Hi, everyone. Good to see everyone. If everyone could mute themselves if they are on the Zoom. Thanks for joining. Just wanted to welcome everyone. I am Jill Hall, Head of U.S. Small and Mid-Cap Strategy with BofA Global Research. We are very fortunate to have a great two days of executive insights, fireside chats with almost 20 small and mid-cap corporates. It is a great annual event we have every year, and our analysts have a really great breadth of coverage in the small and mid-cap space. They cover about 1,000 small and mid-cap U.S. companies. Excited to continue to hear from them. Feel free to reach out to me or to corporate access if you need the schedule, if you still want to join any of the sessions that you are not already signed up for.
I have had some people reach out to me today, so not too late. Or if we can help you getting in touch with any of the analysts for any follow-up, or if you would like to sign up for any of the research on the companies today, as well as our small and mid-cap research, or we also put out a daily compilation on some of the small and mid-cap research from our analysts. With that, thank you for joining. Hope you are able to join some other sessions as well, and I will pass it over to Arthur to do some introductions.
Thank you, Jill. Good afternoon, everyone. My name is Arthur Chu, and I am on the U.S. Internet Team here at BofA. I work with Omar Dessouky, who leads the video game and advertising technology coverage. The ad tech stocks that we cover are Magnite, AppLovin, and Unity. We are going to have some time for a live Q&A later, but if you have any questions in the meantime you would like to ask, feel free to email me at arthur.chu@bofa.com, or you can just use the raise hand function on the Zoom call anytime that you have a question. It is a great pleasure today to have Nick Kormeluk, Head of Investor Relations at Magnite, with us. Magnite is a leader in programmatic advertising on the supply side, particularly on CTV. We are buy-rated on the stock.
We think it is well-positioned to be one of the biggest beneficiaries from the CTV industry's transition to programmatic advertising. Welcome, Nick.
Thank you, Arthur, and thank you, Jill Hall, for having me today.
Yeah. Nick Kormeluk, we have some generalist investors in the audience today. Maybe just for people less familiar with Magnite or supply-side platforms in general, can you introduce Magnite to the audience?
Sure. Magnite is what's traditionally labeled as a supply platform, right? Effectively, where we initially kind of enter into the industry is really working for publishers and working for publishers that have inventory across the open internet, connected TV. Think of any digital ads that are serviced. Those are ads that we are in market trying to bring demand to, right? We operate a marketplace.
Yeah.
We bring in demand to match up to be able to get inventory sold for publishers. But our strategic value is really making sure that we can bring it to as many buyers as possible, make a very active bidding on that inventory. A lot of that's tied to bringing in and importing data in to allow data to find users that buyers and brands are specifically looking for. In doing so, that's the value that we inject into the market, and the value that we drive for publishers. Whether it's in connected TV, streaming, broadcast TV OEMs, or whether it's mobile app or mobile web, or even web browser or digital out of home, we service all markets as an omni-channel player.
I think the area that is most exciting that has a much different market setup than historically the web has is connected TV, where the amount of inventory is really concentrated in the hands of 30 large global partners that represent about 80% of the world's inventory. It seems like every week that goes by, there is less of those people because they are consolidating, like Roku, Fox, for example. But we have relationships with all of them, with the exception of YouTube. So we have become the de facto place that anybody that wants to buy connected TV comes to to access the CTV market, which is a far cry of where we were a couple of years ago.
Got it. So you guys definitely have a very diversified sort of business. I think if you look at the ad tech industry probably three to five years ago, I would say the industry has obviously changed a lot since then. So has Magnite. I think you guys have transformed yourself from a traditional sort of open web SSP to now, I think you are more of a programmatic CTV ad leader. Maybe just tell us a little bit more about that evolution.
Yeah, I think the beginnings of CTV, and we made a few acquisitions to get into CTV, but I think it is the nature of relationship of many partners that view themselves as walled gardens have brought us in as a trusted technology partner, right? So I think that is much different than simply looking at a supply-side platform and saying, "Hey, we will invite you guys in. Throw a bid in. If you guys win, you will get paid for it. If you guys don't win, then go away and try again next time," right? So I think the strategic nature of the relationship that has evolved comes from initially working with a Disney or a Roku or others in our early days, call it three to five years ago, and showing that we have been brought inside.
They want to operate and sell their inventory, but they don't want to simply allow everybody to bid on their inventory. They want their data protected. They don't want anybody to see their user IDs leak out to buyers, but they want to monetize better than they can on their own, either through a direct sales force and insertion orders, or doing something self-service on their website. They want to capture programmatic demand from all sources, and their aspirations to grow revenue more has had them lean on us to be the one party that is invited in to be their tech partner, which is why we define the market as almost a winner-take-most. You may have a couple of partners that you use, but for the vast majority.
W hether it is ad serving, whether it is mediation, whether it is demand generation, whether it is yield management, whether it is identity strategies or audience creation, you do not choose two partners for that. That is where we have won, and I think what is happening is our evolution over time that people believed or investors may have believed that there was risk to when this industry grew up, right? Connected TV grew up. There was always the risk or the fear that people would either, A, do this themselves, or B, go to a much, much larger player. Now, there is not a larger player in the space. Our market share has expanded dramatically. Our position in programmatic is large, and we have expanded our Disney relationship. We have expanded a Fox relationship. We have dramatically expanded Roku.
We have won Netflix during that particular time. So we have shown that we are that partner that can execute and bring people to the programmatic market. The very interesting component to that kind of rewinding to the past was it has also had a halo effect on the DV+ market. The reason that I say that is you now have people on the DV+ market that looked at how we have partnered on the connected TV side, that look at DV+ and have never relied on programmatic demand. They have said, "Hey, can we have that type of relationship with you?" In the past, in open web and in what we call DV+, which is everything excluding CTV, we would be one of a bidder, and then it would not be exclusive, and you would have five or six other guys also trying to sell and rep that inventory.
What has happened by that halo effect that I am describing is people have said, "We want the same kind of relationship with you that Netflix has. We want you to protect our data, you be the one that connects us to demand. We do not want to have five SSPs. We want to hire you exclusively because we trust you, and we think you are best positioned to monetize and bring in all the demand, and we do not need two people to do the same thing. That inserts risks, that inserts confusion, so we want you to be the partner." Our list of commerce media partners has exploded from a United Airlines to a Pinterest to a Best Buy to a Redfin to a REMAX to an Expedia to a PayPal.
All of those things have exploded recently to now 21 partners that are now relying on us exclusively to be their programmatic partner, bringing in new demand once they have decided that that is a path or a source of revenue that they would like to be able to tap into.
Got it. It sounds like, in the CTV world, the supply side is much more consolidated because a lot of inventory are either with these premium streaming partners or companies like Netflix. You mentioned some of the key partnerships like Disney, I think you mentioned Roku, Netflix. I think for the latest, you guys also added Samsung as a key partner. Samsung is adopting SpringServe, which is your ad server for placing programmatic ads on the home screen. I think you talked about retail media, and I think Walmart recently also took you guys as the sell-side partner. It really seems like Magnite has become the go-to for these big brands when they're looking for CTV opportunities. Maybe tell us a little bit more about these recent announcements. What are you hearing from partners that they're looking for that perhaps they can't get elsewhere? What makes Magnite offering unique?
Sure. No, absolutely. Let me do it maybe reverse order, since we were talking a little bit about commerce media. Let me tackle Walmart.
Yeah.
Vizio first. They've been a longstanding partner with Vizio, and we've been their ad server, we've been their primary SSP selling inventory for them, and that's been a phenomenal relationship, and they've been one of our strong customers. What's changed recently is they in the past used a white label DSP that they used from The Trade Desk, right? They kind of parked all their data in that DSP so that you had to buy through that path essentially to be able to access Walmart shopper data on all their Vizio inventory. They kind of broke that data lock a while ago, back in April, and said, "Hey, now you can get that same Walmart data on Vizio inventory from a handful of different parties." Basically meaning it's not just going to be centered in one DSP. Then it turns around towards Cannes, which just happened in June.
They announced that they are buying Vibe. Clearly they bought a DSP. They are focused on having and owning their own DSP versus using a white label DSP. They tapped into us to say, "We are going to use Magnite to be able to take our user data and be able to take that not only on their own inventory and bring in new DSPs like Yahoo," that was announced at the time, "but also be able to use our data across being a DSP that could buy on other sources of inventory, open web inventory, non-owned and operated inventory from Vizio." That is a very, very big expansion of what they would like to do in the market, none of which is running through our numbers as of Q2.
Again, that is a future opportunity that we think has a lot of promise. Clearly, I think they have got aspirations to want to do something similar to what Amazon has done and to stand up and buy a DSP, one that is very tightly integrated with what we are doing with them is a very, very good guide to our long-term prospects of continuing to grow with Walmart Connect. Samsung was another one that you referenced. I think because Samsung was a prior customer, I think to a large degree, the Street really is like, "Oh, that is great. You got some home screen inventory from Samsung from an ad server perspective." Just to clarify, historically, we have done no ad serving with Samsung. They use Publica, which is private equity-owned, part of Integral Ad Science, and it is kind of the only account that Publicis has served recently.
They went to market to bring in somebody to do ad serving for their home screen. Their home screen historically was sold through a direct sales force through insertion orders, and it was ads that they actually ran without having to tap into an ad server to do so. They went to market, ran an RFP. We won that business for their home screen. You say, "Hey, great. That is fantastic." You get a little bit of fees for doing that. Samsung has the largest amount of television, smart TVs installed globally at hundreds of millions of televisions. Some numbers say 100, some say 300. It is their number to post and publish, so you can publish the number that you find from whatever source that you have. They are the largest install base of televisions globally, smart TVs globally.
For other TV OEMs, what we have seen is the home screen can represent up to 30% of their ad revenue. There is a very established channel of buyers who are buying home screen inventory, and a track record of that inventory performing from Samsung being the publisher of that inventory. The fact that we are now the ad server and we are able to sell that inventory as their primary SSP to do so opens up a tremendous amount of additional growth within Samsung as an account for us that again had no revenue contributing from ad serving and SSP in their home screen in Q2. That is another future growth opportunity that is pretty sizable that I think is probably bigger than what the Street has recognized or realized when the announcement hit because folks recognized that we had a relationship with them already.
Got it. That is super helpful. I think, Nick, you mentioned something that is really interesting, which is platforms like Magnite, they allow the data on users to be activated across a variety of different platforms for the buy side. I think you are talking about audience curation. Historically, I think audience curation is a value, sort of a creative component that used to happen primarily on the buy side, right? It contributes significantly to DSP take rates, I think north of 10%. But increasingly we are seeing these audience activation taking place on the supply side. Maybe can you talk a little bit about some of the drivers behind that shift? Why is sort of audience creation moving from the demand side to the supply side? Secondarily, how accretive do you think this could be for SSP take rate?
Yeah, no, it is a great observation, and you are 100% correct. A lot of it has to do with what we talked about earlier is market structure, right? In open internet, when you have millions of websites and you have millions of apps, it is really, really hard to scale if you are a large marketer like a P&G. It is really, really hard to find scale within one publisher, right? Because the market is so broad and so diverse. You really looked to DSPs to be able to find signal across thousands of different publishers and find the users that you wanted to be able to effectively run your campaign. You kind of leaned in there. Publishers were more open to sharing their user IDs on the other side and embracing different identifiers so that they could get access to that demand.
If they had their preference, they would never share their IDs because they believe that that data leakage and loss leads to somebody else being able to buy their user without them having to come back to the same website over and over, whether it is an ESPN, or it is a Disney website, or a New York Times, or a Wall Street Journal, you name it. If you learn that ID, you can find Arthur in five other places, maybe cheaper than The Wall Street Journal, and they have the ability to now target you outside of just their website. In CTV, because the market is so concentrated, that fear that they have in open internet, now they can mandate that that ID does not pass to a user.
They have the control, and they are controlling how data is used, and they will import identifiers into their walled gardens, but the match is happening inside, and then they report out that, yes, the users were found. Because of where the IDs sit and where the matches take place in CTV, DSPs do not have the same ability to charge for datasets and build datasets to identify users. They still have their IDs of who their buyers are looking for that they can pass, but they are not being sought out after for those datasets. The other element that exists there is you have buyers that have decided they want control of what inventory they are buying.
In DSPs, when you would plug in what you're looking for, you might not always know on the other end of that black box where your inventory or what the preferred inventory sources were. Buyers have wanted more control. Agencies have played a role in becoming data providers and even bought data assets like a LiveRamp and market out there. You've seen them partner with other data vendors and providers. As agencies have tried to play a bigger role in data, that's also moved it in that direction, and I think they've started to curate and sought to connect with us to be able to then say, "Hey, we want to on the other end of that, we want to make sure we have which data sources, which publishers we are seeking.
We want to make sure that those paths are visible, seen, and we can buy across what we would like," and more importantly, what they would not like, right? In that seek for control, that's also moved DV+ to have more creation by the agencies playing an increased roles and brands playing an increased role in doing so. For us, there's two parts of our data business that I would reference. If we're using somebody's first-party data, which we think is the most valuable data, meaning a publisher knows most about their users. They have credit card info, they have addresses, they have emails, they know shopping behaviors, they know intent, they know viewing patterns. They know a lot about them. That first-party data is extremely valuable. We do not charge that first-party data. We don't charge a fee for that first-party data. There is a benefit.
The better we are at clearing that signal and making it and matching it up with buyer demand, you get a higher CPM for the publisher. Our incentive is to drive higher CPMs to have better targetable inventory to be able to bounce up against. Our take rate then, for example, if you had a Pinterest ad at $4, on top of that, if we could refine and have a very clear signal of who that viewer is and who that Pinterest user is, and we could sell that impression for now $6 because a buyer, Restoration Hardware, really wants it. For that increase, and I'm just making it up, if our take rate was 10%, now we would capture $0.60 as our revenue per impression as opposed to $0.40 if it was without their data signal coming through.
We get a nice lift from selling the inventory for a higher price, and the publisher gains most of those economics by carrying, again, the higher CPM lift that they're getting for us doing so. What we're also doing is allowing data vendors, whether it's a United Airlines, or whether it's a Best Buy or a REMAX or Expedia, we're also helping them monetize their data off O&O, off their owned and operated inventory. In doing so, LG is a perfect example. LG has ACR, automatic content recognition data, from all the information and all the video that's seen through that glass, they capture that data. We help them sell that data to non-LG inventory.
What that does is you might find a category that you are looking for or a viewer that you are looking for something that you would like to market in open internet, on another streaming channel, on another TV device. We will sell their data, and in that case, we will actually get a rev share for them for reselling their data and driving and contributing new revenue to them that they ordinarily wouldn't realize. Because we are as broad as we are, as omnichannel as we are, we are the best positioned to be a reseller or data broker for that data elsewhere and share in those economics with them.
Got it, understood. Nick, if we think about the traditional sort of DSP pricing model, out of, let's say, take rate of 20%, maybe 10% of that is data bundling and all these audience activation. Are we talking about a pricing lift or take rate lift in sort of the same similar magnitude, or how should we think about that?
Yeah, we haven't quantified it particularly, so it's really hard to understand and try to tease out what CPM lift is with that higher take rate. In that case, it wouldn't be any higher take rate for us, it's just CPM lift. That CPM lift comes through-.
Right.
And zero difference to the take rate. We are selling that inventory off property. In that, you could have very favorable rich economics, a 50/50 split for reselling their data in other locations and other places. That's something that's becoming a very nice contribution to overall revenue. It's not massive, so by no means is it anywhere close to half of our overall take rate in revenue. But it's definitely been a very nice addition, and it's something that has other partners seeking us to be a data broker and vendor on their behalf because it is revenue that comes to them as a new source from nothing.
Sure, got it, understood. Maybe just switch gear a little bit to AI, which is obviously a very highly followed topic. I recall back at the BofA Tech Conference in June, we talked about this new agentic ad buying path that you guys were just rolling out. I think you guys brand it as Magnite Orchestration. So maybe just to start for investors who aren't as familiar with agentic development in ad tech, what specific problem does it aim to solve? Why is Magnite uniquely positioned to solve this problem?
Yeah. Embedded in that kind of series of lots of questions, we probably have a 2-hour webcast that we could go into. Let me try and stay at a level, make the key points that I need to, but not get too far into the weeds. We play agentically, and we made some announcements back in April with real partners that when they signal that they're a partner of yours, they get a lot of customers hitting them consistently and investors asking, "Where are you in your journey?" Because you're willing to put your name in a press release. We announced capabilities across mediation, agentic capabilities across seller agents and across buyer agents. Then we follow that up in, call it the June timeframe, right around your conference, with an orchestration layer on top of that.
Broadly speaking, even if it's not and again, we think we're positioned really well to have our buyer agent, seller agents, and mediation agents used. But even if we're not used in those cases and it's a transaction where somebody's buyer agent's talking to a seller agent, we are the ones that are sought out after in order to be the infrastructure layer on which to monetize. You still need things like privacy protection. You can't just share an ID and publishers can violate privacy risk, and they're the ones that have to pay fines across the globe for doing so. Those agents don't know how the money's going to transact and who's going to pay what funds through what workflow, and then who's going to end up with the money as a publisher at the end of the day.
There's a lot of infrastructure, even if you make the assumption that we are not involved in a buyer or seller agent. That being said, we think we have a lot of traction, and we've announced a lot of partners from an agentic capability. The mediation part of it, the ad-serving component from a agentic workflow makes a lot of sense, right? That's just optimizing inventory and ad units to perform the best that they possibly can with demand, different criteria, different workflows, speed to setting those campaigns up, et cetera. That's something that makes a lot of sense for what we do daily because that's in our purview and that's our general workflow. The other side is the seller agent and the buyer agent. Because we are the end-to-end broker to get somebody desiring to buy to somebody desiring to sell, we sit across that.
If you're a buyer only, all you see is your component of the workflow. If you're a seller only, you only see your component. The fact that we sit across the entire transaction from intent to offering inventory and the match between them, we are able to solve the workflows across start to finish. That being said, we do not have religion of whether it is our buyer agent, our seller agent. We do have them out in market. You can use yours, you can use ours, you can white label ours. We can give you some of the features that we offer. What our benefit is removing friction and then taking share in the agentic world. We will optimize anybody's buyer agent that connects to ours or another seller agent, our Magnite Orchestration product, make sure that they all work seamlessly and very well together.
You're thinking about this and you're like, well, agentic is really a workflow solution, and you're right. Our workflows generally are people setting up and removing friction from APIs. Really what agentic is replacing is APIs that require a lot of manual configuration, monitoring, and compliance and change. We are ideally set to be the agentic replacing APIs and workflows of all end-to-end solutions. What I would tell you is that if we're successful in doing that, as agentic share increases over time, whether it takes it from programmatic, which doesn't bring us any net new spend and gives us the same take rate, there may be a cost advantage. But if we're able to take disproportionate share in agentic versus where we sit in programmatic, that's a good guy from a market share perspective to us.
If we fail, then there's some risk in market share and that others could start to peel that away. I think we're advantaged from the amount of inventory we represent. There's a lot of unique exclusive inventory that we only have access to, especially in CTV. So there I would bet, I would say that the probability based on the supply that you have and where you can optimize is probably in a much, much better market position. There's some risk in the non-CTV side because others have access to a lot of similar inventory that we do. I would say that market share gains are probably the opportunity from the easiest perspective, broadly speaking.
The one opportunity that we're really targeting that we talked about at the conference is in agency spend today, there is a significant amount of money that is still tied into manual buys that are coming in through insertion orders. That's tens of billions of dollars that still live and sit within agencies. Let's take a step back and peel back what that means. Today, there are brands that walk into their agency partner and have a campaign goal that they'd like to achieve. They hand that over. It's normally a two or three page brief that they hand over to the agency.
The agency spends two weeks running around and finding and creating a media plan based on all their extensive history and all the brands that they've worked with and all the campaigns that they've run and new inventory that's out there and all the upfront presentations they've listened to, they come up with a media plan. That's the secret sauce that says, "Hey, this is where we think you should run your ads to give you the highest ROI of your ad spend." They do that, they come back. The brand then says, "Yes, looks good. We approve. Let's run a test." They run a test where they then knock on our door and say, "Hey, we'd like to do a test campaign.
It's only a few thousand dollars or 10 or 20 or $100,000, and let's run it." We want to do a very deep dive on how this is performing. How is the creative performing? Are we hitting the right audience? Are there refinements that we can make? Are there certain inventory that's not performing? Are there others that are performing at a really high level? Let's run a test and then analyze those results, refine that campaign, and then rerun a test based on joint adjustments to that campaign that we've now agreed to. They do that 3 times, each of which is another two week process.
Now you're looking at eight weeks from the time that you walked in the door with your campaign goal to the time that you're actually comfortable that you're deploying and running your full ad spend that you'd like to, and it's not this quarter anymore, it's next quarter. What we do is that entire process in 10 minutes. Because we're connected to inventory, any LLM and an agency can do this on their own. You can convert a proposal or program goals into a media plan quickly. But without being tied to inventory, to real live inventory, you can't test, run, rerun, and then modify your creative because we give you your creative tool as well to version it anywhere it needs to go or create different versions of your creative, A/B test those in real time.
The fact that we're doing all of those things, and you can run three tests in 10 minutes, and redo all of your creative and have it ready to go, that's what we're solving for, and that is new spend that already exists within the insertion order market at an agency and allows them to be better, faster, cheaper, and bring it in to the ecosystem quicker, and it's all net new revenue or net new ad spend to us that runs through agentic workflows. That's really the initial area that we think has dramatic inefficiency in what agencies have told us they want help and assistance in unlocking.
Got it. Understood. I think your ClearLine product also sort of aims to target at the agency spend that is currently tied in user orders. Is this agentic buying path sort of considered as an upgrade to the existing ClearLine product suite? Is that how we should understand it?
Yeah. ClearLine for us now is an entire product line. There is ClearLine curation, there is ClearLine audience activation. You can effectively run an entire deal where, if the deal is negotiated in an agency marketplace, they negotiate the price, the volume, and what inventory will be purchased. We execute all that programmatically, make sure we find your targets, make sure you find the users that you are looking for. What ClearLine allows you to do is at the end of the day, be a payment method for that agency. The brand then says, "Okay, I have done my transaction," and either ClearLine or the DSP basically becomes a payment method at the end of the day. That is really kind of what that final role is.
Yes, you can do that without having to pay a DSP take rate or do it at the DSP that you would like to either, A, get credit for that ad spend as part of a commitment, or be able to run it most efficiently for purposes of reducing your ad tax.
Got it. Can you talk a little bit about the economics of these agentic buying solutions? Obviously, you are monetizing your supply better. Can you share there? Does the agentic sort of path or the agentic tool also carry some sort of economics by itself?
Yeah. I would say today, even if you are running it programmatically, which the thinking is it is all machines running it, there is still a degree of friction that is involved and matching up or sharing deal IDs and making sure those pass through appropriately. There is more on the cost of revenue side in order to support those from an economics perspective. We do not charge more or less of a take rate to the publisher. Our deals with publishers are we get a take rate from them. They are interested in what CPMs we are able to drive for them and how much overall spend we are able to deliver.
If we are continuing to grow their spend and grow their CPMs and be an effective partner on the base criteria we are measured, they generally do not come to us and say, "We would like you to do it for a lower fee." They are really happy that revenue continues to outpace and grow relative to other players, and that market share position is something that is highly attractive to them. So it is really more on the cost of revenue side than there is an incremental pricing to using those tools versus using our people.
Got it. So how should we think about agentic buying versus DSPs as sort of two different routes of ad purchases? Do they compete with each other? I guess, just based off of your client engagement so far, have you seen advertisers sort of consider moving some of the budgets over from DSPs to agentic?
Yeah. Maybe let me even take a step even further back, right? I think what traditional DSPs have done, especially in CTV, is they have done a very, very nice job of converting very large enterprise-type clients, whether it is large brands and large agencies, trying to move and migrate budgets from linear into connected TV. They also have done a very nice job for those same clients to be able to access and find their users in the open web in order to be able to broadly target and satisfy their campaign goals, largely through brand advertising. Much of that has not necessarily been mid-funnel or lower-funnel. So they have done a masterful job of being able to service that customer cohort, which is very concentrated from a buyer perspective, right? It is logically who you would go after.
Why wouldn't you spend your time going after the whales out in the industry and secure that business? I think that has played itself out in open internet, right? You've kind of had a maturing of that space and that industry in open internet. I think in CTV, you've also started to see that cohort mature, right? They migrated over, they're actively spending, and now you're kind of subject to what happens within specific verticals, whether CPG is up or down, or autos are up or down, or financial services, or healthcare, or drug manufacturing. You're really subject to how those verticals are performing to kind of what your future is. What's happening in connected TV specifically is you're having new demand come into the market, and that new demand doesn't necessarily come through a traditional DSP path, right?
Small and medium businesses don't even know what a DSP is, right? So they want to be able to transact, they want to be able to find an audience, they want to be able to generate their creative, and at the end of the day, they just want to find the fastest way to spend their ad dollars for the highest return to generate demonstrable and visible increase in sales and earnings and performance. Right? So they don't go to market as a small and medium business and say, "I need to hire a DSP, I need to hire an agency, I need to hire somebody to do my creative. I need somebody to version it.
I need somebody to be able to analyze this data, tell me if it's working or not." They can't hire five different vendors to do that work like a large brand or a large agency can. So what's happening is agencies are trying to cater to some of this on their own, small, medium, and mid-sized agencies, by offering all those tools and being a one-stop shop. You're having DSPs like a MNTN, or a tvScientific, or a Moloco try to do that in market. Amazon's trying to do it because they've got a large installed base of customers. You've got many of our publishers that say, "We don't even need that path. We want to offer this self-service." So you have Roku's curated marketplace, you have Warner Bros. Discovery NEO marketplace, you have Disney DRAX.
I would assume that every large streamer out there and broadcaster will eventually have a self-service buying option. By the way, those are marketplaces that we white label and build for them to allow for that self-service to take place, and then mix that demand with all other demand sources. From a demand perspective, we are agnostic to where that demand comes from. In fact, we're tapped as a partner to help build out all those different demand paths and have no blind spots of how demand enters into CTV. What they're all doing, however, is they have to access that supply from Magnite.
Our relationships and being effectively the mediation layer through which the world accesses CTV supply has become increasingly important and doesn't just relegate us to a few DSPs, either ups or downs, or a new entrant isn't a risk or a benefit to us. It all flows. What we've seen in our inflection and overall CTV growth rates is really demand has broadened. SMBs have entered. You have guys like Genius Sports entering into the market as specialists in buying live sports inventory. The market really is broadening, but they're all coming to us to access the inventory, which is the good guy from the market position that we hold in CTV.
Got it, interesting. I have some questions coming through from the audience. The first one is, "What are the overall targets or KPIs for how you manage the business? How should we think about the building blocks of the earnings algorithm in future years?
Yeah, I would say that at the very highest level, we are very motivated to grow ad spend. From an ad spend perspective, that is our lifeblood. That is our fuel. I don't think you'll ever see us in a position where we're willing to sacrifice ad spend and market share loss for take rate expansion. I think even now when we're probably in a position from how strong our tech is and the breadth of our partner relationships, we probably have room to raise prices on our rate card for services that we offer in CTV. But we have not, and we don't plan to. I think we're very comfortable.
Our take rate is something that's not really an important KPI for us, to be honest, because we're not trying to influence or push buyers to buy unnaturally where they don't like to, because generally that doesn't end well. I think that where we'd like to play is being the best tech at the best price, and even when you start to consider, "Do I want to do this internally? Do I want to build this myself?" We always want that answer to be no. For this cost, there's no way we'd even consider this doing internally or doing this ourselves or hiring somebody else to be able to do this. And we're measured by how much revenue we drive, and it's not really a, "Hey, I'm really evaluating.
This is a service that I could easily solve for, and I could do it for a cheaper cost. I think for us, really it's ad spend, it's overall revenue growth. Those two things being heavily linked together, but again, not doing it with take rate expansion or trying to push price increases through. We will bring new services to market that drive new value and things like doing home screens for a Samsung and a Vizio and an LG. That's an ad unit that they did internally on their own. That's new incremental revenue. We'd rather define our revenue growth and our success by doing more things for them that we can do more efficiently, drive more revenue, and do it for a cheaper cost. Yes, we'll share in those economics, but that's also not a take rate increase.
That's just broadening the purview to service more inventory for them. I think, margins, we've talked for years about our margin performance, right? We've said, "What does the operating model look like?" We've said above 10%, we start to get very, very high flow-through from revenue conversion to EBITDA. This quarter is a great example of what we've said all along is we think that above 10% growth, you start to get about 70%, 80%, even higher percent flow-through from revenue to EBITDA. Our beat in Q2 was $10 million on the top line, and that equated to an $8 million EBITDA beat. That's exactly that 80% flow-through that we've talked about. We'd rather continue growing revenue, and as a result, we've taken our EBITDA margin guidance up now three times. We started the year where The Street was under 35%.
Now we've guided to at least 37%. That's a pretty attractive way to see that flow-through happening commensurate with our revenue that's inflected higher. We do look at internally very, very carefully. We don't share a whole lot of this externally, but we look at cost to serve an impression. We are constantly inserting rigor to how do you best serve, most efficiently serve each impression that you have in the business, whether it's DV+, whether it's CTV, whether it's live sports. That's a critical function. You've seen us move to a hybrid model in connected TV from being cloud-based before. Even though being cloud-based is a bit more agile, the cost benefit of being on-prem for a predictable load and a predictable volume is it's a 70% - 80% cost reduction for your predictable load.
We'd love to hit the cloud for inventory spikes and peaks that we hit, but for our base minimum load annually, we'd like to service that all on-prem because the cost advantage there is massive. Same thing on DV+. We do almost that entire market because you don't have those spikes as you do. You have a more predictable pattern of ad impressions that you're supporting and then auctions that you're running. There, it's all on-prem because you don't have those massive sport events where your volumes can go from 2 million to 50 million when people get an alert about a particular game.
We also look for cost per impression to continuously drive that down, and generally that has been dropping in the very, very strong double digits, meaning, 30%, 40%, 50% on an annual basis as we continue to drive costs out of the ecosystem and get more efficient.
Got it. The second question is: what are some of the main areas of pushback on the story that you hear from investors?
This quarter we did not get many, so that is a nice place to be, at least this quarter. I think the one question mark that people have is: Look, what is the long-term trajectory of our DV+ business? There is a part of the business that is open web that may be declining annually in the call it high single digits to maybe 10%-ish . There is a part of the business that is mobile app. Mobile app actually grew 17% last quarter. Mobile in total grew. The streaming parts of our business in DV+, TV streaming over desktop and mobile is healthy. Audio streaming over those devices is healthy. Digital out-of-home is healthy. Commerce media is healthy. So I think not the one pushback, but the question is: what do we underwrite for growth in DV+?
I would say in the near term, as you saw this quarter, flattish is probably the best way to kind of underwrite and de-risk what your outlook is for DV+. We might get a good quarter where it is up 5% or so or more. You may get a quarter where it is down to a similar amount like we have seen. But thinking of that generally as flattish, then over time, the good parts of it will continue to get bigger, and the small parts, which are about 40% of DV+, will get smaller. You will get a nice healthy trade-off that skews you towards growth from a mix perspective. That is excluding if we get any Google remedies that come through or we sign any AI search-related partners that we are actively trying to market and sign.
I think that's absent any of those type of growth vectors that start to contribute in that market that are, again, completely outside of numbers today.
Got it. That's super helpful. And guys, again, just a reminder, if you would like to ask questions, please feel free to raise your hand or just email me at arthur.chu@bofa.com. Okay, cool. Let's keep going. Nick, you mentioned some of the pushback from investors. I think one or two years ago, there were a lot of discussions, or maybe I should say like fear around these premium publishers like Disney or Netflix. They're potentially building out their own sort of programmatic tech stack. I think we've all seen that sort of not being the case. I think some of these streaming services, they are increasingly reliant on tech partners like Magnite. I guess, first, are you still hearing some of the same sort of concerns from investors? Number two, what are some of the takeaways from that episode?
Yeah. Let me just say, I love that we live in that fear every day, and that's what makes us better. That's what makes us hungrier. That's what helps us innovate. That's what helps us develop new products and features, and to continue to push programmatic. The more and more people rely on us and the more spend we're pushing through programmatically, the lower the chances are that somebody's willing to completely rip and replace and take a chance of doing that on their own. Not because they couldn't, could somebody try to do it on their own, even if the cost. We would challenge anybody that our tech spread over all of our clients, if you were one client trying to stand up a similar amount of tech to do that, not leveraged over 30 large partners in the industry, it would cost you dramatically more.
From a cost equation standpoint, I don't think it makes a whole lot of sense. Our modest take rates, I think is also something that plays to our favor that we can still generate very healthy margins on. The speed of feature development and touching more inventory, I think is also the thing that just starts to dizzy people to think, "How could I. Maybe I could do these one or two things," but Magnite just came up with 10 other things on their roadmap that they can deliver in the next three months that's going to unlock more revenue. I can't possibly keep pace with that if all I see in my purview is what I'm doing myself. The reason I said I love that is, again, it drives us to continue to innovate.
Our history has been that fear has existed, but we have never lost a partner. We have expanded every single partner relationship that we have had, either geographically or vertically or more services that we provide. We have probably won every account that we have pursued in the last two years. I asked our chief revenue officer, could he recall a single pursuit that we have not won, and his answer was no. Maybe there was, on the tail end, something that we did not necessarily win. I think that speaks to our continued innovation and our continued fear of just what you mentioned happening, and living to make sure that we ensure that does not happen by being on the cutting edge of bringing more to market, developing more, and leaning in with more and driving more revenue for partners.
Got it. Let me pause here, see if we had any questions from the audience. All right. Let me start with one more. Nick, you mentioned the in-app part of the business within DV+. We have seen some of these LLM companies like OpenAI or Anthropic venturing into the advertising business. I think over the past two years, we have all seen how search referred traffic has declined. I think with the rise of agents, there is a possibility that we could be actually looking at a future where even apps could be reduced to APIs or MCPs. I guess, just curious how you guys think this could evolve over the next few years in terms of impact on the digital ad ecosystem. Also, as you guys have these conversations with the model companies, what do you think their strategy is?
Do you think they want to become closed ad ecosystems like the walled gardens, or could they be more open source, open platform where these traditional ad tech intermediaries can also play a role?
Yeah. No, I think it is undecided. I think everybody is going to have their own path. I would not say that there is an opportunity within Google, who has the entire stack and has ad server and SSP capabilities and demand capabilities. I think all others, it is kind of open territory for how they choose to build up their own ads business. Either it is on a short-term, medium-term, or a long-term basis. I think the fact that you have heard that some of them have talked to DSPs out there is a good sign that they are exploring that source of demand. I think that plays to the advantage to say, "Hey, how do you reach the world demand quicker?
How do you start covering some of these infrastructure costs, not through financing in the debt and equity markets, but do so with actual revenue? I think that's a helpful accelerant and a question that's posed that allows an opportunity at least to exist for us to pitch. I think the market is one that we never played in historically in web search. We play in open web, but we've never had any business in search. If we get tapped, that would be a new incremental market that, again, we've never serviced and played in historically. I think that our view, at least, is that the app market is pretty well insulated and secure, and everybody providing access to their apps through an LLM is probably a lower risk than obviously what exists in the open web.
We're clearly seeing open web traffic reduced in the marketplace, and that's what kind of drives that area that we talked about earlier, about 40% of our web business that is seeing some degree of pressure out in open market. Even if you're trying to create an advertising plan, the recommendations of LLMs are that you should not spend all your money in one medium, and you should look at different mediums. The ROIs, if you're augmenting a CTV, a search, and a social program with some open web advertising, probably even bring some dollars back into open web advertising because they carry low CPMs and they actually have a decent return, and they may actually be performant in lower funnel. So you've got different, I think, ebbs and flows that I think exist from the market. But we don't see it as all risk.
We see there being a definite possibility for opportunity. To the extent that they end up as front-end engines to be able to start creating campaigns like we're using it, we're using LLMs as the front end of our buyer agents, I think to the degree that we sit and connect that to inventory that exists, either existing or new, I think there's a role for us to play.
Got it. I think we are at time, so let's maybe just wrap it up here. All right, Nick, thank you so much for your time, and I really appreciate your insights today.
Yep. Really appreciate the time, and look forward to talking to any and all of you in the near future. Thank you.
Absolutely. Thanks, everybody.