DoubleVerify Holdings, Inc. ($DV)

We're going to get started. We're going to get started now. I'm Laura Martin, I'm the senior media and big tech analyst over here at Needham & Company. And I'd like to welcome the CEO of Double Verify and his introduction is over here and the CFO of Double Verify. So we're going to turn off Okay. So Mark Zagorski is the CEO of Double Verify where he has spearheaded the company's growth as a critical trust layer in digital advertising since joining in 2020. Veteran of ad tech industry, previously served as the CEO of Telaria, where he managed its merger into the Rubicon project, which was then sold to Magnite prior to Valarie. Mark was the CEO of Exelate, which he sold to Nielsen. And this is Nicola, and he's the CFO. And so we will create some questions for him on the run, which is fantastic, glad he's here. Because there's always 1 I ask him that he hates so we'll do that 1 first. And I would just like to say as an introduction, we rate Double Verify a buy because it's acquisition of Scibids, which it does AI optimization and Rockerbox for attribution are pivoting to Double Verify towards high-value performance workflow optimization with broader ad tech platform economics and deeper integrations into ad campaign executions. We believe that Double Verify will successfully transition from vendor verification middleware to mission-critical AI advertising infrastructure, which will give them upside pricing power and accelerate their revenue growth. So that's why we really like the stock. So I want to start with a leadership question, and you're going to get to answer it too because you're on stage of me. And that is that when we think about generative AI, it threatens to disrupt a lot of jobs. So how does your notion of leadership change in this environment where there's so much uncertainty? What kind of demand does it put on you as a leader that are different now.

Well, that's a great question that you did not send in advance.

Yes, I'm sorry. I apologies.

Answer that no Yes, it's a good one. I think AI has really upended how we look at talent, how we look at growth and innovation. And how we bring those 2 things together in a very different way. I mean, in the past, when we thought about innovation, it was how many engineers can we hire and how great engineers can be hire. Okay. And how best can we get the most out of them -- now it's more focused on how do we take the tools that are out there and apply them in ways that are going to drive growth in our business, but also create efficiencies, which the market is absolutely expecting, right? And from a leadership perspective, what that's done is created massive fear amongst teams. And so the way to frame it less as this is something that's going to take your job as opposed to -- this is something that's going to give you longevity in a space that's going to change rapidly if you embrace it. And I think we've seen a really great embracing of AI within the company. And I'll tell you, when we first started, folks were -- especially an engineering team, probably half of them were like, I'm not doing this, right? I'm a coder. I'm not going to knock code because we told them, hey, by the end of the year, you're going to be a manager. You're going to manage 5 agents that's your job. And like I don't want to be a manager we said, "You're not going to have a job there." They changed that. You can see the people embraced it. And a lot of times, the normal thought was, hey, your senior executives are the ones that are going to be like. Keep that away from me. I know what I'm doing, but it wasn't. It was -- they were the ones who embraced it first because they've seen this change before. Many of them had been through multiple cycles of change, and they're like, if I don't embrace this, it's the end for me. So it's changed to [indiscernible]

[indiscernible] jeff Beso, right before they're going public. And they said, "You've got all these guys that are about to be worth $10 million each. How do you motivate them? And Jeff Besos, because the best employees never work for money. But the point is the best employees don't just work for this job. They work for something bigger, something more interesting, something creative -- and if they don't, then they're not a best employee. What about you in finance, and I'm going to actually add something. So 1 of the most interesting things [indiscernible] is talking about, and I think they're right after you is token cost. He says, "I got everybody in my organization using tokens to try things in HR or in legal is up and exists, I have no return on capital metrics. Like in coding I can put a return on capital metric. But my costs are going through the roof of these tokens and they are telling me in HR, how this is making me the CFO more productive. So talk about how you think leadership, but also this cost notion that maybe tokens need to have a capital spending cycle around it and put budget and legal, you can use this many, so there to prioritize -- but talk to me about that.

So we are actually doing that, which is we are managing the use of token centrally through our technology group. It is spread across the teams in a way that is manageable and that we can monitor. So right now, we're just taking a look at where exactly it's used. Is it using in a way that is efficient, that is actually accelerating our own product road map and our efficiencies rather than just letting everybody trying out and seeing what happens. I think in terms of the responsibility of finance around these tools, there is -- we do need to try them, and we are trying them and the ultimate result of using them will be a higher, more efficient business model, which we are already showing the signs of through our first quarter results of the year. And so there will be an ultimate measure to see that it is actually happening. The way we're rolling them out is in a more centralized fashion.

And I'll just give you the Fubo answer. He said from now on, I don't look at resumes, anybody for interviewing new people. He says, we put them. I expect them to make a video using generative AI tools to make an app and then explain in the video what they did and why they did it and why it's useful. He says because the 2 skills I need or people who could actually use Gen AI, my next employee needs to be able to use Gen AI is less and explain to people why he used it to do that and why it matters in English. And he's like, "I'm back to generalist. I don't want to point solution people. I don't want generalists because I don't know where gen AI is going. I just thought it was an interesting to hire your next guy. Maybe you can impose it on your current guys. I thought it was an interesting idea. Okay. So let's talk about how does DB remain independent while integrating deeper into the walled gardens? Because some of your fastest-growing stuff is sitting within meter sitting within TikTok. And it feels to me like this gravitational pull to the big pools of money that are sitting walled gardens does threaten your independence or trust in your brand as an independent brand.

So our role in the walled gardens is exactly the same, the role as the Open Web, which is exactly the same with [indiscernible] streaming, and it should be the exact same role that we have on the LLMs, which is that as an independent trust layer. We -- just because we're analyzing something behind the wall, it doesn't mean we're not objective in doing so from a way that has only our advertisers desires at the forefront. So that role doesn't change, and I think it's critically important because -- that's why people are hiring us. They are hiring us because they need trust in those platforms. They've seen their ads run against stuff that they don't want it to be, and how look, the reality is how can you trust someone who's selling you media to tell it's good tell you that it's good. CBS, there's a reason why Nielsen exists, right? Because CBS came up with their own ratings and said, "Hey, I get my ratings. -- who would trust them? Like -- so I think -- our role there, and as you noted, it's an increasingly growing role is important. The first -- last quarter, we grew our social activation business by over 90%. We grew a measurement business by 23%. So it's we are leaning in there. There's demand for the solutions. And I think our innovation and our products are just catching up to that demand.

Yes. Okay. Fantastic. Retail Media, CTV, 2 biggest TAM categories for [indiscernible]. How do you prioritize resources and which 1 gets -- and which 1 will be bigger in 3 years from now, do you think -- and this can be for both of you.

Yes. So I mean, we've leaned really into streaming right now into streaming television as the bigger opportunity for this -- and maybe not for the reason why you think, but first of all, streaming would barely scratch the surface on in our opportunity there. We mentioned that our CTV impressions grew 28% last quarter. And we still have strong kind of growth trajectory there. Our products are catching up there. We've launched verified streaming TV, which is a pre-band postbid application that ensures that ads are being run in high-quality full episode players, not like on a mobile gaming site or something like that. It's running in Hulu, it's running in Paramount. We created a product called automated do not airless which give the advertisers the ability to avoid specific programs or specific genres or programs.

Or Gen AI -- or is that a different product?

That's a different product. But this is specifically around streaming. We launched that through Trade Desk. It actually doubled the attach rate on Trade Desk of our ABS product because you buy the do not airless through ABS, which was pretty significant in a quarter in less than 3 months. Now we went -- we're starting from a low rate to a high rate, but it was great. So I think the reason why we're leading stream , we just think there's a lot more of a gap there for what we can delete Or market need to . What we've been able to deliver in the past, what we deliver now and kind of what our attach rate is there. .

Okay. Okay. All right. And Retail Media really does retail media network just have an outcome, so people don't really attach as much the double barrel.

If you look at some of our biggest retail customers or folks like Walmart and Amazon, which are also the total media play and Amazon is interesting. I think last side, I saw something like Amazon is like 80% of the Retail Media Network business.

Yes, that makes sense to me. I would guess higher actually.

Yes. So it's huge, and we work with Amazon as a supply platform as a demand platform through the DSP and as an advertiser. So on all different aspects. So that part of our business is really growing. But when you talk about retail media for us, it's really about Amazon. And we're meeting kind of all of those demands I think the interesting thing about RMs though, which I don't know if people are fully starting to embrace it, an RN is basically just a giant retargeting network, right? You're taking data from a site and then they're using that data to retarget somebody when they leave your site. If you're Best Buy, if your target, that's what it is, right? .

I thought it was -- I thought an RMN was TLM network was sort of I spend an ad dollar and I can actually just have an outcome of an actual physical sale at the other end. And that's what the RMN did is tied the ad dollar I spent to the actual outcome of a sale.

It's part of it, but the money comes from me as Best Buy, me as hard Saying, you came to my site and you looked for a television right? Now Sony, I'm going to sell you that same user someplace else, and I'm going to extend that data of that impression offsite. That's where the money comes from in retail media. The spend of impressions or dollar impressions based on data that I get from that retailer. That's -- so but if you think about that, the buying impressions off-site. Most of that's basically mobile online video or display -- it's not in a walled garden. . That is -- that business is slowly going to be starting to be eaten by the LLM who are going to be focused on commerce, a lot of retail media is moving to the LLMs, so that universe I think is going to start to be cannibalized first by the LLM The media part of that for sure.

Okay. Interesting. So let's integrate you into our play here, Nicola. When you think about like you prioritized by product category? Like how are you guys prioritizing in the back office.

In terms of product development? We're prioritizing about the channels where we're not yet fully penetrate. Would be social and CTV. Those are really the ones where -- we spent a lot of product development resources last year to launch their products that are not in market, 1 that's growing the fastest is social activation are now in the market. And I've already mentioned the 90% that's where we're prioritizing the product development. And of course, in the background, we're also getting ready for a verification on the platforms on the AI platform.

On the AI platform. Is that like open AI? Is that what you mean? Okay. Is Open AI let you in? That sounds like a no.

We can't talk about anything. [Audio Gap] Is that we're in conversations with LLMs around.

Since Calude doesn't have advertising. So narrows them down. We're not letting Google.

I mean our perspective is this. When we talk to advertisers, they expect us to verify media transactions everywhere they spend. And it started off in the open web and then moved programmatic and then to social and then to streaming TV. The next phase will be LLMs. They've made it very clear to us and to those LLMs for them to move beyond test campaign budgets. They need verification. When Reddit went public, what did they say? We want third-party verification. When Netflix started selling ads, what did they say? We're going to have third-party verification and measurement and so they called us. They called Nielsen, it's just history repeating itself that.

So long as you have brand advertisers, I do not think you need it if it's SMBs, because they have clear performance or outcome orientation.

Look, I think if you look at every platform has some percentage of brand advertisers right? So even the social platforms have 30%, 40% on branded advertiser.

Because I would because they do all SMBs, they have 10 million SMBs, but ...

Right. But they have 30% of their billions in revenue. I mean you look at someone like P&G. P&G spends $1 billion a year on YouTube, billion just on 1 platform. .

That is crazy, and that's just P&G.

Yes, on -- so you have to assume that there's going to be a -- and if you go to some of the LLMs now look at the advertising, there's big brands there. We've seen brands there.

Well, an Open AI doesn't have the capability of doing granularity. They have to do big brands. It's almost like a billboard sale right?

They made recent deals with demand generation companies right? Cargo impact you video because they're trying to attract dollars from brands who spend across their platform you'll have a sales infrastructure. So we -- and we know we have dozens of our current advertiser brands are testing campaigns across the LLMs. So we just feel it's inevitable at some level that verification will come there driven by advertiser demand.

And by the way, Claude is saying, "Hell, no." And I'm like, "I've heard this with Netflix, like it's not a thing. Like ultimately, you guys need money. And advertising is -- targets a whole part of the U.S. that likes less than $70,000 a year. That's where advertiser targets. You just can't ask people to pay a fortune. That's the top half of the -- so you need to reach everybody and that means you need advertising so initially....

Especially outside the U.s. remember, Netflix hit the wall is when subscribers started to slow in the U.S. and they moved outside the U.S., and you can't charge somebody in developing countries, Southeast as are $24 and spend $20 billion on content a year.

All of their new ads are here or ad-driven because people have 7 streaming services, and they want to pay $10 for Netflix $24. So it's not only offshore, but I take your point about the slowing.

As you expanded to markets where you can't pay $19.95 a month. .

You must have an ad-driven option. Yes. Well, and also, to me, it's more fair because if you have a hit piece of content, why are you paying a fixed price to something crappy like advertising allows you to make a little more money when you have stranger things and it blows everybody away or some outbreak out it. It allows -- and same with sports, like they now have 5 sports, in this case, 5 different live sports for the NFL. If you have programmatic, if you have advertising, you get to capture the upside of it happens to be a great game, right? Whereas if you just have a flat subscription fee, like it's like you got -- I mean the consumer got much value in that particular in a sense. So I like the combination of the 2, like a downside protection layer of subscription and then an upside warrants from advertising revenue. If you do a really great job at your core business, which is making content, so but anyway okay, great. And let's go to social and do activation and measurement products on Meditec talk YouTube Snap. They continue to grow at double-digit rate and does growth slow -- are we in 1 of these periods. Sometimes you guys have this fabulous first year or a fabulous year because you have a new product, but then it sort of goes like normalized growth rates of sort of single digits. Do you see that happening in social.

I think we have a while to go before we see that kind of pacing. We're -- we've just launched our social activation tools on meta early last year. And to be honest with you, like the V1 of that was not even close to where we are today. So the cool as advanced. You've seen an acceleration in revenue on social activation in Meta. We just enhanced our TikTok activation tool that's gotten better.

And when you explain to the audience activation versus measurement, how you distinguish?

Yes. So think of activations, what we call Prebid. So it's actually filtering out a violation before you even buy it before you bid on it. And measurement is what we call post bid. So I've already bought an impression in some places, but we have the ability to block the actual rendering of that. Obviously, advertisers would prefer not to buy than they would to try to get a make good and get money back. And get money back -- so activation is pretty new for us on social. We just started kind of scaling it and launching it over the last several quarters. But what we found, which is what we saw on our Open Web, our activation business in Open Web is significantly larger than our measurement business. .

Makes sense, because they want that add to run next inappropriate content.

So right now, our activation business on social is smaller than our measure. Our measurement business is half social, half open web right now. And that was -- which was that last quarter? $60 million . Yes, $60 million in measurement, right? But our activation business is considerably smaller than that. So we see acquisition continuing to grow on the social side to where a ratio that we see in the open web is similar, whereas almost twice as much revenue from activation as we do from.

So 1 of the things I remember when you guys did the Prebid product, I was very winabout why can't we raise our price from $0.08, and you said, "Well, the good news is in order to buy a prebid product, we require you to a postbid product, right? So you're calling them activation and measurement. But -- do you do that in social to? They can't buy the prebid product without buying the postbid, so it sort of doubles your take rate in a sense because you're right.

[indiscernible] work together. We catch violations in measurement and then we filter them out in prebid. And if they have just a postbid, let's call it, measurement, which is your word, -- what is the benefit they get? Like on average, how much does it save them? How much does this measurement number go down for the fraud and when you would add a pretty bid to it? So we see anywhere from kind of like 3% to 6% or 7% percentage point increase in suitability, for example, by using a pre-bid filter. So if you were at 90%. I'm SP1.

You think you started in '19, it goes to like 96... .

But think about that -- it's fair running a campaign that has 100 million impressions. Yes, 6 percentage points is a lot of impressions that you are keeping away from start. It's incredibly valuable. And we see those rates going up and getting the product getting better over time. The product learns, right? .

So we're getting all this crop like -- let's call it AI slot by the way, I use that term because you guys called a product, AI slot on the data panel this morning like, we should have data slop, because there's too much data, data is confusing it. At some point, it's doesn't add value. I'm like, well, let's take AI op and apply it here to data slope it's just a good it's a great title. You guys made a product out AI. .

Shop stopper, a .

Slop stopper, I love it. But on this point, specifically going from the 90% to 96%, when you have a slop stopper, when you have more slap, could this 90% be structurally under see going towards 80% because generative is going to make so much more content that's just bad, not viewable or -- it's a different.

That's -- I mean, look, the volume of challenging content continues to grow, right? But our filters are still our filters, right? You're just jamming more into them. They're still identifying everything, right? So volume is never our product problem. Sophistication, for example, around AI, cyber fraud, which is becoming mortality and we saw 140% increase year-over-year in fraud variance between Q1 of '25 and Q1 to '26. Think about that, 140%.

And what does that mean?

That means there are more fraudulent attacks using AI. These are like bot attacks, things like that, that we've seen year-over-year. It's becoming more sophisticated. And this is where like there's good AI and bad right? And bad AI is being used to impressions spoof advertisers feel dollars. That's beyond that's beyond quality and suitability it's actually fraud. True fraud.

Like true fraud. Yes. Okay. Is it coming from offshore? Is it state actors from offshore, which is what I assume -- so this is bad actors in America .

This is not a kid in his basement. These are sophisticated. No, because people somebody think like, there was 1 this is like sophisticaed.

Some 15 year-old that's taking time off from his lunch break to like do fraud....

That's always that. But like these are sophisticated criminal enterprises, they spin up literally thousands of servers -- like this costs money to do -- so they spend money to steal money. And this is a multibillion dollar industry, billions of dollars get lost every year.

It's a trillion dollars, that's really great. Okay. So we're going to continue to -- the answer was we got to continue to grow at double-digit rates, and we're going to continue to have 1 for monetization. Okay. So Nicola starting with you, then we'll do you a second. How are you using AI today to lower costs and how much AI-driven productivity gains will be reinvested versus being allowed to expand your margins.

So how we're using AI to lower the cost on the efficiency of what we do in terms of how we classify content, AI is allowing us to do it faster and for more content at a more efficient and that's part of the way we've maintained gross margins over 80%. So that's on the top of the funnel in terms of what we classify. Below that, we're able to -- what we said at the beginning of the year, we'll be able to grow with fewer employees and that we already started to see it at the beginning of the year, and we'll end up seeing at the end of the year. And the ultimate result is that our margins are growing. So in Q1, we had 31% margin. First time we had a margin over 30% in the first quarter. So we're seeing the results at the bottom. We're obviously being responsible on how we're doing it because people still need to trust us. So we're using the AI tools in the right way to classify. But the efficiencies are great they're large. Up until now, we've always said we're going to continue to invest and reinvest in the business. That's not changed for us, except that the way we're able to do that creates a lot of efficiencies, which is why we're able to expand margins.

Okay. I'm glad you're managing tokens. I think a lot of CFOs are not doing that yet -- so you're really ahead of the game on the token cost. What are you seeing, Mark, on maybe on the product development side in terms of -- by the way, I will say there was a time there where you guys were super innovative on product and then like everybody went to sleep for like 3 years. And now it feels like you're back in the new product introduction game. That's how it feels to me from the outside. I don't know if you see I think -- you know what it was. .

We -- the innovation was done incrementally, and we are doing things around the edges as opposed to big swipes, big hits. And I think that's where we've moved to now, which is like taking on big challenges like streaming TV and transparency around streaming TV like Gen AI swap like social and the opportunities within social. So I think we mentioned last year, at the end of last year call, like we launched more products in '25 like in 2 quarters than we had launched in 2 years before that. And I think it's -- part of it has to do.

Why would you ever work on incremental if you can work on a big new category.

I think -- we still saw a big pool of money there that we could go after, just tweaking things and investing small. That wasn't like we weren't doing anything. We were building the base layer from which we could then launch deeper into social, deeper into streaming. But I do say, over the last several quarters, AI tools have allowed us to do that so much faster, right? Nicola mentioned classification. We can classify content 2,000x faster now. 2,000x faster.

It feels like a cost thing, not a new product thing.

Its cost, it allows us to expand to new languages faster into new markets faste, so we employ semantic scientists. We employed linguist translators. All of those roles are now being done by AI, right? I don't need to have someone translate text and then and label it and then put it into a model. I had AI doing that. We're going to be down 100 contractors this year. by just using AI to do labeling as opposed to have a human student labeling and feeding our models. So it's driving efficiency. It allows us to move into markets faster. It allows us to test products faster. So we're now building natural language interfaces for customer service, right. In the past we've launched that on a normal pace where it'd be like, okay, let's do a beta, which will take us 6 months then get it to a [indiscernible.

Why do it yourself? Aren't there guys off the shelf that you can just use theirs after tomorrow.

Well, I'm saying that's the old No, because we can use AI to actually build prototypes very quickly and lost. And we don't use off-the-shelf stuff because when you rely on third parties, that's not your own product, then you're tied to them forever. We build our own stuff.

But it takes longer to get to market.

Not anymore.

Not anymore. Okay. Fair enough. Okay. So I have 1 CEO that is saying -- and maybe you don't care because you guys are sort of hedged, but he's saying that programmatic rails are going to be replaced by agenetic rails. And he says, over time about what he means is like 3 years like a time isn't like a 10-year idea. Do you guys agree with that in your neutral position you sit at?

No matter how someone buys a less media we're kind of indifferent as long as our solution set is there as long as we're verifying that transaction, we're kind of indifferent to how they buy or sell. In this case, though, what's really interesting is Think about how folks used to buy with a piece of paper.

That's true a lot of them are still but that way. It's great and then they bought exertion order, which looks like a trading ticket like this and you had with the video to company you want to run the ad. Right. And then it moved to programmatic, right? Now programmatic at everything and like 85% of display and 85% of mobile is programmatically bought. It will become agents, right? Our role is still going to be the same. And the interesting thing about agentic buying, which if you thought programmatic is going to be fraught with challenges and opacity, think of an agentic world where I'm like -- because everyone is like, I think, you don't need a DSP anymore. You're just going to send your agent out to buy media, right? And then that fit, I'm going to send my age out, and we're going to negotiate those agents are going to negotiate. I guess what's going to happen. That universe is going to be flooded with fake agents, flooded that fraud we're talking about. There's some kid in his basement right now creating an agent -- that agent is going to sell media. And it's going to look just like a Netflix agent, and it's going to say, hey, proctoring at $30 CPM Yes, don't you want to buy this, and they're going to make this looks legitimate, but how can I tell because we're going to be able to tell. And more broad there is a better will be able to tell because we've already built a tool called Agent ID, which is a first iteration of kind of giving advertisers the ability to understand what the role is of that agent and who they there at is interacting. But the next stage will be when agentic buying truly scales, is playing that same role. Is this person really selling you a safe page -- is this personal page at all. Page right -- is it safe -- is it able -- is it viewable? All of those things. We're just going to be employing. And so we joined ADCP which is the Ad standard standardization group. We're involved with the IAB standardization group. But we'll be agnostic. Whatever protocol someone uses, we'll build for it. And -- but we're going to play the same role, which is, hey, need to stop here and we're going to check it first. So 1 of the big -- and I really do want to -- is trust. Like is this trust in an AI world become more valuable.

I mean, I think it becomes as valuable as -- when you continue -- we're continuing to move people out of the equation, right? That's true just said you went from a person you trusted sitting across the table from you to then a machine to then buying through a programmatic platform where you had no idea what was happening -- so transparency is always going to be an issue as you bring more machines to it, efficiencies there. But who can you trust? And I think that's the role we play, which is driving trust and transparency in opaque environments. And I'll give you a good example of where this so think of meta, we started talking about social Yes, we did Meta has an amazing product called Advantage Plus which is a black box solution that you put your money into, and it guarantees you an outcome, right? You don't know where it runs. You don't know how it runs. You know how the context of runs, but they'll say, "Look, we'll get you this many clicks. We'll get you as many conversions. And it's doing quite well. However, there's no transparency in it, right. when we see campaigns running on Meta using DV measurement verification, the attach rate is twice as much on advantage glass campaigns as it is on non-advantaged plus can -- that means when you're buying into a black box, people are relying us on twice as many -- the attach rates twice as high because they want that transparent. So the more things move to black boxes and just trust me outcomes, the more they need transparency and they lean on companies like DB.

Okay, right. And the more fraud there is, the better it is for you because you're like an insurance policy, like a low-cost insurance policy. yes. Okay. One of the things I really liked as I've told you a million times is this move towards percent of take rate percent of like a take rate that's a percent of media. So let's talk -- and I don't think you guys talk about this as much, -- but talk about like because of Siris and because of some of these new products you're introducing, you're sort of getting, I would call it a dual revenue stream where you have, let me call it a subscription fee, which was -- it's really a fixed fee, it's a subscription. It's a fixed fee round number is like $0.08 or $0.14 per 1,000 impressions. Now you're going to a percent of money spent, which I think has more upside. But can you talk about the dual revenue streams and how you think about those?

Yes. So on the optimization products, it makes plenty of sense to be on a percent of media, right? -- it makes a lot of sense because you're showing an immediate return on your investment and what we're able to optimize for the media part of the business. That is the 1 that we are pricing directly on a percent of media. The rest of the business is still, as you said, a fee, we receive a fee per impression that we're measuring and verifying for the advertisers. That is the model that we like for where we are today, which is we are still penetrating new verticals. We're still penetrating new platforms where -- for us, it's very important to be very simple for the advertiser. So rather than having a negotiation every time there is a new product that's available in very simple fee per impression and just tack on more and more products to the impression. It creates a very simple upsell motion for the client with our new products. .

I think that we... Is that different from a 10% take rate 2% take rate, just as simple ...

Because every platform is not the same. -- it will be not everybody is thinking of what they're spending on social the same where they're spending on the DSPs. It's generally not necessarily even the same budget. And so the conversations are different. And it's easy to just say, this is how much we are charging your impression that we see per channel. And it will vary by client. It just makes it very fluid and very simple -- once we are -- I think once we end up in a situation where we are upsold into all the channels, I think then we'll have a better opportunity to differentiate the price based on the CPM.

The debacle in this was the $0.08 when you went to CTV, you never were able to raise the price when you went from a $2 CPM to -- so you were stuck at the $0.08, so -- which is 1 of my big complaints because you just walk away from all this excess value. And you've never been able to raise the price. You keep said, oh, we're going to do it, we're going to do it. No, -- we -- so as -- part of that was the value prop on CTV versus other video versus social. I think we are now leaning into having that value prop that gives us greater ability to raise prices. We have raised price on video versus display. We think there's the trifurcation opportunity as we've launched new tools like very streaming TV like the do not Air list. So there is an opportunity to increase price there for sure. I agree with you. This simplicity costs you 10x. That wassnt the best economic decision.

It is -- it's not just simplicity, but it's actually the attach rates and the value of the product that we wanted to go for, right? Because most CTV is now -- is bought by PMP or PG.

Yes, which is a private marketplace or programmatic guranteed.

And there's a little bit our value prop there is a little bit less defined when you know I'm directly buying an impression from Hulu right? But when as more CTV inventory comes into the fold, as it becomes more opaque as these black box solutions come into play, our value starts to increase, which means we have more room to start to raise prices there as well. .

Okay. questions from the audience. Any questions for Mark or Nicola? Okay, from the audience. Okay. Great. Okay. Okay. When you think about staying on this issue of rev share, when you think about share of revenue versus fixed fee 3 years from now, this is a growing segment, I assume what we're doing over at Scibids. So how big do you think that could be in 3 years? .

Well, if you think about where our target for Scibids by '28, '29, it was like about $100 million in revenue. $100 million out of $1 billion by that time. So 10% or so. Now we have a new catalyst around that which is the product we launched kind of late last year, which was authentic advantage for YouTube. That's a combination of a prebid filter, a postbid measurement and then Scibids as a sandwich in the middle. We mentioned on the call, that's growing really well. It's a $10 million-plus run rate right out of the gate. We see that now we're expanding that to TikTok and to meta as well. So that could also positively influence how fast that -- because that's a percentage of media. There's a piece of percentage of mediation that as well. That will positively influence what percentage of our business goes there. And as you will like Yes, YouTube is video. It is CPMs are higher there. They're not CTV, the compressions but they're higher, so we have a chance to get a bigger chunk of that than just our straight pre-bid costs of those.

And what I would say is the goals we have there are stated is 50% of our business being social, CTV and then AI when it comes. Right now, it's less than 30%, whether it comes because of a percent of of media or not, the goal is to penetrate the channels where we just launched our product. So there will be a motion there. whether it's held by percent of media or not, I think we're more set on just making sure we are where the advertisers are advertising. And now that we have the products, we're on our way to reaching that goal of 50% being social CTV AI. That's more of an operational goal that we have in terms of diversifying our business and going where the dollars are now that we have the product.

I see. Yes. I really love all the new product innovation. I mean it must be more interesting for you guys, too.

It absolutely. I mean, look, we -- as Nicola is like we need to go where the advertising dollars are going, and that's where they're going, those platforms.

Well, and on that, just listening, like one of the things somebody said on stage is the problem Trade Desk is they stop listening to their -- and I think when you listen to your customers -- where advertising dollars is going is code for we're listening to our customers.

And we have to -- like we don't need to be the first ones there because there was always -- remember Clubhouse a years ago, remember... I don't. It was like a collaborative audio thing. Everyone jumped into this could be the next big thing. And we're like, should we be building for that? I like let's hold off a little bit and then no real advertising dollars followed. So we follow the ad dollars. Like if they're like, hey, we're going into the LLM, and we will be in the LLM.

Well, that makes sense -- I mean that makes sense.

That's how we make money, right?

Yes. That makes sense to me. So measured transaction fees have been falling. And is this pricing compression structural, consumer-specific, competitive or tied to discounts for new clients only? Why has the average price...

So this is a story of mix. We're going in categories where social -- the price for the social product is not yet the price that we have for mobile and online and display and video because we're just launching the product. The price on CTV is still the same as video because we're just launching the product, and we are not yet at a point where we can charge a premium for CTV versus a regular video. So what you're seeing on the MTF is really just the mix of the business. As we grow, as we penetrate, especially on activation, you should see an impact positively on MTF. I think the number that we keep a close eye on is revenue per client, right? So the top 100...

Revenue per client or top 100. That was $2.6 million in '22, it's $4.5 million in '25.

Scaling. It's not -- the fee is an output of that, which is we're scaling the opportunity with the clients wherever they're spending their...

Got you. So you're doubling your average spending per client, but you may be tripling their impression growth. So it looks like a lower impact... Okay. That makes sense. Okay. I'm going to call it there because we're out of time. Thank you so much.

Thanks, Laura.