:08 >> I think that the perception that there's a buy side and a sell side and there they're so distinct and there are these two halves of the industry is going away. Um, people are realizing like, "Oh [ __ ] I can actually make everything work way better if I don't have two systems that have different everything trying to figure this out. If I put everything in one place, it works better." Um, which is very much the the cellside decisioning thing. The reason it works so well is because everything is just cleaner when it's in one place as opposed to the throttled secondhand information madness that exists in all DSP, SSP integrations. Um, and I I think that a big part of it is like a change in understanding. I'll never forget had a conversation with a very senior person, a very large DSP at one point about Gamera and about working together and about investment and about all these things. And the response that I got was, well, you sound like you're a sellside company and we only do buy side things, so we're not interested.

9:08 And I just remember my instinct was like, God, what a what a what a limiting mindset like like if you like if you're a religious zealot for I can only be a I'm a byside person. I only do buyside things. And you can still see lip service being paid to it. By the way, like if you look in a lot of press releases from DSPs and SSPs still, there'll be this funny little koda at the end where they'll be like, "Even though we're doing this thing that has to do with the other side, we still represent the interests of X." And like they always say this because they feel the need to say this, but I think I think it's it's a cod for a reason. It's now it's like at the bottom of the page instead of leading where the original pitch was we are a DSP, we represent the interest of buyers. we are an SSP. We represent the interest of sellers. And that was the the headline. Now it's like, uh, oh, okay, we're building things that make advertising more efficient. And by the way, oh yeah, yeah, we represent the interest of buyers. Um, because I just I I don't think that the uh we we've come to realize that if you actually care about optimizing outcomes for advertisers, one, the better they bid, the more they bid, like the more money publishers make. An efficient market is how publishers make more money, not like fragmentation. And two, it just works better. Like you win tests. Chalice wins tests. Chalice probably drubs competition who are running through like traditional DSPs where it's like we can just do so much more. We can integrate this. So >> I think part of that is because we're we're doing a single advertiser. We're doing single advertiser optimization. So we can't be tolerate any arbitrage opportunities. Like if there's any arbitrage opportunity, we just deliver it to the advertiser. But if you're, you know, I means what I mean is if there's a big spread between the value and the cost, then then that's great. Then the advertiser gets a lot of value for their cost. But just because these things agree doesn't make uh, you know, arbitrage disappear, right? It's there's still the question of, right, whoever is doing that DSP, SSP thing is going to have great arbitrage opportunity.

11:16 >> Yeah. >> Right. And you're kind of assuming that if these things are united that it's better that it will be delivered. I guess it's the same thing with containerization and single advertiser. We you know an entity could take that value for themselves. Um >> well and so my uh my my governing thesis here is that spread harvesting right that should belong to the advertiser and I have no problem with an advertiser harvesting that spread. And this is this gets on the article that I was talking about. It is the advertiser's god-given right to harvest that spread because that spread is a deal, right? That spread buyer, >> that's being a smart buyer. That's your job is to find that spread. That is your goal in life to find a big difference between cost and value as a as a trading desk person, as an optimization algorithm, as a buyer, that is your job.

12:07 If that value acrrues to them and they keep buying from that publisher and and the advertising campaign performs great and everything works, wonderful. The problem is when the spread goes to someone else. The problem is when the spread when it's when it continues to be expensive for the advertiser and the publisher doesn't get the money. Um >> yeah, >> that's bad. That's an inefficient market. There's another problem and it's in the scenario you describe where it's maybe it's a you know a very good actor is doing DSP SSP and they want to deliver spreads if they're optimization is pulled it's actually becomes very hard to take advantage of of value because what happens is you get this bunching of you know of impressions everyone wants.

12:56 >> Yeah. >> Right. And the value comes with getting the impressions other people don't want. So there's like an inherent we're finding Chalice is finding and I'm gonna try to I've been writing papers on it that will be available at least to some people that uh it's in a pulled stack. There's always going to be all these arbitrage opportunities left on the table. And I think it leaves open the question of of of who who gets them. like you could switch to single advertiser and get them yourselves or you could be like Facebook and and deliver more value, right? But Facebook could do something that a DSP can't and just spreading the advertisers out onto those opportunities, right?

13:37 Because they have total control, right? >> So if if they see people bunching up all wanting the same, they could do the value cost calculation and say, "Oh, all these cheaper ones are good enough for you." And spread them out methodically. And that's a huge platform advantage that Facebook has over DSPs. They could actually deliver more value through this spreading out of, you know, the good enough inventory for what you're trying to do. >> And I think I think even in the what you're saying is a nirvana of buy side, sell side, comedy, right, and technical agreement. Uh they can't actually act like Facebook unless you take all the money and deliver all the all the arbitrage opportunities, which is just not going to happen.

14:18 I think another component of it that causes that bunching is user ID based buying. >> Yeah, it's a that's a it's a great example of pulled stack decisioning, right? It's exacerbates it. Like if you're only looking for certain opportunities and those opportunities are only within your ID footprint. >> You're going to have real imbalance of bids. You're gonna and you'll over bid on those ID ID opportunities compared to everything else that you have trouble valuing. >> Yeah. And it compounds, right? The issue compounds with pacing where you have one person who's super overvaluing a given ID and all the ID matching companies who are have every incentive on earth to send as much of those IDs as possible.

15:01 >> Yeah. >> Where there's a a 10 10% of the IDs that pass through bid requests, right, like belong to to those people and like they're >> Yeah. Yeah, >> they get all the high bids. And if and if you're on the and if you're in the middle, right, and you're monitoring clear rates and you're monitoring bid rates, you're like, "Oh, this ID tends to have a huge delta between its bid and it's clear." Um because you can see right what what the the final thing is on the page. And I'm not going to give away too much about that, but um [laughter] there's but there's huge opportunity there for companies and the the middleman companies who understand how this [ __ ] works can take advantage of that. And that's really sad because publishers don't know this and won't know this and advertisers are just like put money in [laughter] put money in money blending >> smart. I think that they are too. I I have every hope that they will but right now it's put money in money blender hit hit go blender and then smoothie. Um >> the way to get this story this is a this is the great untold story of adtech of the last seven years. The headline is AT&T at Apple AT is the best thing that ever happened to Meta because of the way they reacted. It got them off IDs. And the story of how Meta reacted and improved without IDs is in the papers and the investment calls where they've talked about Andromeda and their other models. And what they do is they essentially perfected their value function using all the data they have.

16:30 >> It's so cool. It's so smart. Wow, >> it's so interesting, right? From an ID base to like, all right, let's use all the all the data we have to predict value per advertiser and made the most advanced recommener system, meaning when they choose which ad is going to serve, it's taking so much data into account and acting so fast and they've said like the improvement is like 5%. So, they're better than they ever were with IDs, only 5%, but that 5% is worth like $50 billion.

17:00 >> That's huge. And also it means that they don't have to worry about ID stuff. Um it's it's a double whammy. And I think it comes down to the number of features that they're evaluating. I see this in deals today. Deals will buy based on one or two parameters. That's it. >> And there's just a a lot of very unsophisticated buying that happens out there. >> Um and >> yeah, >> how is how is that going to compete with Facebook, right? Like >> yeah, that's another thing people just don't understand. Like if if you look at a company and it's all engineers, doesn't matter how good these engineers are, if they don't have data scientists, they're not doing the sort of modeling that we're talking about. It's a different discipline. Engineers will always flatten out the data science piece so that they can get it to to to scale. And that's why you're seeing what you're seeing is like all these flat models that were really good for scale because the engineers weren't thinking in terms of the statistical, you know, the advanced statistical pipeline that you need to build. You have to start from scratch, right?

18:05 That's what Facebook did. >> Yeah. Well, we're doing it right now, Adam. >> It's a big job. It's a big job. >> Think about it every day. I think about it every day. >> And I and and you know what? You know what else we could talk about? Like what's going to happen next? I I started I started my company in 2020 with a glimmer of what's going to happen next. And the way things have played out now in the summer of 2026 headed to can I know exactly what's going to happen next. It's like one of one thing is going to happen next because all the other >> I'm so excited now.

18:37 >> All the other things have it's it's not it's not that great. >> What's your one of one? Wait, what's what's not that great? Tell me what's happening next. Everyone who's promising to add value from AI, except for one or two companies, the companies you can count on your hand, are going to fail, >> they're going to fail miserably because they're promising promising things that require a fiveyear build. >> And they're not only are they at year zero, but they haven't started.

19:02 >> Yeah, >> it's going to be there's going to be carnage. >> That's that's what's going to happen next. I feel like Linda Hamilton in T2. Uh like like I see it. It's all happening. Like there's like there's people with big egos who are essentially going to be uh facing tough times. Say >> yeah. Um which classes of companies do you think are in the most trouble? Who do you think of when you think of this?

19:32 Are you thinking about agentic companies? What what's what do you think the carnage hits the hardest? Uh agencies >> agencies >> yeah I think I think agencies have not thought hard enough about what do we offer that an advertiser can't do themselves in the age of AI and on the advertiser side there's heavy AI adoption uh and and it hasn't necessarily hit marketing right there there's heavy AI adoption in things like supply chain optimization >> and all the marketing stories are like oh we're dabbling with creative. But no, this is a tech and finance thing. This is coming in from the the CIO and the CEO, the CFO, and if the CMO isn't up on it, they're they're going to get mowed down. Like, this is this is like there's guaranteed ROI in certain kinds of AI adoption. And it's not going to be like, oh, but my agency, oh, but my a it's been seven years of, oh, but my agency is building something for us. And that the bill's just going to come due. the agencies are not doing well. Even publicists, their stock is down at like double-digit percentages over the last over the last year. Um, so it's not just the hold codes. It's it's it's none very few people at agencies have thought what do we offer that the advertiser can't do themselves. And uh and I think it's uh the best reading of publicist acquisition of live ramp is like well if we're going to offer something that advertisers can't do themselves how do we make sure that a critical piece isn't taken away from us.

21:12 >> I think that's that's how I would say the strategy of of of publicist is is they're thinking of doing something advertiser can't do themselves. maybe some advanced data science on per advertising at it out of their CRM. And somebody said, well, if we build all this and it's a five-year build and whatnot, how do we make sure we don't lose a critical piece? And someone said, well, if you buy Live Ramp, you could build it out of Vive Ramp and then it's sustainable.

21:38 >> Yeah. I think um a good horistic for that is uh what will Claude never do, right? >> Yeah. >> Like what will Claude never do? >> You could ask Claude this. This is not a tough question. You could ask cloud like say everything the agency does and ask cloud and it's like well I don't touch the execution layer. He'll come right out and say where the there's probably value >> in the execution layer where the ads are actually bought and I don't touch that.

22:04 >> Yeah. Like like cloud's not going to be live, right? Cloud's not going to create that internet connectivity as a result of a prompt. It's just not [laughter] it's not a thing. like there's >> the the the the business development like the the the integration work involved is too too dramatic. >> There's so much trial and error in this build to get from models into live to live stuff and the feedback flowing and improving the models. It's it there's you got to build that pipeline and there's no use for any kind of consensus pulled stack AI.

22:40 Everything has to be built ground up. Uh, and yeah, like I mean I I'm looking at LinkedIn and more and more people making the chalice claims every freaking week. Uh, and they're not building it. None of them are building it. >> I um I will say across my clients most recently people have started to ask us for our MCP which wasn't a thing for a really long time. Like we we slapped that thing together a while ago because we had APIs and building an MCP when you have APIs is is a trivial The whole point of an MCP is that advertisers can have the exact same access as the agency.

23:14 >> Yeah. And so for our publishers, we have all these data sets for them. And when I'm like, "Hey, we have these UI." And they're like, "I don't want these UIs. Just [ __ ] give me your MCP. >> I'm going to ask Claude what your data set says, and that's how this is going to work." And Claude's going to build me alerts and going to do these things. And I'm like, "Thank God. This sounds great.

23:34 I don't want to want to build." >> When I was agency side, I was worried about in-housing. I don't think I was worried enough. I think I was in denial about it. Uh I think I underrated the in-house teams that we met. A lot of them were very advanced and remembering [clears throat] the meetings now I thought that my team who was using trade desk had similar capabilities. Now I don't think that thinking about what they were doing with data >> and this is this is a decade ago >> and the inousing projects that are going on now with very very big companies I think I think they're going to do great and that the agency should be very very worried they're nowhere near worried enough like having like an automated guarantee workflow going the value of that is going to be close to zero >> yeah I mean it'll be it'll be held up against things that just perform so insanely well >> like >> the the the value of the automated guaranteed workflow right now is in CTV it's a way to get cheap inventory like that's the >> agency rates or something but I mean if you're a big advertiser you can get those rates >> I would think so and or claude could go get them for you >> the people who are in housing this year are like GM and Fizer and if you think GM and Fizer can't get Pubicus' rates like think again >> well and uh another component of this too like I remember my first exposure to agency world was My my wife was the product lead at a DSP that got bought by publicists and she would she was in the Coca-Cola pitch meetings like uh for the the year that this was back when when Coca-Cola was doing these she's flying all over the world going to all these Coca-Cola pitches >> and she was like they're really smart like like like the marketing people at Coca-Cola >> are really smart like they're they're like they're not incompetent by any stretch of the imagination like these these teams know what they're doing.

25:26 >> Yeah, they're it's not a joke. Like these are these are real teams. >> Even even in my agency days before I knew that many in-house teams, I I felt like you could rank the smartitude of of clients by how competitive their industry is. And CPG is always a night night fight, right? You're you're always on shelf space next to your biggest competitor. It's week to week. There's no lock in. Anybody could switch anytime. And the teams are super smart.

25:52 >> Yeah. Yeah. The other end of the spectrum is banks where there's five of them and they split the money. >> Of course, [laughter] it's like G it's like geob based. It's like we are the bank in Nebraska. >> How often do people switch their bank? It's like when when people move states and they need a new bank. That's about it. >> Yeah, man. Uh oh, that's really interesting. Like the smartest marketers are the ones who need to be really smart um or their businesses don't work. I never thought about it that way, but that makes perfect sense. Like where switching costs are lowest. You have to be the best marketer. Um, so yeah, I that's hope inspiring, right? From a from the sellside perspective, what you're saying is very hopespiring.

26:30 Because I uh one thing that I >> it's good for publishers. And I'll tell you why. Um, >> there are a lot of people in the publishing world who still think that the way that publishers make more money is games. And games being when they see the arbitrage opportunity that exists where a middleman's taking it, they go, I should be taking that. >> Right? like when the advertiser is overvaluing that should come to me >> and um >> you see this you see this like it's like a mafia state right like you see this in Google being the mafia like when people have been dominated by corrupt authority for a long time the only way they know how to make money is corruption and it's not I don't think it's acknowledged enough that publishers being obsessed with gaming is because Google was taking a 100% of the arbitrage opportunity for over a decade >> oh my god and really taking it like talking about this harvesting. They had last look. They had the ad server. They knew the clear price and could wait to see what it was before they did.

27:29 >> Yeah. You don't even have to understand those those capabilities. All you have to understand is like imagine you're trying to auction something a billion times a day and somebody on the other side never over bids. >> Never ever. They get to wait for all the other bids to come in and say, "Oh, what do I want to bid?" Like >> my god. disaster. [laughter] Like >> disaster. >> Just never over bid. Yeah. >> Yeah. I know the advertisers paid me a dollar. We're going to wait and see where this one nets out. Oh, great. I can come in at 21 cents. I'm going to bid 21 cents. Um yeah, very real. That's why they lost a monopoly suit. Um >> but uh yeah, I like I like the idea that it's good for pubs. I mean, what we're really talking about is the arbitrage opportunity getting spread out.

28:10 Advertisers getting some share. Platforms will get some share. Platform advertisers will even get a share. There's ways to do it. Uh, and that'll be better for publishers. It'll just mean they'll compete for for, you know, the over bidders. >> Competition is what drives up prices. Like, competition is what drives up prices. Oh, from Lou. >> Lou is there. Look at that picture of Lou. Look awesome. >> Great picture. >> The number one driver of customer retention in banking is reluctance to endure the inconvenience of switching banks. Yes, Lou is the head of marketing at Bank of America for many years. uh knows well like they spend a lot in marketing and they could be very aggressive but they're essentially competing with four other companies or the small number of people who are willing to go through the inconvenience of switching banks. Oh, interesting.

28:58 >> Yeah. Or people who are looking for a new bank for something. But yeah, gosh, like the switching costs are just so high moving or a mortgage or something. >> Yeah. >> Yeah. It's so annoying because I was going to say, oh, it's kind of like auto, but it's not like it's not like auto at all. Because in auto, you're you're getting a new thing no matter what. There's not really a switching [laughter] cost. It's like kind of a pain in the ass, but it's something you have to do. Whereas switching your bank, you never have to switch your bank. Like >> I have hated my bank. I've live I moved to New York in 1993 and I'm still on the same banking relationship.

29:29 >> Uh and I hate >> Can you say >> I hate them? I will not say because I want them to be a client, >> but I hate them. Their their security features are so intrusive. They act like it's their money and I that I have to prove who I am to get it >> over and over and over. >> Yeah. >> No more salazo tricks. >> Yeah. But I um I think another thing that the reason that is good for publishers, just to close the loop on this because this is an important one.

29:57 This is why this is good for publishers. The way for publishers to make more money is competition. It is the way. >> Yeah. >> Withholding data, withholding whatever, withholding transparency, which is how the ad markets worked and still work in CTV. I was having a talk with my CTO about this this morning. CTV, they still govern ad prices with transparency, which is if you want transparency, you pay more money. We'll make it cheap if you buy it blind. That's those are the rules.

30:28 >> Yeah. >> Um that is how banners used to work, right? If you go into every ad exchange, you can see a button to turn off URL transparency and domain transparency. The buttons are still there. Um because back in the day you would buy from Value Click and Value Click would be like here's your CPA, here's your CPC be like where did I run? They'd be like websites. [laughter] And that's what it was. And uh >> yeah, >> and that's uh that's just inefficient, right? Because that means that publishers who have good experiences are not able to properly harvest the competition that comes from advertisers being able to correlate those good experiences to outcomes. And the more correlation of good experiences to outcomes we have, the more good experiences are bid on up, right? The arbitrage opportunities disappear because if there are five biders who all know this is a valuable piece of impression, there's just way there's less room there, right? Now they're going to try and bid against each other and the price is going to go up. And that's the world publishers want.

31:31 >> I know. >> We are finding insane bargains on quality inventory just, you know, some of it is just from getting out of the out of the pool. >> It's just like it's just they're just blind to it. Like we're finding like >> there's good publishers publishing really good placements and and we're a little surprised how >> Oh, don't be man. Oh, this is this is so the this is why publishers I have this conversation all the time have just had to slap more ads on the page. The reason they had to slap more ads on the page is no one rewarded them for not slapping more ads on the page. In fact, they punished them. They just made less money.

32:05 >> No, it's >> Yeah. The reason I made the data set that I did, the reason I'm out there talking about this all the time is I'm like, reward the good placements. Bid the [ __ ] out of the really good ones. Every single publisher on Earth, and I'm going to use this talking point again. I use it all the time. Every publisher would rather make $10 from one ad than 10 ads with for $1. Everyone >> like it's it's better in every way.

32:27 Latency is better. Ad serving costs are lower. Every single everything about it is superior to have one ad that makes you 10 times the price. >> But the reality is crap video is going for 10 and there's a lot of quality video going for three. >> And that's cuz people probably don't like the domain that the quality video is on for some silly reason. >> It's dysfunction. Yeah, it is dysfunction. Well, there's another thing that um I actually I'm curious on your take on this. So, I see a lot of people talk about deaxonomizing things and I always appreciate this where instead of saying we're going to have a taxonomy for content or we're going to have a taxonomy for users, right? Like users who are autointendenters or X Y and Z, we're going to have a list of attributes of features and we're going to score them all. So, no one is in any given category. They just have propensities for categories. And I really like this. This is this seems like the right way to do things to me.

33:25 And the way that advertising platforms differentiate is what features they have and like what they think is important and what they can generate. Like that seems like a real way to generate alpha to me. >> Yeah. Who who' you hear talking about this? Because there's there's two things. There's doing that and there's activating on it. And just like we're talking about quality pages, like the >> finding out which are the quality pages was pretty easy. Like when sincera was public, you could just get the sincera data and find and find out. And now trade has the ability to get there with sincera data. Chalice has has managed to get there with other partners, but um like I said, like things could be known like the true value of an ID or a page.

34:07 Uh and I think it's only very recent that people can activate them. And as far as I know, Charles is the only one really doing it at scale. You know, I keep saying we're the only >> I do want to call out Cognitive, a company that we respect. Uh, and uh, just the only competitor of ours who I it never really lies in the marketplace. Like I I I think we have a very annoying set of competitors that that lie in public about their capabilities. And I just appreciate cognitive is they're the OGs in the space and they don't lie. I mean, I was gonna say I remember when Jeremy Fain started Cognitive and like >> the company's I I mean I the first time I met him was at a Progress Partners event in probably 2016.

34:56 >> Yeah, it was new then. >> Yeah. Um and they uh they've been beating the same drum since then and they've had to build all kinds of [ __ ] like they built their own bidder at one point because they just had to. They all none of these integrations existed. The things that make our life easy like artf Oh my god. Integrating a custom model with the DSP back then was probably like pulling teeth. >> You know, funny story. I brought them into Tradeesk. Uh I was the I was the customer.

35:27 >> Yeah. >> At Horizon who brought them said like, you know, Trade, you should integrate with these guys. It's really important. I want to see what they can do. And I don't know what they did, but when we started Chalice, there were all these rules and restrictions around touching the multi-dimensional bidding API [laughter] because of something that happened in that incident when they brought them in. Like I don't know if they broke something or they I don't know what I don't I still don't know what happened, but it was really funny to encounter obstacles that I helped create.

35:55 >> Yeah. I mean, it's um it's it's a business that should exist. It's just a good business line. I I got respect for him. I think >> Jeremy Jeremy's co-founder is a real senior data scientist, machine learning guy. >> Oh, really? >> Deep learning. Yeah, they're like childhood friends. It's a good story. They're rolling with Michael Kassen now, which is why they're they're getting more share of voice than us tomorrow. But um you know, whatever. >> Share a voice in like marketing.

36:22 >> I don't know. According to Claude. Yeah. >> Oh my god. Yeah. But uh I want that to be my problem. I'm not worried about share of voice in marketing. We're about to pass. >> We're getting We're about to get a lot of earned media. >> Nothing better than earned media. >> Yeah. But yeah, publishers are going to be publishers have a bright future if they can just stick around, man. If we just let don't let chat GBT kill publishers, >> like internet ads are about to start to work and it's going to make everyone money every day. We get >> also content like the value of of like there's going to be so much AI slop that original content the value of original content is going to spike. it is. Um, now this is I have a controversial take on this topic. So, there's just a big fancy article that came out talking about AI slop and promoting AI slop with um fake girlfriend ads. [laughter] And I want to tell >> it's the ultimate intersection of our market with the porn market.

37:19 >> We got this new technology. We get really close together for a minute. for a minute like [laughter] the the the poor advertising infrastructure is is programmatically enabled that exists like >> you real ads >> credit card you know there couldn't be e-commerce until sites could take credit cards and the first sites to take credit cards were porn sites so all the tech people were coming out of porn sites into e-commerce to teach people how to do >> build pipes you build a checkout flow yeah say here >> and then they dispersed again with different sets of people. But, you know, technologies bring bring us together.

37:57 >> I mean, I will never forget when Seamless ran their advertising campaign on porter sites. I think it was Seamless and they were like the ROI was great. Like [laughter] like these ads worked. Like people clicked on our Seamless ads and then bought food. Yeah. >> Um >> we had an agency team that actually thought up this idea for a beer that was only promoted on porn sites. [laughter] >> Maybe that's the future of agencies. Maybe they become DTOC brand incubators.

38:25 Like that's uh maybe they do a hard pivot and they just start brands like something. But wait, let me back to what I was talking about. Um which is one if it's a human consuming AI slump. I think it's still worth showing an ad to. So, and my logic here goes back to what we were talking about earlier, which is the featurization of taxonomies. I still think that a lot of us operate emotionally inside of identity taxonomies where we're like this good, this bad.

38:59 >> That's how the human mind works. >> How we work. >> Picture a mom. You can't picture all of them, >> right? Picture one of them. And if you and if it's a good one, you like moms. And if [laughter] it's a bad one, you probably don't like moms. Um, so I think that we we get on this thing where we publish like these articles about, oh, look at this horrible thing we found. And it's like, yeah, well, wait, the right way to approach this is the infrastructure that you have that's assessing where you should run ads should simply be scoring these things on a spectrum >> where if you have something to describe, oh, this content was procedurally generated. Oh, this content has like fewer words per page than this or this content has XYZ attributes. That's just goes into your valuation algorithm that Chalice has. And it doesn't necessarily mean that things are good or bad. And a lot of our industry still operates in the world of things are good or bad.

39:58 Like there are so many people that just trade lists of things like the trade. >> That's what I'm calling full stack, right? You could score it. >> Oh, that's interesting. calling it a full stack is nice. >> If it's scor if it's scored for everyone, then [laughter] yeah, >> when you apply it, it's not going to work very well. Or if it's scored for your value function, and the example I always bring up is Coca-Cola in a pitch.

40:21 The way Coca-Cola makes more money is by getting the two a month buyer to buy four a month, right? >> Right. The impression that helps with that is nobody in the market knows what that is. Maybe Coca-Cola does, but I promise you no one else and probably no one at all knows which impressions are worth more or less for Coca-Cola trying to get a two cocoa buyer to be a four coke a month buyer. Uh and and that there's no kind of pulled stack scaling that's going to get them any closer.

40:48 >> You have to get into the data. You have to get into that data and model it. >> Uh so I really like this refrain about the destruction of taxonomies. I think it's a thing and I think it applies to inventory too. I don't I think that buyers want to have inventory taxonomies. They want to think these things are good, these things are bad. There's a huge market for doing that and it's like no no just work with Chalice.

41:10 Chalice has features that will assess the XYZ attributes. >> Let's I have more I have more good news about Chalice. So let's uh let's get to the news the story we both want to talk about. >> Oh, please. Let's do it. You got to leave time at the end to play the crunk song again. >> I love that song. >> How many people are in the chat? >> We got We got about a dozen people uh live it looks like.

41:43 >> Can you just Can we ask how many people heard the Krunk song, the Krunk news at the beginning? >> Yeah. Type in chat if you heard the the song. >> You deserve to hear no reaction. >> You deserve to hear the song. The the song was was amazing. >> All right, this story I think is Ronin Shields. It seemed sourced. It was all anonymously sourced. Uh but Ronin did a nice analysis on uh how the live ramp acquisition is going to change can conversations.

42:16 Uh and I don't know a lot of his quotes amounted to hand ringing and say if Publicus owns this translation layer, what does it mean for us? and didn't really have any answers. Uh I think there should be there should be a nice big dose of ad technology added to this conversation. Uh like what is what was the function of live ramp? What makes it hard to replace? What are the alternatives and how do those alternatives score depending on what you're trying to do? With Live Amp really did a few things. It was a collaborative data marketplace.

42:52 It's clean room and there's an embedded identity layer which is extremely useful because there was a embedded piece of tech almost anywhere there needs to be a handoff of the audience right the big part the hard part of activating these things versus creating some sort of model in a lab is there's a lot of handoffs of that audience is handed off to Facebook it has to be handed off separately to YouTube in programmatic there's really a separate handoff to every exchange and uh live handles all of those. So even if you have something that has a uh you know a bunch of ids, an identity graph as they call it, if it's not embedded for all those places, it's not going to have the same functionality. So you have to think hard.

43:38 >> Yeah, I the I think you really hit hard on LinkedIn that they're going to maintain the neutrality of Live Ramp because Live Ramp's value is its neutrality. I'm glad you >> I appreciate you. Um, my question about that is how much do they care about the live ramp revenue versus if they can carve out 30% market share of current live ramp users who are not published as clients in the act of murdering the other 70% as live ramp customers? Is that worth it?

44:19 Right? Like um is it worth it to maintain live ramp completely independently? And this would this would never be a band-aid, right? This would never be something that happens all at once. This would happen in my mind through the addition of like cascading inconveniences where like publicist customers get xyz things that are nicely integrated with live ramp over time and they just slowly make it more and more of a pain in the ass to have live ramp interoperable with other stacks.

44:54 >> Um >> is that worth could that be worth it? I come from I come from a quantive avenue a Atlas and it's like embedded [clears throat] in my DNA that you can be an agency with tech and offer that tech to the marketplace and one of the executives there executing that strategy at econ how current CEO of live ramp so there's very little doubt in my mind that live ramp is thinking about this is saying yes we can absolutely serve the agency and the market because we've done it before and it worked great combination sold for $6 billion. So, it's uh >> I don't have I don't have any doubt that that's that's how >> what was Scott's intervening company. He had an intervening company, didn't he?

45:38 >> Yeah. >> There's another huge one. >> Axiom. >> Axiom. Yeah. >> So, he was involved in the other sale Axiom to IPG when it was >> Yeah. Yeah. >> Yeah. with Live Ramp and then spun out Live Ramp with them. >> So, yeah. I don't know how I So, I I have to I know Live Ramp well. I don't know's senior leadership much at all and don't know how they're thinking about it. But my my my working hypothesis is they they essentially did this so that what everyone is afraid of them doing is not their intention. What they wanted to make sure is that no one could do that thing to them.

46:12 >> Right? It's defensive and it's and it's a layer that's not easily mimicked by AI which all agencies are thinking about and and they're not dumb. Right? The senior leadership at agencies are not dumb. They right they they they log into Claude every day and they go ah [ __ ] like [laughter] like if a brand can go into Claude and type in like >> right >> set up my marketing strategy for me tell me where you're building and running it and then hit go like uh this not good.

46:39 Um >> so what do we have that that that's where that's not a threat right where that is not a thing and it's it's data assets and it's interoperability piping and execution like you said like this um those are the things that Claude cannot have. Um, >> but you know what the the thing about this story that I loved is this story as far as I know is where the phrase data sovereignty has entered the adtech conversation.

47:03 >> Data sovereignty. >> Data sovereignty. Microsoft has been talking about data sovereignty for a long time. This is something been in enterprise conversations that the platforms should worry enormously about. Now that this is in the conversation, [laughter] if someone asks Meta, how do you guarantee my data sovereignty? They are it's going to be like Zuckerberg in Congress. Like Meta has no plan to give advertisers data sovereignty, >> right? Data sovereignty and and to be explicit, right? Data sovereignty referring to the fact that advertisers own their data. It's not misogenated with other data and it's not used for things outside of >> it's not training someone else's AI.

47:48 It's making margin on your bids, right? Like this is this is a very intense concept to be entering adtech and I've I've been sort of waiting for it. Uh and I and I'm very interested and excited that the live ramp acquisition and and dig now have injected it in like if if if people start thinking about data sovereignty like what I was saying like what protects advertisers from in-housing uh this the conversation will be about data sovereignty >> and just by owning live rampus is in a better position now they're say all right you know your your data sovereignty is guaranteed by Scott howen live ramp who this good story about separation between a tech and an agency.

48:33 >> Ah, >> interesting. We'll see if they we'll see if they lean on that. [laughter] You know, traditionally they don't people make the opposite pitch, right? They say, "I have all this data, so I can see all these things that'll make your campaign run better." I know you hate that. That's the pulled architecture. >> I know, but they already have a network. The reason that's the can't be true is because they already have an ad network.

48:54 the same one you talked about before value click now called conversant now called epsilon network whatever the hell it's called but they have one already >> so it can't be for that [laughter] I've never read a live ramp contracts but yeah I would assume you're right >> live ramp contracts are solid on data sovereignty I'll tell anyone who doesn't doesn't know like there's no like you're saying like neutrality is the project is the product >> yeah it's pipes >> it's pipe it's translation >> pipes.

49:25 >> They're not taking your data. They're just translating the ID. >> I like the thesis. I like it. I put nothing past agency holding companies, though. [laughter] >> They do like to make money. >> Yeah. Yeah. Well, >> no blame them. >> You know, we were saying about competition before. Who's competing for agency holding company business? Do you think Accenture wants it? You think McKenzie wants that? an 80,000 person global business on that makes 14% margins and grows 5% a year.

49:57 >> Yeah. No, you're right. They compete with each other. That's it. [laughter] >> Yeah. >> They compete with each other and death. >> Yeah. Yeah. Rashard Debbakawala called them cockroaches because they'll sur they could survive anything and it's like Yeah. They have their corner of the world. If you want billboards up in Jakarta tomorrow, there's only five companies you can call. >> Makes sense. [laughter] Next story. >> Ah, this is a good one. This so this is what we were talking about earlier, right? Dynamic tech rates are a marketwide squeeze disguise disguises innovation. Uh, I think that there's truth in the headline to an extent. I don't think the problem is necessarily dynamic take rates because dynamic take rates can be done in a fair way. So, let me explain what a dynamic take rate is.

50:43 Um, a dynamic take rate is that the ad exchange or whoever is charging the take rate has the ability to change what their take rate is as long as it operates within some kind of bound. And the creation of the bound is the key. So one way to create a bound is to say oh the person who's taking this take rate can charge whatever they want as long as the average at the end of a period of time is x so 20%.

51:16 That is the bound that is dangerous that is the one that is broken. Um, and the reason that it exactly because it's an average and what that does is that means that that person >> can find those arbitrage opportunities and maximize the surface area that they are buying. >> Yeah. >> And that they put so it puts them in a position where they're like, "Oh [ __ ] this one this is an impression that no one else is going to bid on. I know no one else is going to bid on it." And the buyer just bid $10.

51:47 >> This the same thing exists on the buy side, right? those there's Yeah, I remember I I can remember seeing the contracts that it was when seeing the contracts that basically said we have the right to change your bid. [laughter] >> Yeah. [clears throat] >> And Google started it, but everyone did it because in a world where they actually just convey your bid and this the SSP has a stable take rate, who who makes money in that scenario, right?

52:15 Publishers would collect all the money from over bidding, right? they would then that is and don't publishers read this article and go this is who should be getting the money from over bmitting it should be publishers and there is some truth to that like >> but it's much bigger than one >> it's much bigger than one arbitrageer right like >> you got it right right now the whole industry is set to reward this and so dynamic take rates can be fine as long as they can only go down right and that's that is an optimization we ran at Rubicon project so we did that analysis if we roll out a dynamic take rate that only goes down, how much more money would we make um by maximizing our service area, right? Because winning an ad impression at 15% is better than losing at 20%.

53:00 >> Why? >> That's because oh, I make money at 15%. >> Because you get some money. >> I get some money. Some money versus no money, right? >> Um and so and everyone wins in that scenario, right? Because I took a lower take rate. The publisher is making more money. There's no arbitrage harvesting going on there because I could have just charged 20 every single time and I that's what I would have made. The where where this gets really nasty is when they're allowed to swing up and they can charge 90 sometimes and one enough other times that it averages out to 20.

53:27 >> That was the logic. If we swing it down sometimes you got to swing it back up to uh to make up >> you got it. Oh, and they'll go negative sometimes, right? Uh we got to we got to make this margin look good. We're going to pay a little bit extra out. Um, so >> I think on this story was this is this is why SPO fails. Like anything any analysis you do of how much the S the SSV is taking is just a snapshot.

53:48 >> It's a snapshot. >> And people people have done this. People have tried to calculate bid rates from crawlers. Does not work. Um, and >> I'll talk about it another week. We're making like a we are creating a dynamic SPO offering where it's Yeah, I'll talk about it another week. We really have five minutes left and I really want to play this song again. >> All right. Um, but this is to say I also think that SPO has led people astray. Like the the I think that you're right to point out that the concept that SPO is a take rate optimization project is doomed to fail.

54:24 Um, and even worse, SPO has become a barometer of quality. And I actually think in the grand scheme of quality signals, I'm not sure that supply paths is a good one. Um, like I don't like it. >> Oh [ __ ] got a creative corner. >> Oh no. All right. But yeah, don't don't use don't use supply paths as a proxy for inventory quality. Don't do it like or if you do it, do it really smartly and like correlate things into models.

54:51 And >> yeah, we're taking a buy next week, so we better go ahead and do creative corner because we're all Let's do creative corner Junth next week. >> Yep. Welcome to Creative Corner where Sasha Marley and and David Maldo [cheering] >> us to talk about ad creative >> planet ad tech. >> Hello. All right, let's creative this week, guys.

55:23 >> We got a video submission from a viewer. David at WPP sent us an ad that made us feel something. >> Um, which is awesome. So, let's get it up on the screen if you would, David. >> If you read the start today, you might think perhaps there was a printing error. Literally, every issue has this blood going through stories at the center of the page. What goes through? And then it says, "A newspaper can absorb the blood, but not the shame." The point they're trying to make, 4 million school girls resort to newspapers and other unhygienic substances because they can't afford period products.

56:11 What an ad. This is This is quite something. >> So brilliant. I think that's so amazing. Um, >> I also would argue like it's kind of borderline art, not even ads, you know, like the experiential aspect of it. And I just like brings you into that uncomfortable, you know, the it's jarring, it's depressing, but it's like something that you won't forget. So, I feel like that's like makes all the qualities of a good memorable ad.

56:45 >> I agree. I think that a lot of us, like I've met some of the most creative people I've ever met in my life in this industry and I think we get into it because we're creative and we can optimize the pipes and like have a lot of fun with that stuff. I've been chasing it myself lately, but when you get a good ad that makes you feel something, it's just a lovely reminder, you know? >> It's possible.

57:09 >> Yeah. [applause] I feel things when I see interstitials. >> Mostly bad things. [laughter] Was a great ad though. I saw it on Twitter. Um >> yeah, South Africa tough. >> So should we play the song here and and dance in the flowers? >> Yes. >> I think we can close with a song. [laughter] >> Thank you for joining everyone today.

57:40 >> UHHUH. TAKE THAT. COME ON. COME ON. COME ON. COME ON. COME ON. COME ON. THEY WERE SIGNED OUT, THEN THEY'VE BEEN AART. TOO MUCH pride in the room. Too much heat [music and singing] in the dark. One side talking deals, one side talking tr same old table, same old stain. Same old dust, now the [music] wrong got quiet, then they lit right up. I SEEN THE TENSION DROP. When the money called us, [music] you can hear that room. When the eos crack, now they back in the lobby trying to take it back.

58:09 Hands on the contract, EYES ON THE BAG. No more side eye baby, no more drag. Close [music] that wound up. Let it fade. When the beat go boom, WE MEET HALFWAY. PUBLICIST AND TRADE desk back in the room. Back in the room. [music] Back to back. Publicist and TRADE DESK. SHAKE OFF THE FUEL. SHAKE OFF THE FUEL. We be through the fire. [music] NOW WE PASS THAT TEST. Publicist and TRADE DESK. BACK IN the room. Back in the room. [music] Talk.

58:41 Yeah, words got sharp. Whole crew froze up when they split that chart. But the streets don't care about a wounded pride. They just want the winners [music and singing] on the same damn side. So they roll back in with the chest out stiff. New handshake [music] firm or grudge lifted. If you got a seat, then pull it near cuz the loudest room is the one that's clear. Hands on the contract, EYES ON THE BAG. No more side eye baby, [music] no more drag cuz I wound up let it fade. When the beat go boom, WE MEET HALFWAY. PUBLICIST AND TRADE [music] GUEST BACK in the room.

59:15 Back in the room. Publicist and trade guest. Shake off the fuel. Shake. [music] Shake off the fuel. Through the fire. NOW WE PASS THAT TEST. [music] Back in the [music] room. This is so good. >> It wasn't pretty. It wasn't clean. >> Them names kept [music] the same machine. Let the old smoke rise then let it go. If the bag stay full then you already know now the whole flow shouting out. [music] Let it stand.

59:47 Two big players back hand in hand. I can trade this back in the room. [music] Back in the room I can trade this. Shake off the fuel. Shake shake off the fuel. We [music] keep it through the fire. Now we pass that test for publicist and trade desk. Back in the room, back in the room. Hey, [music] black back in the room [music] and trade desk to our friends ato.com.

60:35 It's art. >> Thank you for coming everybody. Have a wonderful couple weeks everyone. >> Two weeks ago. [music] >> Darn. That's >> Yo, the crowd is back. The lights are dim. Front rows buzzing. It's about to begin. [music] and tech at talk to