They Didn't Win Sh*t

0:05 flows in the scene tights up the night. Adam says pivot. Garrett says scale. Together theyave one flawless tale. >> Q&A side. >> All right. >> Okay. Okay. Two women. >> I like creamy cow. Jersey style.

0:50 >> Dub. >> I found it. There's a bunch of stuff like that. >> Welcome everyone. Welcome everyone. >> It's Friday. Thank you. >> You know, I hadn't seen Greece in a while. I forgot that that's what it was like. >> Yeah. I mean, take one up Westside Story with that >> dancing gangs. But hi everyone. I'm Gareth. I represent the sell side. >> And I'm Adam. I represent the buy side.

1:21 This is Adtech ad talk. September 4th, 2026. Thank you. >> We had a lot of news this week. Yeah, it was a lot of news. A lot of news the last 36 hours. >> I was getting sucked into social media, the world of bad takes. I don't know why I do it. >> I just >> Every time you do, I get happy, Adam. Every time you do, I get happy. I I There's There's nothing that makes me happier than when I see you going on LinkedIn and just being like, "This is wrong. What are you talking about?"

1:52 >> >> Like like like why are you saying this? >> I gave hight touch a very hard time this week. I bet you enjoyed that. >> I missed the hight touch one. What was that about? >> Oh yeah. High touches. Hight touch posted this test about how they beat Live Ramp. It's like a case study of how they did better than Live Ramp. But the use case they had was like sending a CRM list into Google, Amazon, Meta, and Tradeesk. and they have more matches than live ramp. And I'm like, stupid. This isn't a use case for live ramp. You could just upload the email addresses to all four of these platforms and they'll match them for you. Like why would you ever use live ramp? And if you did use live ramp and they matched, you know, like if you're m the fact that you're matching more than live ramp is just says that your graph has more matches for this CRM than even Google, which is sus. Why would you have more IDs than Google Google for any email address unless you're right?

2:56 >> Though I find the whole thing confusing, but I don't I don't I don't have to deal in identity. But yes, my instinct is 100% just upload your email list to to Google or >> Yeah, as I as I wrote like hight touch is like breaking into adtech. >> I'm trying to make a big noise, you know, take advantage, you know, they should, you know, with live rent being acquired, they have an opportunity. They have they have matches, >> right? They have something like the asset, but they're they're they're blowing it.

3:25 >> Well, I don't >> interesting. Well, it's it's also just an interesting use case to highlight because it doesn't strike me as like a particularly ad techy use case. like it's way more martekchy and >> and Live Ramp's life was hard adte like making DSPs and exchanges work together seamlessly. >> yeah, the hardest. >> I mean, like I feel like it's only clear now what LiveR's really for. Live ramp is for decisioning above the platforms. If you want to do any user level decisioning outside of a platform or above all the platforms, you need to do a lot of ID handoffs, >> right? Like yeah, you have to do a bunch of them and only live ramp a lot of people have the match table, but you need to do these fast like you need to circulate the matching. So live ramp is the only one that has tech everywhere you need to do a handoff. And that's why they're the choice for Chalice and a a few other companies like us and probably for whatever Pubis is going to build that we think will be some per advertiser modeling, right? That it's they're just the best for that if you want to compete with them. That's the that's what you have to compete on being a modular translation layer that works across a huge swath of the environment.

4:45 It's not about having the most matches. I mean, higher match rate could be good if they're real. >> I mean, in theory. >> yeah. >> I just I just don't know what like the the match rates that we discuss when we talk about ad text match rates are normally platformtoplatform match rates, right? Like that's that's where it's a concern. It's like how much of my audience am I matching properly in this instance of my ad tech? how well Google matches your emails to IDs that they have. That's that is that is a state of the world.

5:20 That is >> they don't need help. You don't need you don't need to pay anyone else. >> There's a there's a number that Google can match to. that was so you think so you think publishers publish is going down the per advertiser modeling route. >> I yeah I mean I don't know why else they would have bought live ramp considering they already have epsilon >> right? They already have a network. So a lot of people were like oh live ramp is going to be like a network play but they already have epsilon. Epsilon has and the conversant stack that they acquired.

5:48 >> Yeah. >> Everything they need to run an agency network and reportedly or supposedly they run a lot you know they run a huge agency network on that old tech and that's profitable. So you know they wouldn't they don't need another piece. So I think my theory is publicist was planning something like chalice and it got stalled in approvals because somebody realized like if we can't do ID translation at scale the rug could be pulled out from under us. So somebody was like well let's buy live ramp and then the rug can't be pulled out from under us and we could go forward with with this plan.

6:24 >> Interesting. >> Yeah. I mean conversions a bitter right? For all intents and purposes, it converts into DSP. So like >> Yeah. And epsilon's a graph. >> And epsilon's a graph. So building custom modeling on top of that. You're just >> tons of attributes. Yeah. They don't need it. There's no need. >> I mean, we will finally see an agency holding company bring their adtech in house, which has been rumored forever. but that's an N10 adte stack right there. Like what we just described, >> it's an SSP. Epson's an SSP as well.

6:54 It's everything the the whole and just all of ADTE. They've got all the boxes checked. It was built a long time ago, so it's it's ancient, but it can still do it. >> They're also not the holding the only holding company that bought all the pieces. So, it's it just shows how much politics interfere with plans. >> Yeah, it's so interesting. yeah, holding companies, they're fascinating that way. >> I don't know if other holding companies have all the pieces, man. I don't I this is the only one where I know where there's an SSP and a DSP in it. The other the other holding companies have been loathed to buy those parts.

7:29 >> Didn't Conversion had an ad had an ad server too, right? It was Value Click. >> Yep. It it exactly it is it is one of the oldest ad networks there is is Conversion. >> Yeah. >> yeah, Value Click. What a name. >> When when I worked at Avenue when Atlas got bought, everyone was like, "Oh Value Click's going to be next because Double Click got bought by Google. Microsoft bought bought Atlas and then the other ad server was Value Click, the number three ad server at the time.

7:56 We're like, "Oh, they're going to get bought for a billion dollars." But it didn't it didn't happen. >> I mean, it did eventually. >> Yeah. >> No, it it got rolled up before someone's going to hopefully someone will chime in and tell me tell us, but I think it it got rolled up into what was Conversant before epsilon bought it bought Conversant. Yeah, it was definitely in conversion because they also bought something called Grey Stripe which was a mobile ad network. and they they rolled that into that thing too. It was >> wonder if that was PE was a strange beast. That was a that was always a very strange company to me.

8:34 >> they were they were like open web minded though. One of the reasons why I guess they built an SSP, right? So they were one of the reasons why Pubis is a force in open web. >> Oh, big time. like they they had all the guts for it. They had they were they had direct sales teams, right? So that was the thing. The reason you worked with Conversion when you were a publisher wasn't because they were a particularly good exchange. They actually were not.

8:56 it was because they had actual IO deals that they were closing. >> Got the code on page from the value network too, right? >> Yeah. Well, and I think they converted that over into header bidding via conversion. So when they rolled out the exchange that that all rolled into it there at the very beginning like in 2011 there were still value click like tags on pages but that that waned. >> so this is a buy side story but did you hear about Pubis winning Pepsi?

9:22 >> That's crazy. >> I saw some tweets about it. why is that crazy? >> Why is that why is that so interesting? Why are people so fascinated by this? it's it's a humongous account. It's been with Omnicom forever. And Coke, parts of I think parts of Coke are at Publicist or in or up. >> Coke was I mean so when my wife was working at publicist they she was flying around the world doing Coke pitches and that was to keep the account. So I I always thought of Coke as a publicist account.

9:54 >> I remember it at WPP. I think WPP lost it to publicist or lost pieces of it to publicist. >> Makes sense. So yeah, and like no one's ever had Coke and Pepsi. >> Yeah, that is I mean it is a funny thing to have in one holding company. Like I can't think of >> I would love for someone to chime in and tell us did did Publix say that they would pull out of the coke pitch now and let it go.

10:17 >> I thought that's what they said. >> That's what I perceived from like the the social media chatter. I'm here on the marketing dive article now and I think you're right. >> Yeah, thank you Marley. So yeah, that's that's weird. We'll never maybe I'll find out, but I can't. If I do, I won't be able to talk about it. So, the public will never know what actually went down there. >> I mean, isn't Pepsi really struggling?

10:42 Like, >> dominating market. >> Pepsi owns Fredo and I think they own I think they own some fast food, right? Someone look up what PepsiCo owns. Fritole is killer. The chip. That's a >> Oh, yeah. I grew up in Texas. The king of cho rice. >> Don't have to explain Fredo to me. >> I come from the land of the Frito pie, which you Yankees don't know anything about. >> And the Pepsi portfolio is great. It has, you know, bunch of top sodas all all and water and juice and everything.

11:19 That's it. And they're they're known for having good data science. So they're they always been on my radar. Publix or Publicus. Meldo is asking. It's French or English. >> It's almost certainly Publix. >> Publix is their proper name. >> But I don't but I don't do that. >> I won't do that either. I won't say CO. I always call his French name CO. >> I'm actually incapable. >> Doing going halfway is even worse. I remember when the Olympics were in Al Albervie and the news and then the Olympic announcers kept saying Alberville like we'll leave off the T as a call it Albertville. We'll call it Albertville. But it's like if you're going to say Albert without the T, you got to say V instead of Vil, but they just split the difference.

12:10 >> It's American. >> Slame, right? You're not speaking any language. >> Nothing. >> Thank you. >> Yeah. >> Let's do the >> Let's do the news. Let's Let's talk about Google. >> Weird shit's happening. >> I mean, it's kind of weird. It's this This is one where like philosophically, I think this is what makes antirust in tech so hard, right? So, Google defeats the US bid to force the ad tech sale. I actually even I disagree like with the the headline.

12:40 >> They didn't defeat anything. >> Yeah, they didn't win. They just like the judge chose certain remedies and not others. >> yeah, they lost the trial. Not only did they lose the trial, but they had prepared a defense. There was like a parade of small business people coming in to say, "Google helps me and Google's good." And the judge was like, "Yeah, we're not going to do that." So, basically, their defense got laughed out of court by the 80-year-old judge. They lost badly this trial. And, you know, because it took a year to get to anything with remedies, I think people forgot that they lost. And because people aren't very close to it, they don't know how different it is to deal with Google these days because of the of the trial in the first place and the loss on top of it and how different Google's pricing. That's the biggest evidence. Buyer direct costing 10% when Google always charged 33.

13:36 >> It's a different world. >> They didn't win >> They did not win and they're also really smart. Like I I wrote this in my article like we slept on them and they rolled out product again. They So basically I I think they started doing this a lot earlier. Like I think Google saw this coming. I really do think that they they had they had an ear to the ground and >> like they really knew that this was going to happen.

14:02 >> They started getting subpoenas which are in an antitrust or civil investigative demands in like 2017 2018 >> when you were a little kid. Yeah. I their their product people were out in market in 2019 in a way that I had never seen before in my career. Right? So in the 10 years before that, I had never met a Google sellside product person and in 2019 all of a sudden they were asking us for phone calls and I was like and I was suspicious naturally as everyone was enjoyed it >> and but it was a thing >> and and yeah I think there's there's >> publishers were so mad publishers were so mad at that time that they were helping DOJ with like what to ask for.

14:44 They'd be like, "I had a meeting with Google on this day where they told me that." >> You should you should ask for the emails that they sent right after that. So, except for Google, you're like, "Oh, shit." >> They're like, "What you said at the what you said about the meeting privately within Google, like they knew they were caught." >> Yeah. >> So busted. And but then they started rolling a lot of product. And this is what makes antirust stuff so hard. is to actually remedy this one would require major disruptions, like major disruptions to how things currently worked. And that's why the the the proposed hard remedy from the DOJ was a three-year plan to try and like do it in stages. And I think Judge Brigham >> say I hope for a breakup. I hope I mean even when when I was asked by officials what remedy, I did not recommend breakup. I said you don't need to break them up. Can you can you >> Oh, you said it too.

15:44 >> They talked to me as well and I said the same thing and that's why I wasn't understand. >> So, we were we were on the side of the recommendation they took. >> you know, we thought it was I'm sure we thought it was important that they stop certain behaviors and get out of, you know, even you know, exit certain markets. This is as far as I went. But I was still rooting for a breakup until I started reading about, you know, how hard it was going to be to find a buyer for GAM or Adtech that they, you know, there's this huge problem with how big they were and how buying it as a standalone piece was like a crappy proposition for any buyer.

16:20 >> Well, yeah. And that was because of how they had designed GAM. GAM was almost a vestigial, not vestigial, it was it was like a a lamprey, right? It didn't it couldn't stand on its own. It's most of it came from they didn't char they didn't charge at all for it. It would they basically gave it away >> and and that and because of that like >> like what do you do with that because if you start charging for it the market share goes away and if you get rid of the tie the market share goes away and I think that where they landed is a solution for something that is just not even really a problem anymore because now they've moved past it.

17:00 They've they've moved to buyer direct which is okay we're not so dependent on adex anymore and the the whole thing of well you use GAM because you want adex if buyer direct succeeds publishers will just use GAM like it's it they're building it into GAM >> and I know Ari said came out and said the same thing this morning I I read his piece it was really funny it was structurally similar to mine and it it is terrifying Google basically rolled out a product in a year or two that totally obiates the need for the tie.

17:35 And I do think there is something to the thesis that the market share that GAM has is ill gotten. >> Yeah. >> And because of that, we we need a different type of remedy than just fixing the structural problem. And that was one of the things the DOJ talked about, right? like how do we how do we structure a remedy such that we prevent abuses in the future that we do not know about right now which is really vague and really hard but I I still think that there Gam is going if if buyer direct succeeds GAM will use its illotten market share to ensconce itself permanently or semi-permanmanently we can compete with buyer direct which is nice we could not compete with addex because addex had adwords and it nobody nobody restore competition, right? Like say like now now we're going to compete on price and functionality. It's not like it's not such like it's David and Goliath because they're so big and we're so small, but it's it's still competition.

18:35 >> Still competition, man. This is I think the the conclusion of my article was that competing against Adwords was impossible. It was like competing against a god. There's there's nothing you can do. You have the number one consumer small business advertising system on earth that just funnels demand into websites as it sees fit. We we're not going to build something that can compete with that. We can't buy direct thing. >> It's it's convenient. It's nice. Does a good job. But >> what about DSPs? Like the the measure the hold on measurement and none of the remedies really are going to help with Google strangle hold on on measurement and buy side gam. And I think it's close to impossible for a DSP to compete on performance you know or for any byside network or entity to compete on performance. Now especially in Europe and this this case doesn't cover Europe and there will be a case that does but in Europe where IDs are illegal except for Google's because it's first party like there's a there you know there's it's really far from having a competition on on CPA.

19:42 Yeah, I I really don't know how to properly decouple Google. I don't know how to do it. It's it's it is such a thorny problem and they have so many business units. The answer is probably we should have stepped in a long time ago. But >> I mean, if if GAM had APIs that released all the impression data that they used to measure, then you could compete with another byside entity. I mean it's it's it's an issue because you'd have to find a way to deal with the privacy issues but live ramp has proved you could do ID based work privately by by encoding and transcoding. So yeah, that's that's something I told DOJ is like they shouldn't be able to keep like the log the gam logs the advertiser side gam logs but should belong to the advertiser. The advertiser should be able to get them out with transcoded IDs and you know and model conversions.

20:44 >> Is this not a thing? >> It's not a thing. >> I never I never work I never work with DFA. I I never work with DB3. >> There's no APIs. Yeah. this any DSP that competes with Google has incomplete information like if it if if the measurement is Google last touch right in order to model Google last touch you need all the impression logs that show how you know when Google lasting an ad server so like when I'm I'm a buyer I'm an agency and I sign up with like DV like their whatever their ad server name is I don't even know what their buy ad servers anymore.

21:21 >> They call it Google Ad >> DCM. Google campaign manager. >> Google campaign manager. Yeah. DCM. DCM. Yeah. Yeah. Yeah. they don't give me log level data for where my impressions served. >> That's so astounding. >> Yeah. You can get the actually you can get the all the pages. >> You can get impression level data by page. >> But I think even that's hard. >> You have to use what they call this you have to use this DTS process. They don't always approve you. But you can get every page which is something.

21:53 >> Why do people use this? >> There are so many things that are so much like better. >> >> they think that their ID graph is so good that they're going to more accurately represent the conversions. Like >> they don't always use there's a lot of innovid ad serving. >> Yeah, I see inid a lot in like creative text. I see a ton of innovid. >> yeah, the I mean the idea was that they're a hub that you need your buy ad server to be a hub so you don't double count across your partners. So agency buying is a hub and spoke model. The measurement has to be in the hub so you don't double count. And there aren't really many alternatives to to Google that are scaled and of course cheap.

22:37 They almost give it away, right? It's like a fraction of a penny per impression or something. And it's, you know, it's highly standardized within the agencies and even in-house teams to be the the source of truth on billing and measurement. >> Interesting. And and a big part of it is just like trust us, bro, because they don't give you the IDs. So, it's like >> Yeah. >> Yeah. And you know what their solution is for for Apple? Do you know what Google the Google Ads server measurement solution is for Apple IDs?

23:06 >> What is it? >> It's called modeled conversions. They they make up credit for conversions. they think they probably got for impressions served on on iPhones. It's freaking crazy. We looked at it really close and like they don't even come in one at a time. They come in in these huge batches. Well, here's a whole bunch more conversions we probably drove. Just accept this. >> I mean, I know what they're doing. I know why it comes in batches, right?

23:36 They're they're running a hash query. So, basically, they're saying, "We see these users over here. I see these user over here. What are the user agents? What are the IP addresses? Blah blah blah blah. And we run this query once a day. We dump out here's here are the probabilistic matches. >> >> but they don't get any like asterisk in the counting. They just count as conversions conversion. >> You know, >> Facebook's the same. Facebook migrated entirely to probabistic conversion tracking even for postclick. Even Even Facebook's postclick conversion tracking technically modeled >> billions of dollars swinging on who's willing to take more credit. And you know what? You know what's and you know it's going to get worse because because there's this now this adoption the first place that Agentic is being adop adopted is budget allocation and the budget allocation machine says oh CPA is down here let me put more money there. And if CPA appears to go down because somebody took a bunch of credit for conversions that they feel happened. Now they get money automatically for for for lying.

24:36 We're going to see it get worse before it gets better. start automating on this >> I don't know how >> great Odin's Raven. >> I don't know how you could do interplatform CPA optimization. It still eludes me. Like >> it still eludes me. >> Just get an agent. >> Yeah, that's the first thing that agents that agencies want to identify the budget allocation, which they've always done kind of arbitrarily anyway. At least, you know, now at least they could say it points to best possible CPA because there's some calculation, but again, they're creating a a very sensitive incentive to lie.

25:18 >> Yeah. I mean, are there third party MTA vendors? Is that a thing? Like, >> no. Google crushed that market because of the G the GCM ad server logs. You needed those logs to do it. >> Interesting. They took the IDs out of the logs for for privacy reasons in 2018, I think. >> And they crushed Visual IQ >> and their competitors. >> I feel like you could do this. >> You could do this with impression tracking pixels. I'm surprised. I mean, Rockerbox offers this, doesn't it?

25:52 >> Yeah, >> I think they do. I think I think they have an MTA solution. >> Imperfect because it's imperfect because of the lack of logs. the yeah you could have an impression tracker but again you can't get you can't match the IDs impression to conversion there >> it has to be probabistic it's the only way >> because there's there's no way that you can like sync an ID between the logs and your conversion tracking >> someday we may see a buyside case around that conduct the domination of measurement all the bull crap that analytics expose like Google's really played a dangerous game with measurement, giving themselves credit and you know very committed to appearing to have the lowest CPA.

26:40 >> Yeah. >> I mean it's fascinating to me that they don't have to have disclaimers everywhere >> like this is modeled. This is this is probabilistic. >> We we don't know. >> Don't use this for budget allocation. It's for entertainment purposes only. >> >> That's the label it should have. >> One of the best artifacts in the Google case was somebody from IA. Did you ever see this? >> No. >> He's like saying it's who's the guy Malone who owns IA like match.com and them. I think it's >> Barry Diller >> writing the letters. What's his name?

27:15 >> Barry Diller. >> Barry Diller. Yeah. >> Yeah. >> Barry Diller. Right. And I don't know if it's him writing it or one of his lieutenants, but he's like he's really really nice to his Google rep and he's saying like over the last five years my Google investment has increased and it's something like 10x and he's like the amount of traffic match.com or whatever site he's talking about gets from Google has been flat. And he's like, I know, you know, this is justified by the metrics that drove it, but if you back up and just look at what I've paid and what I've gotten, it doesn't make sense. And I'm at my wit's end over it.

27:58 It's like you just pull out the attribution fees and look at what did you pay and what did you get and it just completely stops making sense. >> Yeah. I mean, I was on a phone call with one of my buddies yesterday who is a Silicon Valley guy and he pointed me at a one of the new tech startups that one of his own co-workers running and it was an MM company Adam like there are company >> MM right so like a medics modeling company and they are rolling this out like it's a new thing because all the DTOC companies are are realizing like oh I can only use platform attribution so long and now everyone's like oh we actually have run models to do this properly. It's the only way.

28:40 >> yeah, I mean, at least it's in it. At least it's it's causal inference, but at least the word causal is in there. >> Can how does a custom model optimize to an MM? >> Some of them give us feedback, but mostly we just try to drive incrementality. If you if you if you could just get a hold of a set of non-incremental conversions that came in in a state you were in advertising a PSA ideally or you could even just use the loyal customer base that makes conversions anyway and you use that as a negative seed. In other words, you tell the algorithm this is what we're not looking for.

29:23 then you could start then you model something that will drive the mm because it won't >> right because you're just filtering out the sales you know what what fake attribution >> runs on sales that would have happened anyway right so you just make all those negative and and try to filter them out of your model >> so instead of optimizing to like a positive output from the mm what you do is you target away from the negative space so You're like, "Here's here are the conversions that are going to happen anyway, we think, >> and don't buy things like this." When you look look for conversions. So, you still have to >> also get a seed of new customers. Yeah.

30:01 We also try to use a seed of the customers you want. So, there's a positive seed and a negative seed. >> Got it. And so, and then while the algorithm is running for like the reinforcement part, it it looks at actual conversions that happen and says, "I think this was incremental. I think this one was not incremental." And it just and that's how it makes a decision. So it still has to optimize to conversions. It just does optimizes to the right conversions, >> right? Yeah. You you just have some kind of new customer flag that this is someone who bought and never bought before. It's pretty easy in CPG where, you know, there's new buyers and the rest, but there's other ways of doing it in other verticals. but we're very confident saying, you know, this will move the MM because they're looking for they're looking for a causal relationship between spend and and sales. So, we're saying the, you know, we're modeling sales that happen because of advertising. If the ad works at all, you'll see this in the MM and it works.

30:57 >> Oh, that's so interesting. All right. I I've had some conversations with some B2B companies lately that use nothing but MTM, their MM, and they were like, we the reason we don't run programmatic is because we we can't get it to drive incremental. So, >> I say you should go talk to Chalice. >> B2B. Yeah, B2B is super hard. Is it tech B2B at least? >> Yes, tech B2B. There's there's a >> you worried about other B2B like agriculture B2B >> like you know the ones the toughest ones are like office supplies where your target is one office manager and a skyscraper like how are we going to find this person the one person who buys the office supplies that's really hard and then we've done like waste management there's so the B2B world is crazy >> there's like multi-billion dollar companies that just pick up the garbage from offices.

31:51 >> It makes sense. Amazing. >> Like B2 B2B is so fascinating to me because the tickets are so big. Like it's got to be a semi-stressful marketing department to run because it's like well for us to have a significant test, we have to spend a hundred grand. Like that's the like the only way for us to test something is to spend that much money. >> they're very conservative. You know, there's very few flashy B2B brands. Their names and logos and stuff are almost always boring.

32:19 You know, they they spons they do things like sponsor golf tournaments, try to reach business people. You know, they advertise in airports where business people supposedly are. I don't know. I think Vibe Vibe was the most was the flashiest B2B advertiser of all time. >> The founder addressed as Sydney Sweeney and stuff. Well, they also had in their ads how to target people with email addresses at companies. That was like one of their one of their like flashiest subway ad was reach all of these people at Open AI.

32:59 >> yeah, >> and that was that was a thing. So they were a company advertising B2B, but they also advertised that they could advertise B2B. >> But they did B2B. They did B2B. >> Yeah, I think it's I think it's a vertical where it pays to be bold. Hard to make a it's it's it's humongous. B2B I think is bigger than B to B T to C but there's a diverse there's a great diversity of problems. it's very hard to most conversions take a long time and require a phone call but tech is the vertical where there's some scalable some opportunity for scalable solutions because tech companies buy tech and a somewhat predictable pattern.

33:40 >> Makes sense. >> You're small. >> Yeah. They're more comfortable with it. They're more comfortable buying tech. >> Yeah. You learn you learn you need to buy tech. You need to have subscriptions and SAS and then you need to buy cloud and then you need to buy an LLM and so forth. >> Yeah. So yeah, man. Wrapping it up with a Google discussion. I am not such a pessimist because I don't really know what the alternative is and the market has just moved on.

34:12 like between a gentic and like buyer direct the this the remedies kind of are what they are getting them getting addicts to bid into pre-bid and trying to police that that's going to be a hard thing. So one thing one thought I did have on this right the reason I was originally I was originally behavioral remedy and I and then I switched over to divevestigure the reason I switched is how do you police that adex is playing fairly right that when adex buys in GAM it doesn't bid higher than when it buys in pre-bidjs >> but that can be done >> there are lots of thirdparty reporting companies out there there are lots of vendors and we just need a little collaboration here like someone needs a contract with the DOJ where it's like hey Google you're going to submit to this reporting and you will you will prove to us that you are treating outside the same as inside and if it's not you're going to tell us exactly why you have to >> yeah most people don't know thisly disclose >> you know Microsoft lost their case and then one on appeal I think there were then they settled in the f in the end Microsoft settled and they got behavioral remedies and the main thing of the behavioral remedies was they could not use Windows to shut down applications. And it's just the same thing with Adex. It's like, how do we know they're not around with Windows and making it hard for people?

35:32 And it succeeded. They they have they made like an Windows interoperability committee and there was government oversight making sure that if you're building an application that needs to connect to Windows that the APIs are open to you, right? as long as you follow the rules, you could use the Windows API and have a functioning web applications and Google and Amazon and a zillion other companies were able to grow up because DOJ successfully stopped Microsoft from stifling the web from using Windows to stifle the web.

36:15 >> And this is this is really similar. It can be done. You Google, you cannot sit in the middle of the market and use GAM to stifle programmatic advertising. And they're gonna I think they're gonna watch them close enough to make sure they don't. >> Yeah. And that's definitely an easier thing to bite off than how do we break this up without damaging the publishing businesses that are on there? Because every publisher I talked to was like, "Yeah, sure. I don't like that Google has so much control, but they're 50% of my revenue and I run all my direct sales in this and we're not exactly crushing it right now. So, if something breaks because of this, it's a real problem for me.

36:59 >> Yeah, no one Yeah, that no one wanted that. And I think it's underappreciated that they had a they had a solution. They had a plan B for Breakup 2. They were going to move all the functionality to Chrome and Android. >> I have no doubt in my mind they were going to move all that functionality to Chrome and AMP. The privacy sandbox was like a tipping of the hand that they were going to move, >> you know, they just have another middle of the market to mess around with. So, I think I'm glad it's not a breakup for that reason because that would have been a horrendous outcome. Oh, it's that would that is the worst case scenario when that was happening. I I I'm I'm not surprised it didn't go anywhere because it's it's a vertical integration that is unfathomable, right? integrating a browser with an ad server with an ad exchange, like coupling those three things together, it's just it's it's such >> inconceivable.

37:49 >> It's such an absurd advantage. >> yeah, >> no, no one could compete with that. Like the native functionality available in that ads will render faster. Hell, ads could render before the content of the page does. Like they they could >> they took over pages. They took over mobile pages. like the amount of control they were starting to assert in in Android and were threatening to do in Chrome was was like the Death Star. >> Was still possible.

38:21 >> So, so in the end of the day, it's not the worst outcome possible. >> We can tech is still fighting. It's still kicking. >> Yeah. I think people should be rubbing it in Google's faces more that they lost this trial, that they were caught cheating in a zillion ways. They're going to face a zillion civil suits that they'll lose if they bring to trial. They're going to pay out millions if not billions to publishers, rival ad servers. Like, you know, there's, you know, whatever. If if you want to see Larry and Sergey suffer, you don't get that, right? They're going to be fine.

38:59 >> >> There's there's truth to that though. A lot of there is there was a lot of in the initial replies there was a lot of emotional investment and I just want to see Google suffer. Like that's that's really like how I >> that wasn't the government's job. All right. >> No, I hope >> that man >> I don't want the government being victindive and just making people suffer for the sake of making them suffer. It's not >> Yeah. And we we like we love success.

39:25 Microsoft's better company today. They make way more money than they did in 2000. you know, THE GOVERNMENT >> MAKE IT >> shepherd them off conducts that were unproductive and then they focused on things that were productive and they're the engine that that drove cloud computing and open source after that. And you know that's what we want for Google not to suffer to be a better company and spend their resources and energy on new products and innovation instead of trying to stifle new products and innovation.

39:58 >> >> Yeah. >> All right. What's our other story? We got one more, right? >> It's delicious. >> Yeah. The FTC sues Amazon for deceiving advertisers to drive up prices through a conduct we saw in the Google trial, which is telling advertisers you're running a second price auction and then placing that second price bid yourself to drive up the winning price.

40:30 >> it's rational behavior. >> It is. >> And it it's legal under certain circumstances. The circumstance where it's not illegal is where you lead everyone to believe it's a fair second price auction and it's not. So that is why that is what FTC is saying is not an antitrust case. It's closer to a fraud or a consumer abuse case where the consumers are Amazon sponsored ads advertisers. This was not the open auction. It's the sponsored links auction in Amazon ads, right?

41:04 >> Yep. >> So a totally private auction where I Yeah. Like legally Amazon could do whatever they want, but you can't tell advertisers it's one type of auction and then run another. >> Yeah. Well, and I think that this is really important. I mean, this is why we moved away from second price auctions and programmatic or one of the many reasons, right? One reason was because of the ad server, but the other was because the there was just rampant gaming of second price auctions.

41:28 >> I thought it was fun. >> Oh gosh. the the when you ask some people what yield management is, they would say that gaming second price auctions is yield management. That is that is how some people would define it. I and let's let's back up and explain it a little bit. Right? So the thesis behind a second price auction, the reason this is so repugnant, the reason this is so naughty is because in a second price auction, you pay only what it would take to beat the second price. The incentive being we want you to bid what you think it is worth knowing that you will never have to overpay. That is the point of a victory auction is bid. You can bid as high as you want. You'll never overpay.

42:14 >> All right? >> So if you tell people that and then you charge them more arbitrarily. >> Like we tricked you into bidding what you think it's worth and now we're going to make you pay that maximum. It's a cut and dry fraud case, right? Like it's just just like cut and dry. That is >> amazing. >> You you you have you have lied like that. And and they're saying that they didn't. I'm sure that there's like some sort of like excuse for why this mechanic existed, but don't tell people it's a second price auction. That's an own goal.

42:45 >> I wonder about their defense, right? I was wondering all about what Google's defense was going to be and then it turned out to be, oh, we're going to bring up a hundred people who say that Google is helpful with their small business. like what does that have to do with cheating in auctions? >> It was it was it was a non-defense. I know. I mean, everyone was pretty flabbergasted with the defense Google put forward and like pretty I was I was super impressed with the DOJ. The DOJ like really understood how this worked and they were they were there and they were ready to talk about it and Google's lawyers were somehow talk trying to like play this emotional game. It's like what are you doing?

43:17 The DOJ is coming in here armed to the teeth. I so I expect they'll do the same here. Like they're going to be like this. >> You think Amazon's like that? >> No. I I Well, I know the DOJ is going to be armed. I don't know if Amazon's lawyers are going to >> Well, it's FTC, not DOJ, which I'm aware of some nuance difference there, having met some of the people. What >> is I guess they're different lawyers.

43:41 Like, >> they're totally different groups of people that have totally different jobs. I Yeah, the DOJ lawyers were incredible. They learned adtech back and forth and like I teach I've been teaching ad people adtech for 25 years. So I have a good sense of who's learning it and how fast. And these lawyers were at the upper echelon. Some of them quite young. Some of them right out of school. So I don't know if it's going to be like that at FTC. It's not the same people.

44:07 >> Interesting. When you have a case brought to the FTC, is there like a separate like is there like an FTC judge? Is that how that works? Like does it go in front of like a tribunal? Is there a committee? >> I don't think so. No. I think it'll be a reg I think I think it works just like DOJ where they they they file a tort against you and you go to civil court. >> So you go to civil court. Okay.

44:26 Interesting. Yeah. I mean >> I know Google used Did didn't Google use like a third party law firm too? Wasn't it outside counsel? >> >> yeah. >> I mean it was it was it was a big fancy law firm, right? That that was another reason that I was so flabbergasted with the defense because I was like this is >> tens of millions of dollars. >> Was her name Karen Dunn? It was the same woman who prepared Kamla Harris for for her debate with Trump.

44:50 >> Oh, no. That's not true, is it? >> Double loss. >> Oh, that's not >> I'm sure she's doing fine. Don't feel bad for her. >> No, I don't I don't feel bad for her. I feel bad for the rest of us. >> but yeah, that's it's it's really interesting. I I think that people tend to play fast and loose with telling people it's second price auction. I'm not >> Trump we got Nautil. Oh, okay. I thought it was something I said. but it's one of those things where these perceptions really matter. And part of why I think programmatic has been so burned is because advertisers perceive things as not being what they are.

45:34 whether or not that's the case, but there aren't a lot of other advertising mediums where you can say, "I want to buy a video ad." and then you end up in like a 320 by 50 on a mobile phone with a video ad like in the tiny little bottom corner. There aren't a lot of advertising mediums where that can happen, right? Where you can log in and you can just end up with something so utterly different than what you thought you were buying.

45:54 >> Yep. >> And people don't and so many auctions like the money really piles up. But it's going to hurt Amazon no matter what the case because one of the things one of the things economists have been able to prove about auction behavior is if people lose confidence in an auction even a little bit everyone bids down, >> right? And if you're a seller and there's a case of everybody bidding down, it really will hurt your monetization because >> the mon auction monetization depends on some people being so gung-ho for some for some impressions that they over bid.

46:28 Right? That's how that's how you make that's why people sell things at auctions. So an auction people lose confidence in losses the the thing that makes auctions valuable for sellers. >> Part of what I don't understand about this one is there's no one else in this auction is there. So couldn't Amazon just say this is what it costs? >> Yeah, Facebook too. >> It's their thing. I well in Facebook there's I have never once seen Facebook tell me that they have some sort of internal auction methodology. I've never seen it allude to anything.

47:04 >> They say it's an auction. Yeah. >> I I understand right. Why would it be? >> Absolutely not. Facebook is optimizing a curve and this is what they think this should cost for me and that and that and that and that process of optimization is like okay how much can we charge while it still backs into rorowaz. That's that's the that is the optimization here. We don't want to price this person out so that it doesn't perform and they stop spending, but we want to charge them as much as possible. So, there's like a a curve, right? There's a there's a maximum point where it's like this is the maximum they're going to spend to achieve the maximum number of conversions at the optimal price. And >> I think they need an auction. They need an auction to get you to enter your maximum price, right? like they need a signal from the buyer what they're willing to pay or how much ROAS they'll how you not a bid but they need a target rorowass right >> you need a target rorowass yeah I mean because there there are different ways to use Facebook right Facebook you can put caps in there and you can say this is I think this is the cap CPM I'm willing to pay which is it's astounding me that that's necessary because everyone who uses that cap feature all they say is well a conversion is worth $3 to me so what what CPM M is a $3 conversion impossible, right? Like where like like you know I need like every ad would need to convert. Yeah, I don't want to pay that.

48:28 >> yeah, didn't they just someone just announced that the cap CPM they could go over it as long as the average comes out to that cap. I forgot who that was, which is, you know, just another way of taking that signal and pushing it as far as you can. I I at the end of the day all all of these like tweaks and things that's why I'm so confused about this Amazon one is if I'm Amazon why am I saying there's any transparency like I I just want to say >> put the number in the machine it'll deliver or it won't deliver that's that's what's going to happen if it's not delivering put a higher number in there like >> I think it comes down to the people right I think I think I I think that the people who work at Amazon and Google want to think of themselves as good and fair and meta.

49:15 They don't have that problem. >> Hey, that's why they're so good at what they do. Okay. >> Beyond it. Yeah. That's there's no pretense of being of being good and fair and moral. >> To be fair, if you look at the likeformational assets that these companies have, right? If you compare Amazon has the most astounding purchase data that's probably ever existed. Google has our entire lives, right? Like just just the whole whole thing is in there, including a stream of consciousness feed of what we're thinking about in this moment. Meta has who we went to high school with.

49:53 >> Like, >> but they are still an advertising behemoth. >> Yeah. >> Like when you think about their actual assets, they're not the same. Like it's it's it's nowhere near as like powerful a data set inside of Meta. and they've just created these spectacular advertising platforms. Got hats off. Like this is it's a real thing to have done. >> and Marley says, "Is is it is it attention?" And I would argue that there are lots of media companies, right?

50:24 There are lots of places where people invest a lot of attention. So, Facebook has a lot of attention, sure, but there are lots of other places, too. Facebook has done this spectacular job of converting the attention they do have into advertising revenue. and it's it's it's not a trivial thing. I remember >> they'll never come out. They do a bunch of sneaky things. >> They must right there. There must just be this this horrifying laundry list of here's how we make this make so much money. because I remember >> the Jedi Blue Jedi Blue is one of the best artifacts of the Google case because it's the only view into Meta's behavior and they negotiated a deal with Google that indicates by the deal that they knew every single thing Google was doing. And in the deal, line by line, it's like, you can't do this, you can't do this, you can't do this, you can't do this. It's like, oh, this everything that Google got caught for DOJ, Meta knew about and negotiated their way out of when they made it when they made their own deal.

51:19 >> Jedi Blue Jedi Blue was such a loss for mobile advertising. It was so dumb. >> Why? >> If because well, Facebook got burned on the live rail acquisition. So that happened. They spent $300 million on an adtech company. They wrote it down to zero because they bought they bought an arbitrage video ad network and somehow thought it was something else. and so Jedi Blue, they pulled out of pre-bid and that meant that the mediation or the mediation landscape was left wide open and then became inhabited by like Iron Source and one or two other >> companies landscape. so the mediation landscape is when you have a mobile app, how do you maximize the revenue that you get from the ads in that mobile app? So they do not have header bidding. They do not have a true unified auction the way that we have it here. What ended up happening is they had waterfall monetization and then the companies that offered waterfall monetization tacked a header bidding unified auction type of thing onto it, but they maintained their pole position as the like primary mediation partner, the primary monetizer of a publishers's inventory. So in the web, you have pre-bid is like the lightning rod that every publisher uses to surface its inventory out to buyers.

52:40 That doesn't exist in mobile. In mobile, you work with one of the three or four big mediation providers and that company is how you interact with the rest of the world. They're you're it's gatekept. >> and Apploven is one of them, right? Like like Apploven, they have their own like format that they run and this is a thing. But Apploven is a new thing. This is a this is way after Jedi Blue. Apploving Apploven when Jedi Blue was happening was primarily a publisher.

53:08 They primarily was their own apps. they hadn't built their network into what it is now. And Facebook in theory could have used pre-bid to jump the unification the unified auction thing and dominate it and like place Facebook and Facebook audience network at the center of mobile advertising which would have pre prevented Apploving by the way. Like it I I I believe this in my heart of hearts. It would have prevented the rise of apploven if Facebook had thrown their weight behind a unified auction and offered like management in the middle of the open source would have been such a good play. Would have been so smart but they didn't because part of part of Jedi Blue was you can't talk to pre-bid anymore and in fact you got to pull out of pre-bid.

53:50 >> that was part of Jedi Blue. And >> why do you think it'd be better of I mean you're saying it would have been great for Meta and it would have been better for app publishers because it would have been open. >> Correct. Yeah, >> Meta would have met Meta would have had this control because if they were the primary force behind the open source and they're the one who hosts it for people like that's that's a business unto itself and for publishers they would have had a nice un >> yeah it would have been a tremendous source of data for them right why didn't >> tremendous source of data huge >> why didn't they do it yeah because they're pretty friendly with Google is you know there a lot of ex Google people's probably >> yeah there was they Google hated pre-bid Google was pissed off about pre-bid And they and Facebook was a a rapidly growing pre-bid buyer. Like they were top three in many many publishers pre-bid stacks. Fan on web was huge.

54:43 >> And yeah, Google was like, "Don't do that anymore, please." Facebook was like, "Oh yeah, we don't really give a We're putting ads on websites. We don't care. we make less money anyway." So yeah, this is we we'll see what happens. I I just don't understand why they told people it was second price auctions. It still confuses me. >> There's definitely a problem with the with pun, you know, like there are a lot of bad takes on Google, but there's definitely a problem with the enforcement the penalties being too small. Like it it seems like these companies are so big that they could write the penalties down as cost of doing business. like their profits dwarf the penalties and it starts to become rational for them to do illegal and pay the penalty. and we've seen a few examples of that and I don't know why they they should just charge them hundred billion dollars.

55:37 >> Yeah. I mean to be fair that that would have also been a compelling behavioral remedy for the antitrust case. In fact it was one of them wasn't it? Didn't they have to put together a slush fund for like money cost publishers that's gonna get paid out publishers eventually? >> And Adex would have to bid into pre-bid, right? Did did that happen? >> That's gonna happen. >> That's going to happen because Google Google's response to DOJ. DOJ said you have to sell adex. Google said we can have adex bidding into pre- bid within a year.

56:11 >> That was Google's reply. >> so that's how >> that'll be fun. Wouldn't that be fun for >> Oh, it's going to be so much fun. It might it might open up the ad serving market. It's possible, right? This buyer direct thing, it might not work as well as they think it does. And like >> I don't think it's going to work for enterprise at all, right? Do you I don't reserve on a zillion sites. We'll see.

56:33 >> I don't know, man. I don't know how the I don't know how buyers buy atom. The buyers are mythical creatures to me. >> I just I just I just give people data, man. I I'm >> No, there'll be no decisioning. Yeah, it won't work. Buyer the big advertisers need decisioning. I don't know why I had to check on that. It's like my thesis, but I thought it through again. I arrived at it from first principles.

56:58 >> Yeah, I you're right. I think the enterprise buyers are a thing. But I think every mid-tier agency, right, who works with like not small advertisers, but not big advertisers, >> and they're just like, I want to I want you to put my ad on CNN. It's gonna be >> locally. Yeah. >> Like they can just do that. They just log in and be like add on CNN. >> It'll be cheap impressions and Google's attribution predictions will make it look like good impressions, >> right?

57:29 >> Yeah. >> This is just like a static version of GAM, but it's not even that. It might it could change every day. And not GAM of a GDM. >> It is. It is. It is 100%. One thing that I think is really interesting about it is I wonder if it's going to murder publishers direct sales teams. because you'll have to figure out how do I compensate our sales I guess you can compensate your sales people for deals they drive via buyer direct and it will make their lives easier.

57:57 >> >> but there's another middle there. >> It's just going to be weird. Going to be fun. At least it's change. >> What about competing with it? Is there a way me and you could like do a huge guarantee and you get all your publishers to reserve inventory in their ad server for us to optimize? I mean, >> so this where this gets interesting. >> Does it have to be GAM? It could it could be another ad server.

58:19 >> so if GAM is there, it's a pain in the ass because I have to go build GAM line items. Ask anyone who's trying to do ADC CP today if that works. It doesn't. Building GAM line items sucks, >> right? >> We could build this into pre-bid. Like if you were to like if if if some buyer and if we were to if we were to coordinate this and somebody were to come to me and said, "Gareth, we we want to do this and we want to do it outside of GAM."

58:43 We could build a guaranteed interface inside of pre-bid in two to four weeks. End to end ad serving doesn't scare me one bit. We've talked about it a gazillion times at pre-bid. Pre-bid has so much of this logic in it already. You could build all of the ad serving infrastructure into pre-bid. The problem is a publisher has to be willing to migrate their direct sales out of GAM into pre-bid. And I would have to think through what that process would look what about the decisioning, right? Every time that we would need a decisioning layer on the publisher that's that's allocating between this guarantee, >> pre-bid could do it. So pre-bid could >> it's like a floor, right? It's like if you can't find a bid above this, give it to this. You can include the okay step one would be a priority and a price.

59:29 Yeah. >> So, >> yeah, >> just just introducing the concept of priority, right? Where when this buyer bids, it trumps price. >> That's enough, right? To do a to do a guarantee to do a >> direct sale header. The old creto header. >> Cretto. Yeah. So, that's enough. Round one, that does it. When it gets more complicated is when you have now you have three things that are at this priority. They have to paste all this stuff. That could all come later.

59:55 >> but let's do the creteo model. We like any RMN we could do the equivalent of like a huge holdco principle buy by by just reserving in all the gamma ad servers and let chalice optimize the crap out of it. >> I this is I could find publishers do this. I know publishers who do their own version of this today where they'll call pre-bid and they'll say pre-bid if you can hit this price I'm not going to hit GAM.

60:22 >> yep >> because because there's a huge premium on speed. pre-bid is faster than GAN. >> I'm sure all the RMN product people watch this and they'll be calling us soon to to to do this, but I mean it's a slam dunk. Lower CPMs and better performance you could bring to your RMN boss right away. >> And publishers will like it because they get guaranteed revenue, which they >> Yeah. >> need. >> And the public ones could tell them they're participating in RMN guarantees.

60:50 >> Yeah. I so I think the the future the competitive landscape for this is you use you you move this logic into pre-bid and don't get me wrong there are some really big pubs who have built like these magnificent creatures on top of GAM using GAM's APIs like entire internal workflows are built on top of it etc etc moving them I said it in my article right like >> moving ad servers is on this like podium hemorrhoids >> hemorrhoids in moving houses. it's the same category. But >> I was impressed you could spell hemorrhoids. That's a really hard word to spell.

61:29 >> You know, I actually got it on the first try and I was really pleased with myself. I actually waited to see if the red line was going to be under it cuz I was like, there's no way I spelled that. >> Must be a great speller. >> I I was a regional spelling be champion when I was in middle school. Thank you. That's right. >> I rebelled against spelling. >> Why speller? >> >> Adam rebelled against a lot of things in high school.

61:58 >> Started early. >> Yeah, some reading teacher. >> I later got good Marley. That's why >> I I think this is a future that could exist though. I hope some pre-bid people are listening to this. I think that there's been a leadership change at pre-bid. There's been a lot of work there, but this is something that could be really interesting. it's it's agentic adjacent, right? because the agentic stuff could all fire into this new priority system. And this actually makes like the promise of Agentic attainable if pre-bid is in charge.

62:34 >> GAM makes GAM is what makes everything so complicated. Gam is building that automates line items in GAM is just a task that I would wish upon precisely no one. >> you heard it here first, folks. Adam and Gareth's ad network competing with Google Buyer Direct. Lower prices, better performance, agentic activation coming soon. Call now. >> I'm in. We can build it. It won't take long. Let's talk to Patrick.

63:06 >> Just Yeah, we need a big guarantee. Can't We're not building on spec, you know. We need a big guarantee in in a big fancy pub. Yeah. All right. >> Deliverable. Thanks everybody. Thank you for joining. >> Thank you. We'll see you next week. >> We'll see you next week. >> Hallelujah. Hallelujah. >> Darn. That's >> Yo, the crowd is back. The lights are dim. Front rows buzzing. It's about to begin. Ad talk to taking the stage. Adam and Gareth on a knowledge rampage.