The Grand Bargain (ft. Matt Sattel, OpenX)

0:00 Listen, get my raw owning the day. Programmatic [music] flows in the scene real tight. RTB dreams lighten up I'm so pale. You're on it. Today's snow is crippling much of the Washington lowlands. 1 to 3 inches of snow fell in Seattle and other areas.

0:34 Let's do the same thing, guys. >> [cheering] >> I'm so pale. Red. Woo! Hey, everybody. Welcome back to AdTech Ad Talk. We are so excited to have you here today. We're so excited to talk about AdTech. We're so excited to do two shows in one week and to show off Adam's new haircut. It's wonderful. Mine was just wet. I just got out of the shower. >> [cheering] >> Thank you. I'm Adam. I represent the buy side.

1:00 And I'm Gareth. I represent the sell side. Thank you for joining us today. We did a live show. We did do a live show. Um it was a lot of fun. You know, the the bright lights the bright lights are a thing. I I see people get addicted to the bright lights. They like to be up there. Um but a reminder since we are not on a stage with bright lights right now, um please ask us questions in chat. We have a guest on today who will introduce in a moment. But as we are going through things, do not be shy. Yell at us.

1:28 Disagree with us. Say we're full of it. It's totally up to you. There are a lot of things to talk about in AdTech this week. There's been a a little bit of news. We're going to try not to spend the entire time talking about the Trade Desk, but we'll we'll talk about it a little bit because we we should. Um You have to ask. You have to ask. I really think this is a time in AdTech where people have to ask. They have to be afraid unafraid of asking the questions. Like AI is changing everything. Like everyone has questions.

1:57 And this is just a terrible time to be reticent to be a to be afraid to throw your question out there. Me and Gareth don't know everything but we'll talk about almost everything with some knowledge or reference to someone else who knows and there's a lot that we do know about optimization and and AI especially so Let the questions fly. Yeah, and the pipes man. I think um I have something I spend a lot of time thinking about is the spring from which everything flows Mhm. is buyers.

2:30 It's your people Adam. That's right. I am a passenger on the ship and you guys are the fuel. So whenever we talk about AI people are getting lost in a lot of AI implementation things um that I am like oh just ask agencies what they're doing man. They're just ask them what they're doing. That's that's all you got to care about. When I get a question from an advertiser or an agency about the future like you know will Google open up to custom algorithms and I've been answering it's up to you.

3:02 >> [laughter] >> It's your money and guess what if Google doesn't open up to custom algorithms you don't have to spend money with them other than for YouTube. Um They you know they definitely won't if you never bring it up. They're they're very happy just pushing you into PMAX. It's the truth man and I I really do like that train of thinking cuz going on the trade desk I've been reading a lot of Jeff Green stuff recently. Um and all the the ad exchange article about the user ID thing aside cuz I think that that was a nothing burger. Um his point he he hits it really hard is that I he's banking on sophisticated CMOs.

3:42 And he is like I'm taking a 30 year view here. Where sophisticated CMOs are yeah I know it's a long view. It is though because he's he's banking like The The could be our children. I hope not. Like his his thesis is basically the arc of the universe is going to bend towards sophisticated CMOs wanting what the open web can provide, which is a level of control and transparency that a company spending $250 million a year on marketing should command.

4:16 And I actually think it's an interesting thesis. I think it's a pretty good thesis. It's just a really long one. Like it's it's like a Those those things are not moving fast. Those things are moving slow. And there are so many other businesses. Like what's the average lifespan of an ad tech business? Like 7 years? So we will have multiple crops of ad tech. I [laughter] think that's about right. Most of them croak, right? Most startups croak. Most people don't realize this.

4:45 Most startups croak. Most of them go out of business. The vast majority. That is why VC firms cast such wide nets. Like there's the there's a survivorship bias in what people consume where they're like, "Oh, that's a successful startup. That's a successful startup." Like they're reading about the good ones. For every one that succeeds, that has lots of revenue, that does really well, there are 10 or 20 that failed. And that went absolutely nowhere and nobody made any money.

5:12 So like the cycle we're going to cycle through a lot of ad tech companies before this new like universe comes into being. There's just going to be a crop of people who make money, extract value, probably don't add that much, and then disappear. Um And then there'll be a small set of ones that make it. I'll say one thing and I'll bring our guest on to comment about this because I think that the smarter the problem what I what what I think of when I hear the smarter CMO theory is that I mean that was my theory when I started Chalice, and I was definitely evangelizing that with a a quieter voice than Jeff's, unfortunately, and you know, in 2021, 2, and 3. And I think what I'm seeing now is that now there's so much pressure on them. Like now, you know, that they're so you know, like stockholders are you know, telling companies with their money that they're not growing fast enough. If they're they're not, you know, unless they're firing people and replacing people with AI. Like there's some really extreme signals to tech companies and agencies from Wall Street uh that are essentially, you know, AI or else. So, you get more efficient and effective from from AI or else. And it's I think I just think it's unfortunate. Like the last 2 years were the years to start adopting, and now I think the idea of the smarter CMO making better decisions is kind of threatened by this idea of the panicked CMO, or like just the, you know, stand-in for a CMO having to make uh you know, fast decisions under pressure, which might not play toward the what what we would consider smarter.

6:52 Mhm. Well, I tell you what, let's introduce our guest. Yeah. So, today we are joined by Matt Satell, the CEO of OpenX, and who we will let run through his background, but now that let's bring him out. Thank you so much for joining us. >> live from Chicago. Matt Satell. Great to meet you, buddy. Great to meet you, and and Adam, I I think you are over 7 years right now with Chalice, right? Uh no, we're actually just turning 6. We founded in the pandemic. So, 6 years, but it's our 7th year. Well, you're not going to be a you're not going to be a stat.

7:24 >> [laughter] >> I know, you're being mean it. I know. I tell my team all the time, we're [clears throat] in we're in exalted territory. Assuming you're not running a process right now. >> [laughter] >> Not. When companies are under process, they kind of disappear cuz they're so busy. So, look for that. If I disappear from LinkedIn. >> [laughter] >> Um, but yeah, I think so. I want to ask Matt about his background. So, Matt, would you mind telling us your life story, please?

7:52 Yeah, I'll go all the way back to childhood. Actually, we already did that, Gareth, a little bit. We got ad tech kid? Yeah, maybe raising some. But yeah, I I think I come from a little bit of a different background than probably a lot of the the standard SSP leaders of today. I came much more from the buy side. So, prior to OpenX, I was at a company called MiQ for quite a few years running their in-housing team. So, working very closely with with large Fortune 500 brands on bringing more of their programmatic in house. That was ultimately a big trend in those days and it was pre-COVID into COVID. So, a lot of conversations there.

8:34 What's interesting is when I was talking to to these brands, when we first started building out this in-housing team, it was really heavily centered around how many DSPs should you be working with, who are the right DSPs. And then we got a lot more sophisticated with like S-chain analysis and really understanding that, you know, some of these brands we were working with that we were doing audits on were running on, you know, 74 plus SSPs with average of five hops and really like focused on how do we eliminate that ad tech tax as much as possible. And then that was again on my kind of buy side MiQ days.

9:09 But that actually is really what led me to to the move to to come over to the SSP side and start focusing on supply was I saw just how much potential there was for more working media dollars and driving true innovation with SSPs. And that's what ultimately drove me to make a change from the buy side over to the supply side about almost five years ago now. And I think it is actually just continued to evolve and change even more than I could have ever imagined over the last 5 years. So, I think I can be proud to say that I sit somewhere in between the two of you guys where I I have had experience representing both the buy and sell side. And um so, maybe I can maybe I can mediate some of your conversations and that can be my role on this.

9:52 That's what the show's about. Great. So, >> [applause] >> I I have a question for you. Did anything surprise you when you came from the buy side to the inner workings of one of the largest ad exchanges? Yes. You know what's funny? One of the biggest things that surprised me was how people on the supply side define SPO or how they historically defined SPO. So, when I was at um on the buy side and and at MIQ, when we were talking about SPO, we were talking about how do we get the most direct path to the most premium supply. Um and then commercial terms came second. So, it was like, well, if we are, you know, if we're going to push more money to you and strike a bigger partnership, let's come up with favorable commercial terms. Um when I got to the supply side, um you know, I would be in meetings and I would hear people talk about SPO as like a financial or commercial strategy as opposed to a supply strategy.

10:49 Um so, I would say like my first 3 months, I was very confused on every call being like, what I don't understand why we're describing SPO in this way. Um because inherently I don't believe that SPO is a financial agreement. I think it is a strategy to have clean supply. Um and the commercial strategies come because you're consolidating your spend somewhere. Um I think that actually is probably one of the biggest shocks I had is how the buy side and the sell side think about SPO. Um and you know, I think we've taken at OpenX a strategy where we want to focus more on the innovation and a clean and healthy exchange. And that leads to SPO. Um and it it it has more so with transparency over the last couple of years. Um But truthfully, that was probably my biggest what's going on moment when I when I switched over.

11:41 That's super interesting. So, when you say that pricing, I feel like the pricing agreements that publishers are striking are rarer in like banners and app, but are not rare in CTV. Where in CTV, publishers really think about this as like, "Oh, okay. You want a direct. This is the price that we're going to do this ad." Yeah. And yeah, I it's really it's fascinating to me cuz it reminds me of old banners. Like if you like when like ad exchanges first came out, people were like, "Oh, you want page level transparency in your bid requests? Great, that's going to cost you this." Yeah. Cuz like there were literally buttons in exchanges like exposed page level exposed pages.

12:22 By default, you just got the domain and you were buying on the domain. And that was that was a sales technique, right? The removal of transparency was a sales technique in old school ad tech for domains. And yeah, it's it's back. What's old is new again. Yeah, I think it's it's interesting to think about the like dilemma between CPMs and outcomes and how they very rarely are match, right? I think about it back in the days, you know, I worked at Publicis on the buy side many many years ago and that was early days of viewability. And I remember, you know, starting to do buys with, you know, 70% viewability standards or something like that. Again, I feel like we're doing like an ad tech history lesson because it's just like isn't what you would normally talk about, but I think about it a lot in kind of what's happening now, too, is viewability became this thing where they're like there's actually just less supply now that is meeting the standards that you're looking for. So, you were going to pay a premium on it and there's a lot of pushback there. And I think the same thing is happening in the ecosystem right now in ad tech where there's a much more push towards premium supply, especially in CTV.

13:28 Um and there's an expectation that CPMs change, right? Less supply, more demand equals changing how we look at pricing. And I I kind of think that a lot of the things that we face right now can be tied back to things that we faced, you know, 8 9 years ago, maybe a little bit more, 10 years ago with things like viewability where it's not so different. It's really around the more premium or the more you're getting what you want, um there's a difference in price.

13:55 Mhm. I want to zero in on the Yeah, the clean supply comment. Like it makes a lot of sense to me what you said to OpenX when you came from the buy-side that, you know, buyers are choosy and, you know, if you want them to commit, you should show that you have something worth committing to and then talk about price, right? That makes perfect sense to me. But I don't really know what what it means on the supply side to say we have clean supply or what values they would put forward besides, you know, low CPMs and the most efficient path.

14:26 Yeah. Um so, I think there's a couple examples for us. So, you know, we can obviously look at some of our public competitors. We process about 500 billion requests daily. Um and, you know, one of our competitors does about 1 trillion requests daily. If you look at like a Jounce report, directness is pretty much within a percentage point or two, um very similar, right? All the top SSPs for the most part have a similar amount of directness in supply.

14:52 Um so, how does a um an SSP of OpenX size have 50% of the reque- or like processing 50% of the requests as another SSP with with which are one or two percentage points? It's MFA, it's indirect, multi-hop, it's how many paths can you get to the same supply or the same inventory? Um and for me, um that is a really key stat. 50% of the requests processed daily but about the same amount of directness, um that to me signals a very a difference in the cleanliness of the exchange. Um, I would also say we actually exclude about 50 billion requests daily because of our supply standards.

15:32 So there's a lot of focus that has been done at OpenX over the years and you know in some ways it's made things more challenging. It's it's not particularly easy to do the same commercial incentives as others when you're heavily focused on premium supply. Yep. So are you guys taking on what are the big by side bugbears of the moment and big duplication and ID bridging? Yeah, I mean I think we've been pretty public about request duplication and the fact that we I mean I think that there is a place for request duplication in certain areas if there's a particular signal attached to CTV that might deem a different price, right? But we do not you know there are others that have buttons within their publisher platforms that allow request duplication.

16:17 Those are the areas that we don't play in. So we've taken a pretty firm stance where we do not do request duplication as a standard practice across the exchange. And ID bridging Gareth gets really fired up about this so be careful. Yeah, I mean ID bridging is um He has to do it. He doesn't have a choice. [laughter] You want you want OpenX to be competitive in the prebid auction you got to do some ID bridging man. It's the world we live in. I do they outsource it or do it?

16:44 Outsource it. Outsource, yeah. I think yeah, I mean I think listen, I think there's a if we're having the honest conversation there are things that everybody in ad tech has to do to to compete and and whether we believe that's right for the future or something that we want to continue work on with some of the standards. I'm sure you guys are pretty familiar with like the grand bargain that's going on right now across some major agencies and tech partners.

17:09 For for me it's around how do we continue to have a voice? Say that again. What's the grand bargain? The grand bargain is it's a conversation across is that a rhetorical question or >> No, I'm not in I'm not in this conversation little >> I am I am You're in it? >> I'm outside. Yeah, um I think the grand bargain is something that is ultimately a is a grand bargain between the buy and sell side where it says here's what the buy side is looking to get access to in terms of transparency things like ID bridging request duplication and here's what the publisher world is willing to the supply side is willing to kind of come with and DSPs and SSPs play in that and it's with the goal of creating a standard that makes both sides happy.

17:55 And for me I think we have a hard time with standardization in this industry and I think partially that is because a lot of the industry bodies are made up of you know primarily like product and tech teams or the commercial teams and often those teams aren't communicating of like this works from a product and tech side but this doesn't work for like how I communicate with my clients. So for me the grand bargain is interesting because it's bringing together different areas I can put you both in contact with some of the people that are that are kind of I mean it's definitely it's part of the scaled tech world that that I don't know I can't speak for Garth but us as a modular AI company like we don't those aren't I just I don't want to say those aren't problems for us but they aren't as big problems for us ID bridging and and bid duplication bid duplication isn't a problem for all the ways Chalice operates today we would answer the exact same bid to the exact same bid request with the you know very low energy expense so we don't worry about that and the ID bridging since we measure so much offline outcomes like it comes out in the wash if there's if it it just looks like a non-conversion like you can't trick us for a conversion through ID bridging the way we're we're doing things so I understand that it's important for the scaled software companies to have standards so they can operate but it a little bit to me it's like, you know, you know, it's trying to fix an old world that's that's going away anyway.

19:19 Yeah, I mean, I think request duplication often is gamifying DSP algorithms, right? To to get different pricing based off of you know, that gamification of it and I think Challis does a really good job of weeding through that based off of algorithms and the work that you know, you do with things like our real-time bid stream API. Yeah, so I have two thoughts on this. One thought on this is if you really get abstract with it, a really good ID bridger is actually an optimizer um because a really good ID bridger is going to send you user IDs that convert.

19:53 So, not just ones that you will buy because they know the ones that you will buy. [laughter] But they're going to send you ones that perform >> But that has nothing to do with the graph? So, it's not that you That's what you're saying? If you take probabilistic ID matching to its full conclusion, you will build an optimization system that finds users that buyers want to buy, which means you will build an optimization system that finds users that convert for campaigns.

20:18 Um it is Now it's it's a functional system, right? It's unaware of the conversions. However, it is watching buyer behavior across lots and lots of impressions and it's finding which users are valuable and which users are valuable and >> your SSP has an algorithm All of the probabilistic ID bridgers have algorithms. Right, you say but you're saying they have a conversion prediction algorithm and they just serve IDs that are going to convert even though that's not the user They don't look at conversions. They don't look They look at bid price, >> [laughter] >> right? They don't need conversions. They don't need to know the conversions are happening. If you're bidding high for them, they're likely to convert. Bingo.

20:51 Um and if they're real smart, they're going to send you users that are good for you. And I mean, you can This is a big rabbit hole you can go down. There are There are optimization systems on top of optimization systems. I'm just telling you this is a This This is a thing that occurs. And there is like ostensibly value added there in some strange, twisted way. Um but like it's a thing. Um But my other thought on this, uh when it comes to the standardization, all of these things that we have now, right? The the ID bridge environment, right? The probabilistic ID graphs where you know, Procter & Gamble's out there, they probably have a list of 50 million user IDs. If people can figure out which 50 million those are, they can make a lot of money because they can pass them into the bid stream and and resolution at the end of the thing is really like haphazard. Um But two is this this stuff makes people money today.

21:42 And it's really, really hard to shut things off that make you lots of money. It's a really hard thing to do. And so, the people who are in these rooms having these conversations about the initiatives that would shut these things off are normally people at the companies that are making lots of money doing these things. >> Yeah, and the advertisers whose money it is are not in the They are not. The advertisers are not there, and the publishers, right? The publishers who do lose working media budgets to this the shenanigans that occur in the middle.

22:13 They do. Um like I I truly think, and we've had this conversation, Adam. If advertisers really knew their working media, and I talked to Fiducia about this. Um if advertisers really knew their working media, they would be upset. Um and and a lot of them don't. Yeah, you don't think so? You don't think they'd care? They'd be Smart CMO theory. Smart CMO theory. Um but I think I think they That's why these conversations move so slowly uh and why people get so embattled over them. Uh because prior to the new infrastructure, right? So, prior to ARF and containerization and bidders moving into exchanges, I want to hear your thoughts on that, Matt. Um but prior to that existing, the way that SSPs differentiated themselves were be was being good at these things. There was no other way. Cuz like you said, Matt, the the top publishers work with all the same exchanges. They do. Um and so if you're going to you're you're in the header with like the three other big guys on premium domain A, how do you get better bids than the three other guys?

23:15 And the answer is you're good at these. Um even taking lower fees turns out moves the needle less than being good at these things. That's something that I When I was at Magna, I saw that. They were like, "Well, if you're the one who's able to get the high bid, doesn't really matter what your fees are." Like you're you're able to solicit bids that other people don't get at all. Um and at that point fees don't matter.

23:35 They maybe average out over time, but the scale's too big for that to be a thing. Everything you're saying I think of as an artifact of Google dominance. And I want to ask Matt, has any of that changed as Google has backed off a little from >> I So, I think I see it a little bit differently in terms of not disagreeing with you, Gareth, because I actually >> know. Please. When when there was a I was hoping for a couple of pauses through a couple of things that you said. Um But I would say it's around um >> [applause] >> Thank you.

24:05 Uh that was for Gareth, not for me. Um I think that the the big thing that I've noticed is it takes um all sides of kind of the ad tech system to agree on something for it to work. So, um I think one like clear example of that was MFA. You know, the you know, the ANA report came out a couple years ago and that was that probably like four years ago now, um calling out just how much MFA was um was taking on programmatic spend, which a lot of people don't consider to be like proper working media dollars, right? Um we took a pretty firm stance and had previously that we were eliminating MFA from our deals business and and often our exchange. And what happens is DSP algorithms are optimizing to um and and this is, you know, back when the DSP algorithms in that decisioning layer was almost always exclusively on the DSP side, it um it had a big impact on our business because, you know, media metrics and what people are optimizing to and scale is often what DSP algorithms were looking at. So, by us doing, you know, what was right by the industry and what buyers were saying and industry bodies were saying we wanted, it didn't necessarily benefit us. In fact, it had a pretty negative impact on spend and revenue for a period of time.

25:15 So, I think all things have to change, right? Like, SSPs can eliminate MFA, but if the SSPs who don't eliminate it are benefiting from DSP algorithms not making those changes and buyers aren't changing the KPIs that they're holding you to, it doesn't work. And I think that's often the challenge that we face is good guys often get penalized because there are people making money off of the things that are exactly what you said, Gareth, or are making them money. And so, it's that to me is one of the biggest challenges we often face as an SSP and transparently why I think a lot of SSPs are looking at um building bidders, partnering with bidders, and getting closer to the buy-side is it does allow us to make those changes um and have a voice in the room, which I think historically we didn't always have a voice in the room with the buyers to say, "This isn't going to work if you don't change your KPIs, right? If you're looking at video completion rate or if you're looking at, you know, viewability as a standard, MFA is going to win every time." Um so, it's changing the way that we're looking at things. I wrote an article about this called MFA is programmatic's black mirror.

26:16 And people really love to yell at the MFA publishers. They really love to blame them. They say they're scams or X Y and Z. And to which I reply, "They are giving the people what they want." That's it's hu- it's normally human traffic. It is. It's normally human traffic. And if you think that that MFA site was hitting your KPIs and you have no KPIs that would convince you otherwise, that is on you. Like, oh my goodness. That is if the if the buyer thinks if if it took Jaunce saying, "Hey, you're spending money on this website" for you to figure out that you were spending money on it, like, come on. Who's fault Who's fault is that?

26:54 >> measurement dominance. You're just under playing it. Like Google is a trillion dollar company that measures on last touch in their ad server and has access to your EO if you move spend because you don't believe in their measurement and yeah, that's how it went down in a lot of cases. There's a lot of people who wanted better measurement. There's a lot of companies that were trying to provide it, but there's there's, you know, the most powerful industry in our space by 20X was insisting on the system that coincidentally pollutes the open web with MFA and makes YouTube look better.

27:28 I mean, to be fair, they also led to like the search slop that the internet kind of is, right? Like the the search slop is a thing. It's gross. Like anyone who's used Google search knows that they're the out of the first 10 results, five and seven of them are just two ads. >> no one no one talks about the the Kristoph [laughter] Franek reports, the analytics reports on the Google search network and the YouTube partner network were like the most explosive >> search network.

27:56 Those were the best reports, the most revealing and the timing was so good like around or right after during the antitrust case and stuff. I think it really like I I don't know. I think if it didn't swing spend, it really swung perceptions in a way that I found useful for for my sell and my evangelism is to be like yeah, of course, like they're polluting the open web and they have just as much pollution in their in their own networks or worse.

28:23 I mean, he had a report that had like how they were monetizing, you know, Iranian porn sites or whatever. I know, which of course the most evil porn sites in the world. I was going to say, >> [laughter] >> Iranian? Um, but uh yeah, I I really do like Matt's point though because I think that I don't get me wrong. I sell a product to detect inventory quality that does it very, very well and I have to be code on page for it. It's a whole thing. It's it's a great undertaking.

28:49 Um but the conversation about MFA, the fact that it got legs is very inspiring. Um I actually think Adam's in in light of what you're saying now about Google creating this last touch attribution environment, the fact that the ANA and that Jounce and that I mean like you have to give Jounce credit. You have to give Chris Cain credit. He did such a good job. Like it is such a a mitzvah not to necessarily right not MFA in and of itself because I think MFA is it's like a a Pascal Lamb for caring about >> It was DeepSee who did the who did the research on the on the ANA report. Well, it was Jounce's as well. Yeah. Oh, Jounce I think Jounce coined the term MFA. Oh, yeah. Well. Yeah, I think I think that was Chris. I think they've also kind of held the gold standard of of like an ongoing list of what's considered MFA versus not and you know, it's it's challenging. It pops up all the time so there's no there's no way you can ever say you've eliminated MFA completely, right?

29:45 >> We we never claimed No, it's coming back. It's going to be back. Um there are some people who still want it. It it drives value for people who believe in last touch attribution and they don't care about context. It's it's it's not valueless. Uh it's not right or wrong. Um yeah, I think that There are a lot of really bad websites out there that have lots of ads on them with really crappy content that are not MFA or I mean well, everything's MFA anyway. But um this is a but that aside, the movement to care about inventory quality right? There a whole there's a whole slew of vendors about this, right?

30:17 Like attention vendors, the notion of attention um buyers saying, "Oh, actually I do care about the quality of the sites that I run on, right?" The the PMPs that you build, Adam, right? Then the you ingest the signals that you ingest. The the fact that ad buyers care about this is actually a little bit of a breaking of Google's hold, right? In in your >> Absolutely. in your drawing. Like this is a break of Google's hold. Is that Let's uh before before I let Matt get to that? I was I was something I almost said to publishers was that your market is splitting like the coffee market split when there started being premium coffee.

30:52 It's like it used to be every coffee farmer got the same price for their coffee, right? Commodity coffee was all there is. And then Blue Bottle and these other companies start being like no, we want to buy this totally different product. They get that they have to rip out their plants and grow again and you know, grow different plants and the way they're cultivated are different and they get a much better price, but it's very risky for them because you could blow it and not get that price and then you have nothing.

31:17 Or, you know, the the premium coffee market just is unpredictable, right? You might not be able to find the right buyer or selling it to Blue Bottle and you lose Blue Bottle. Well, you know, so it's is very I read a book about this, the The Monk of Mokha by Dave Eggers, it's really good. But there's it's very risky for the farmer to switch from commodity coffee to premium coffee. Uh even though premium coffee is drives a much better price, right? We I buy I bought an $8 cup of coffee yesterday.

31:43 Like it's it's so much higher. It's 8x the cost of uh of of commodity coffee. And I think that's that's happening to publishers where they're going to have this split market. It's like everything you've been doing is essentially been selling like MFA and now that these people now that these companies coming along who can tell the difference and though it's smaller, we're bringing we're going to bring much higher CPMs and you have to decide, are you going to start to yeah, grow the premium product?

32:08 It's look looks a lot different. Yeah. I mean, >> [clears throat] >> uh we, you know, in talking about I know we haven't really talked much about our like rebrand to the intelligent SSP, but everything you guys are kind of saying right now is almost exactly why we have been thinking about our role at OpenX's role in the ecosystem and how we needed to pivot away from kind of the that commodity dumb pipe mentality that people had around SSPs.

32:35 And I think a lot of the work in fact that we're doing with Chalice is a really prime example of that and Adam you gave an example when you came in and talked to or an analogy I would say and when you came in and talked to the OpenX team a couple weeks ago around Legos. I don't know if you remember that or if it's something you say often. I'll I'll attempt to to do it first and then you can you can tell me I'm wrong but um but thinking about about everything as a Lego and I think everyone knows that Legos always have a base plate.

33:06 OpenX is really centered around us and our infrastructure being that base plate and allowing others to come in like Chalice and build upon that. And that's what OpenX build and our containerization solutions have ultimately started to do. And what we've seen is when the SSP is helping to define because we're closer to supply and much closer to the signals that matter. When we are directing the right supply to the DSP, we're seeing massive increases in performance. And sometimes performance looks like look like efficiencies, right? It's lower CPMs. Sometimes it is a real outcome but what we're but for me it's the more of that decisioning and intelligence is brought to that infrastructure of an SSP and we start focusing more on how do we get the right supply that is going to drive performance to the DSP, everybody is winning. You know, we don't have to have conversations with DSPs around QPS caps.

34:00 We don't have to have conversations around, you know, you're sending me the firehose. It's really is around we're sending you what is likely going to drive performance. And I think there's a Chalice case study that's coming out I think in the next week and you know, we're not allowed to say who the brand is but ultimately the Chalice AI models and being able to actually send and identify the supply that is likely to reach an auto shopper which is this was for a major auto brand. It was a 60% lower CPM.

34:30 Which is ultimately the metric. It was insane. great example of when the intelligence and decisioning sits with the SSP and partners like alice, you actually see the right publishers getting spend because of our focus on, you know, driving to premium publishers and the cleanliness of our exchange, but you're also driving better performance. And all of that can be done without raising the price of your CPM. And I think that Ch alice case study that's coming out next week is just such a prime example of how if we really start to pivot the way we think about traffic shaping from the SSP side and the DSP side, that there's huge wins to be made in the efficiencies in programmatic as well as working media.

35:12 You know that and the only thing I'd add is that, you know, what you called Legos is the true [applause] That's the true agentic, right? Because in order to achieve what Hyundai achieved, you know, you don't have to make a bunch of different deals with OpenX logs and Ch alice optimization and we use DeepSee for quality. These things are are arranged just because the advertiser is saying, I want it. And that is the true agentic. You don't have to force the integrations, you don't have to wait for certain companies to be integrated, you know, you could you could swap out Ch alice for another optimizer or swap out the DeepSee for another attention company and get and you know, maybe get just as good results if they're just as good as us, which they're not. But the But the idea is the true agentic means the advertiser chooses which partner they want to work with and those partners are integrated on a a platform, you know, a shared platform like OpenX. So, um yeah, I hope that I hope I hope that case study uh gets a lot of interest and I hope we successfully connect it to agentic cuz I think the message that agentic means automated has confused a lot of people and it's actually much simpler. Agentic means you get what you want.

36:24 Yeah. Yeah, I mean, And think ultimately his historically when we have thought about traffic shaping from the SSP side, it's looking at historicals and understanding what we think the DSP is going to buy. Um done at kind of the total DSP level. And I think what we're doing now with uh this this us becoming the intelligence layer and working with partners like, you know, Chalice, it is actually um taking on a different form and we're actually sending the supply that matters to that deal or that account as opposed to that DSP. And um and there's just a lot of efficiencies that are made with that and >> Yeah, I mean, you're elected to call it agentic because that's the definition of agentic. It's like we don't send you everything. We send you what we intend to send you, which is what you intend to use. Yeah, no, we ultimately I mean, OpenX Build, which we launched a couple months ago, um one of the uh AI solutions that are within OpenX Build is that real-time bid stream API, which is ultimately what both of you guys will are, you know, looking to use. Um but we also have two others, which is um our identity resolution API and our auction insights API. So, for me, when I'm talking about AI, I think about it in that form. I mean, listen, we've been doing traffic shaping for many, many years. That's been AI, right? Our you know, we've had a data science >> containers, right? And you do it in a Kubernetes container, right?

37:39 >> Right. Uh so, I think like what my pushback on like calling certain things AI or not is a lot of the stuff has been happening for a very long time. It's not like new to an SSP to be utilizing AI for traffic shaping. I think the sophistication is is increasing and I think we're doing it more intelligently um because we're working with partners. And again, not to keep bringing back that LEGO example, but like as our infrastructure and our exchange becomes that base plate and allowing other people's other companies to build solutions on top of that, um we're seeing again, I just like I can't say it enough, we're seeing better performance at or similar CPMs. And in some situations, if the goal is lower CPMs, it's lower CPMs. And to me, like those are the things that we should be talking about in the ecosystem is the efficiencies that we're driving.

38:26 I think we can get very negative on what's not working, but I think we should also take it like a step back and think about some of the evolution we've had just over the last year, right? Like Yeah. this whole idea of containerization, our partnership is hasn't been that long and it has the ability to really drive an incredible amount of change in the industry. You guys are super fast. Your tech team is the fastest we've ever seen. I want to ask you about that. How do you have such an amazingly fast tech team? So Meyer is amazing to work with. He's amazing.

38:53 But just the the speed you got going is is really something else. Uh and uh yeah, Marley's chiming in that we we love that team. Uh and Gareth is Gareth is one of the more optimistic about the open web people that you you could find in across the whole landscape. So Gareth, I'm going to call out this in this case study we had we were getting OLV for like YouTube prices. Like the OLV was quality OLV at the price you pay for open YouTube, which is unheard of.

39:19 You don't And it was all because of the the slicing we're able to do at the URL level with with a whole bunch of text data. I am an optimist. But I'm going to rain on the parade a little bit. Yeah. Because every time someone uses the word throttle or QPS cap, I vomit in my mouth. >> [laughter] >> And the reason for that is we are having tremendous success here making pipes that are broken work better.

39:48 And I think that the and that and it's very encouraging. It is. It's wonderful. But like it's time to be done with them. Uh and I think that so the the bedrock release with Index was the first time a bidder said they were going in there. I actually don't find it to be that significant. I think that the the infrastructure is there to do the same thing in Open X. Like I could like I could integrate a bidder with your real time API. Like I can say it would it would be doable. Um I've looked into it.

40:15 It's kind of a lot, but anyway. No, it's it doesn't it's it doesn't scare me at all. You want you want a bidder Adam, I'll build you a bidder. I'm Um but it's I I think that the the next step, right, is to do away with the concept of throttling and QPS caps. They should they're just unnecessary. And if you can have some >> QPS. If you only count a a true users getting a >> how I You know what? I think I know what it is. I just say you know. I'm on about 10% of the internet I'm code on page. So deduplicated page views on 10%. Um and I don't work server to server with people.

40:47 I don't do that. That That 10% generates tens of thousands of QPS. So you think what do you think it is? 50,000 a second? No, wait. I Uh well, QPS is per second. It's queries per second. Um I think that the deduplicated QPS of the open web is is in the hundreds of thousands. I do not think it is in the millions. Um yeah. I >> but you have to subtract bots, not just duplicates. Yeah, you know, bots are a weird thing.

41:17 They're kind of in the eye of the beholder. Um Let's just say viewers without screens. Yeah, I I don't know. There are a few vendors out there who do this like it's There certainly is a large percentage of bot traffic. I don't think it's what Foo says. I think to to get Foo numbers, you got to go out there and try and buy bots. Um but which some people do. Um but like I I certainly think there is a percentage, but I I just think that they we've created an environment where we reward high QPS, right? Uh because there's there's volume bias like Garrett who I used to work with always had a volume It was volume bias, right? The more a QPS I can hit a DSP with, the more revenue I get. And that sucks. That is that is an environment that is like uh oh, am I including app or just desktop or mobile web? Uh just desktop or mobile web. App is its own horrifying monster thing.

42:07 Uh cuz there are there are waterfalls in apps, so deduplicating QPS is quite difficult when there are waterfalls involved. It's similar to CTV. Um like if you were to look inside the OpenX Exchange, I'll bet actually a at least half your QPS is CTV. When in reality, if you think about the quantity of CTV advertising opportunities relative to the banner economy of banner advertising opportunities, like oh my god. I have >> [laughter] >> I have 10,000 TVs in my house showing Pluto, so don't say those aren't real.

42:34 They are real. >> [laughter] >> 10,000 ads on every every ad experience. Um it's wild, but it's cuz waterfall environments inflate QPS really badly. Um but I think that this new architecture where no longer do we have this DSP thing that we make all our money from, and we live a life of just thinking how do I get it to pay me more? Um like cuz that's that was the exchange's job for a really long time, just staring at a DSP and thinking how do I get it to pay me more? Um and smarter throttling is a way to get it to pay more.

43:09 Let's just uh Once that's gone, right? And we're talking, you know, a million QPS for all the things, like the the cost of compute for that is just not that astronomical if you have 10, 20% of advertiser budgets flowing. And you know how long the average prebid timeout is, Adam? This is a good one for you. You know how long the average prebid timeout is? Prebid JS, how long you have to respond? It is anywhere between 600 600 milliseconds. Wow. Why do they take so much time?

43:38 >> 1.2 seconds. That is how much time we actually have to play with here. Because in the world of five or six hops that Matt was talking about, And you needed all that time. You needed all that time. And if we're just talking about now there's just OpenX, and there's Chalice using Gomera data inside of OpenX, we have a 600 milliseconds to respond? Like A lot depends on the whole code >> every data point in the universe. Like [laughter] that's a thing. A a depends on the hold codes investing in this stuff. And that Have you seen any signs that the hold codes are interested in what Gareth is talking about and and taking this efficiency leap? And, you know, sharing some of the benefits with advertisers so they finally get engaged in cutting costs out of the middle.

44:22 >> could keep them. Yeah, I I I was going to say there's a there's a couple questions in there. Uh yeah, I mean, I think that we're hearing a lot from the holding companies. I actually think one of the more interesting areas too is that we're also hearing a lot from independent agencies, probably more than we ever have in the past. And you know, I think Adam, you you and I have had this conversation quite a bit around how do we bring this same level of kind of sophistication to independent agencies that, you know, don't necessarily have the same resources. And I think, again, this whole building or base plate building block um metaphor really helps with that because we have a lot of ability to work with, you know, different partners to create better programmatic campaigns for independent agencies, too.

45:08 Well, let me ask about this scenario cuz it feels like this is what's happening at the moment. Like, Azure and Google Cloud and AWS are trying to do big guarantees with the holding companies around the Gentic, right? That we're going to be a Gentic portal. And Anthropic doesn't seem to be in there, which seems like a miss, but maybe they'll come back in the second round after >> [laughter] >> after whatever happens in the first. Yeah. But, if uh if they become like the Gentic seats and the way that they want to access supply, then I think it really does become, and Gareth will object to this, like, a smackdown between ad CP and ad and the RTF, right? Because then the agencies are like, "Oh, should we just move everything to guarantee?" Cuz now we have this portal. Cuz if you try from Azure to buy in real time, like, you're not going to get good results.

45:54 Um and there there's there's a lack of connection there. And I'm worried they're going to make these big deals and you know, and then it's going to break or otherwise upset the real-time market. Yeah. Um, you know, I think uh Omnicom just recently had an announcement where they're already testing agent-to-agent like real-time campaigns. >> they buy a lot of guarantees, so if it's for guarantees, whatever. Yeah, I mean, I think the holding companies have an ability to do that in a way that independent agencies just can't, right?

46:21 They already have kind of marketplace teams where they're forming direct relationships. They've been historically much more um focused on kind of their those one-to-one buys. Um, but I think that that it it's really challenging for even for publishers to to maintain that many relationships. So, for me, it is actually um I get really interested in kind of the agent-to-agent conversations that some of these holding companies are having, but the way that I see it is that the data that we have access to, even just kind of the um real-time bid stream API, can continue to be like a very important part of that, and that's what we've seen um with holding company conversations is actually a big emphasis on AI, but it's utilizing things like containerization to actually help power decision.

47:02 >> Right. That's encouraging. Now, and Garrett, let me ask this, if if they sell all these guarantees guarantees agentically, are there any sell-side ad servers that can handle any, you know, specificity in these guarantees, or they just going to be like upfront guarantees where I, you know, a million dollars of whatever you have on ABC? Like, I feel like the promise is like you're going to be able to be much more granular, and you can use your agency resources, like maybe your agency audience buying to say this is what we want and don't want. But from the publisher side, I hear, "We can't manage guarantees that are against different audiences and different geo preferences, and like we don't have the tech to do that."

47:40 >> there there are vendors, right? Like me. So, I have a direct sales product for publishers. That's actually the number one paid product I have is I have a direct sales tool for publishers to help them improve the performance of their direct sales. Uh and I can do that cuz I'm on page, and I understand publisher inventory better than anyone. Um, and I have an MCP server. So, in the theoretical ad CP world, an agent could query me inside of a publisher and say, "I want to deliver a campaign at 90% viewability across this publisher."

48:11 It would generate a targeting segment in me that I would insert into the publisher's ad server and would deliver on their direct sold campaign. But then it's ultimately Google that's doing the selection. Right? >> Yeah, well, Google's ad server is the ad server. Google is the ad server. >> I just think that's the hard part. Like if they have 50 of those with 50 different parameters, is Google's ad server going to deliver them all? Well, you have you are now building line items on mass in Google's ad server.

48:37 Um, it can do it. It can handle a lot of volume. Um, now, can it handle an infinite amount of volume? No. Um, and cuz cuz this automation does already exist. >> mean I just mean that every impression that comes up will qualify for some and not qualify for others and they'll have to decide in the 6 milliseconds or whatever, you know, what to serve. Like is Google's ad server going to It'll be real binary, right? So, in the situation I just described, I'll generate a key value pair that attaches my performance targeting. There'll be another vendor that has an MCP server that handles audience with the agency somehow.

49:11 Um, via like a clean room or via something. They will send a segment down to the page as well. So, in the line item targeting, it will say, "I'm going to deliver at a price of $2.50. I'm going to target the Gamera performance key value pair. I'm going to target the DMP audience segment and when both of those signals are present, I will deliver this line item." Um, that's a thing. >> the best price. If it's the best price, right? If it's the best price, maybe maybe not. You can you [laughter] can have the concept of priority in there where you can give people like this ad servers are made to do this. Like they they have OMS integrations. There are already companies that automate this. Like when I talked to Media Ocean, Media Ocean's like, "We don't understand. This is our business model."

49:54 Like >> [clears throat] >> Like, we've been doing this for 20 years. We integrate via API with Google Ad Manager and set up line items based on RFPs. We do this. But the delivery is Yeah, I'm talking about the real-time delivery. I Yeah, the ad server is pretty good, man. They're It's It's It's a lot less complicated when you don't have to consider the entirety of the internet, right? They have like their publisher objects. There are not that many of them. And you know, here's the thing.

50:23 AdCP could become its own enemy, and if it has to deliver via GAM line items, that's a nightmare. The latest AdCP stuff I've heard is they're going to deliver in via prebid line items, which is its own Now, it's just a prebid bidder. So, you basically have an API layer for buyers to set up bids via prebid. >> All right. And they're just going to bid flat cuz it's a guaranteed. That's That's not going to work. Well, you can add in priority. You can do all kinds of weird stuff in prebid. But this is This is all to say that I think that the value of guarantees is way more interesting in CTV because CTV has this horrifying price asymmetry that exists where if you set up a campaign buying in like premium in open market CTV, sometimes it will deliver it at a wildly different CPM than if you call up a salesperson at the pub and go, "Hey, I need you to like traffic a tag for me at this CPM." And the salesperson just goes, "Yeah, sure."

51:14 That's a mess. So, I think this this automated direct sales workflow could have teeth in CTV. It could be a thing there. But in banners, I You said this, Adam. Bundling is just going to perform worse than per impression decisioning. It will always be the case. There's no way to avoid that. Gareth, one thing This is a little like slightly off topic, but you might find interesting. One of the things that we've been seeing hearing from our like the ad servers that we work with more is differentiated demand for publishers as well. So, finding SSPs that are driving ultimately differentiated demand to the publishers and prioritizing those bids because it's not, you know, one SSP could be or like, you know, one all the other SSPs could be sending the same publisher the same campaign, the same deal, whereas you know, someone else could be a completely different advertiser, completely different campaign.

52:05 And so we've actually seen our relationship with ad servers change because of our kind of unique demand story. We have a large independent agency business, but that's one area that I don't think ad servers always were something an ad server wasn't always doing to advocate for a publisher where it's how do we get you unique demand that's actually going to help drive incremental revenue for you as opposed to that you would have gotten the same thing whether it was one SSP or another. Um I mean, there were I it's really cool. I love to hear that. I mean, there's going to be a proliferation of ad ad servers. They're coming back. Yeah.

52:39 >> Um as long as long as like the Google ruling actually comes to bear. Yeah. >> Um and that's It's the big if, but >> a year. I think it's been a year. I know. I know What I think is interesting I think about it is like, you know, the role of an SSP has evolved quite a bit over the last five years. I think the same is happening with ad servers where the value of the ad servers are starting and will continue to become more and more interesting even in the AI world, especially Could you could could you host a containerized ad server?

53:10 We'll have to ask Joel. >> [laughter] >> Yeah. Oh, you definitely could. Oh yeah, you didn't answer how how's the how's the team so fast? Can you have any Do you have anything to say about being the the fastest tech team Charles has ever worked with and we've worked with a hundred at least. Yeah, I mean, there's there's a couple of different answers when it comes to speed. One, Joel is absolutely amazing. I think the way that we have structured our engineering team. Keep in mind like we are while we are a top SSP, we are significantly smaller in terms of the number of people that we have.

53:38 So I think we just don't have the same bottlenecks. We can make decisions pretty quickly. You know, Joel and I talk almost every day, so it's not a, you know, five people have to make a decision. Um, the other side of it is we are cloud-based. Um, so, you know, if Chalice comes to us and says we are expecting a traffic spike, we can increase, Garrett, sorry to say this word, but uh, but QPS almost immediately, um, which actually helps companies like Chalice actually start to scale their own campaigns and and their optimizations, which again drives better performance and more you have access to the higher performance. So, I think from a speed standpoint, there's the just the element of the team that we have and Joel just being an absolute like rockstar engineer and overseeing product. But I would say on top of it, it's, you know, the decision in 2019 to move to the cloud has been uh, it was a huge bet. It was before, you know, really anyone else was moving to the cloud. We made that bet. It was a an expensive, challenging, um, Yeah. shift to make.

54:35 >> It's not a cheap shift. No, and you know, we we had the we had a strategy behind it, um, but I don't think we ever could have imagined the focus on AI and all this, which, you know, being in the cloud really does help us be AI ready as well. So, for me, I think, you know, moving to the cloud in 2019 is probably one of the biggest answers to speed because we can react very quickly to But it also means there's probably significant re-architecting in 2019, which was, you know, pretty late in the SSP story, right? You probably were able to rip out a lot of tech debt by making that switch, yeah.

55:09 Yeah, and then continue. Uh, you know, like I think there's a lot of value in having Joel be across engineering and product. Um, I know that's often, but, you know, I think it's often pretty unique that your the person that's running engineering and product come from an engineering background. Yeah. Joel comes from an engineering background and I think he is evolving to be like one of the stronger product people and building a really strong team there as well, but I think being so engineering and infrastructure led, uh, the you know, having a strong base plate infrastructure is is a huge advantage for us. And something that I will I think will only continue to become more and of an advantage as containerization and our products Open X build and our APIs continue to kind of grow in market.

55:50 It's having that solid infrastructure is so incredibly important and and moving to the cloud is a core reason why our infrastructure is what what it is today. If our listeners want to hear more about containerization and hear from Joel Meyer himself, just search container explainer at Tech Ad Talk and you'll find our special episode explaining containerization with a commentary from Joel and and the other engineers who pioneered it. You did that at CES you guys [applause] You guys had a panel at CES too, I I believe. Yeah, that was Yeah, the video came out of the panel. Yeah.

56:26 You know, I worked at a very large exchange that really wanted to migrate to the cloud. The CTO at the time really wanted to make it happen. And I can tell you it's very daunting when you look at cloud egress bills and like how much money that will cost and you have your own on-prem ancient servers and you look at what those cost. The the the differences are in order of magnitude. It is a real amount of money to bite off and that just creates a lot of lethargy in like people wanting to actually do it. And I'm not sure they ever did it. No, they really didn't. Uh this is I mean, it was it was a constant topic of discussion where I was like, "God, how do we get off of this? How do we get off of these ancient boxes that we own?"

57:15 Um like how do we do this? It's really hard. It's it's it's uh It's I think that one of the refrains that we're seeing now is the exchanges who had were able to do that and make that bet are the ones who are here to ride this wave now of oh, buyers want to get closer and we have a way to bring them closer. Um and the legacy ones who don't have that and that's not there.

57:40 Let's let's uh use the last 3 minutes to let Nat explain uh the OpenX positioning and rebrand. Yeah, please. Yeah, I can uh probably don't even need 3 minutes at this point cuz I think we we covered a lot of it throughout the conversation, but yeah, we uh we launched our rebrand um as the intelligent SSP and um I'm not going to get too deep into it because we've I've I feel like I've uh highlighted it quite a bit throughout, but Let me Let me just comment on the positioning strategy. By calling yourselves the intelligent SSP, you're calling the other ones unintelligent very subtly. Me and Marla have studied this back how this how this works, so Let's uh I feel like Adam I feel like Adam said it, not me, but >> [laughter] >> I think that um you know, Gareth, you've talked a lot about the SSP bloodbath or um that they they could potentially happen. Um I agree with one caveat that I think it is a consolidation of all of ad tech and I think the bloodbath can happen on the DSP side just as easy as the SSP world as well as data partners. Um but I don't disagree.

58:46 I think that there are many SSPs out there that are doing innovated innovative work um and I think those are the ones that are going to continue to succeed. I think for us that um bringing that intelligence layer to decisioning to um directly to our infrastructure is what's creating the narrative around the intelligent SSP and I firmly believe that what we're doing is actually making a dramatic change in the industry in an intelligent and strategic way. So, for me, I I stand by the intelligent SSP. I don't believe that that means no one else is intelligent. Uh All right. But I But I do think the SSPs [laughter] that are surviving are going to be the ones that survived that um like and and for me it is the evolution of infrastructure, the building block mentality and and I think one of the biggest changes is being a part of the conversation.

59:41 SSPs kind of lived in one world not talking to buyers as much, not as as active in certain industry groups that we are now, our industry bodies that we're working with now and I think having SSPs as a voice in the room is I Listen, having Joel sit on a industry board is bringing intelligence to that board and it's bringing intelligence back to OpenX and I think that's the pivot that we're looking for. >> [applause] >> There's the clock and I hit 3 minutes exactly so I lied.

60:10 >> Yeah. 3 minutes. I love it. It's just going to be a much more competitive arena and going through in the next couple of years. I think I think a lot of people have no idea expecting incumbents to continue to reign but is there's there's a big shakeup coming. Yeah, I mean the DSP bloodbath might have already started Matt. Might might be in the works like as we speak. It's crazy out there. Yeah. Yeah, I agree. I think I think it's a a shakeup for the industry not SSPs.

60:38 Got to go forward to go back. Exactly. Cool. Thank you so much for joining us Joe or Matt. Oh my gosh, sorry. Joe. I look like a Joe. Who is Joe? Where did that even come from? Where did that even come from? Like what is what is my brain doing? Um You were just so heavily invested in the conversation. I I don't even know. Um but thank you so much for coming on. It was really nice to meet you. Yeah, thanks for having me.

61:06 >> Thanks Gary. Yeah. Yeah. See you guys. Yes. Darn, that's the >> Yo, the crowd is back. The The are dim. Front rows buzzing, it's about to begin. [music] Ad Tech, Ad Talk are taking the stage. Adam and Gareth on a knowledge rampage. >> Ad Tech