All right, great. I think we're gonna go ahead and get started. I'm Steven Gee with the UBS US Internet team. Sitting to my left are the management team of AppLovin. We have founder and CEO Adam Foroughi, and CFO Matthew Stumpf. So thanks for coming, guys, and welcome to Arizona.
Thanks for having us.
Thanks. Awesome. So let's put you, Matt, on the hot seat first.
Sure.
We'll get to you in a second, Adam, so don't you worry. So let's talk about the gaming opportunity first and the gaming advertising and the growth and the durability there.
Sure.
So that continues to grow pretty well. It had a broader sort of industry benchmarks. I think we're all used to thinking, like, the audience there is already tapped out.
Yeah.
Maybe, like, we're done expanding in the sector here. But, you know, what's driving the outperformance there, for AppLovin? How durable is it?
Yeah.
Right. And, you know, what specific benefits in MAX is actually contributing to that growth?
Yeah, sure. So, to understand that, to break it down, there's a few factors that are going on. First, you have to understand that the supply's grown.
Yeah.
So most gaming companies now use MAX, right, the MAX marketplace, which is a marketplace that's growing on a double-digit kind of pace annually. So there's a large opportunity for continued growth for all of the demand-side platforms, including AppLovin's own solution, but every demand-side platform, to continue to feed into that supply-side opportunity that's continuing to grow.
Yeah.
And then on the demand side, you have the AXON model, right, that sits behind the AXON Ads Manager that's really doing the matching, right, for the Ads Manager. So what's happening there is that the technology's continuing to improve. We've broken that down for investors in the past. There's multiple components there as well, the first being ongoing learning. So that's the model just continuing to learn as it shows impressions, what happens, was it correct or incorrect, you know, in its predictions around, you know, what would happen after the impression is shown. Then there's directed model enhancements, which we previously said would happen at a, you know, kind of a one per year, which was an oversimplified, you know, explanation for investors just to understand the basic concept of what's actually happening.
In practice, it's enhancements that are happening, you know, multiple enhancements, dozens, really, over time. And then those enhancements, each one of those lifts, compounds over time. So the addition of both of those pieces continues to grow the scale of spend that's possible for the advertisers. And then on top of that, you also have additional factors like demand, right, which is continuing to grow in terms of the diversity.
Yeah.
So as you add on those technology enhancements on top of the more diverse demand and more data, there's this multi-year opportunity for us to continue to grow into that mobile gaming supply.
Got it. So that's what's informing the long-term growth outlook that you've articulated previously.
Exactly.
Okay. It's not like a one-time we're done, benefits done, staircase function.
No, I mean, the technology is very nascent as well.
Yeah.
So those directed model enhancements, I mean, there's a very long list that our engineers are working from of things that they wanna continue to test and see whether or not they could potentially provide an uplift. And then there's research that's going on all the time, right, that's continuing to add to that list of potential changes to the models that they also wanna test and see if it will perform. So we feel like because the technology's so nascent that there's a multi-year runway for us to continue to improve the model.
That's good. Good. Adam, spotlight on you now. So, we can't let this meeting go by without talking about the web advertising opportunity, right, the broader opportunity. That's about 17 months old now since you guys rolled it out. So you've been bringing in new cohorts, you know, on an ongoing basis. So how is the new cohort performing? You know, what have you learned, right? And how do the newer cohorts compare to the old, you know, the older cohorts? And, you know, how do tools like the new one, prospecting model, and Gen AI affect performance for all of the cohorts?
All right. Cool. You got a ton of questions in there. So maybe to answer, I'm gonna back up to when we first released the product and what were we really looking at to know that we were on the right track? And then what have we done since? And we can bring it to the present. We really focused on last year, bring on a fixed cohort of customers and sort of lock it in. I think over the last decade, a lot of companies have tried to go into this space and have prematurely rolled out products and probably not really focused on optimizing those products for what the advertisers needed and ended up seeing some revenue growth, getting really excited. But the premature rollout made it so that the products weren't scalable.
What we wanted to do to reverse that was really say we've got some products from companies across a whole bunch of different categories, so reflective of the broader market, huge market. I mean, we're talking hundreds of customers, and Shopify alone has millions of shops on it. So the very, very large market in terms of end customer count, but you can extrapolate from small to big. And what we wanna do was in the small set, go in and say, if they're able to spend on us on a performance basis, and what I define performance as is drive someone, a consumer, to buy something they didn't know they wanted to buy before and sell the advertiser's product to them and predict every component of that conversion funnel so that you can price for the advertiser on a return on ad spend basis.
If you can do that well, you drive true performance, you turn the marketer into an arbitrage marketer, and so the dollars they can spend on a platform become really limitless with the constraint to that return on ad spend goal and the money in their bank account is the only two constraints, and so we wanted to execute on that. We wanted to know, out of the dollars that they're spending, how do we index against other media channels? Of course, in this direct-to-consumer, shopping category, Meta's been the best over the last decade, and so we wanna know what's our performance relative to the other players, and we wanna know what is our ROAS, return on ad spend that we're driving for them relative to the other players.
If the spend scale was high and the return on ad spend was competitive, we knew we had a scalable product. Now, we also looked for qualitative feedback. What did they not like about the product? We just rolled out the product a year ago. We didn't even have a dashboard. So a lot of the last year was spent building the basic plumbing, the dashboard, the flows, the integrations with attribution companies, the Shopify app. So the plumbing is pretty obvious, and we had to build it and time allowed it. But a consistent point of feedback we were getting were the customers that weren't scaling didn't appreciate that our platform had only one campaign targeting type. It was universal targeting. And in that universal targeting, they would get new customers and retargeting.
The customers consistently said they wanted to be able to target only new customers. In a complex world of attribution, if you run a business, it's much easier to know if you get a customer you haven't seen before that buys your product, it's very incremental. If you get a customer that had bought your product before to buy another product through a media channel, you're not really sure would that customer have bought next week because of an email they got, through internal CRM. And so was some of that money or all of that money wasted on the media cost? And a couple weeks ago, we rolled out this product on our product blog, called Prospecting Campaigns. I would urge investors to try to talk to clients about usage of this tool. It's working phenomenally well for the customers that are already on it.
We had some testing it out, and we scaled it out because we knew that it worked really well. Universally, it's allowing customers on our platform to share data into the system and ask the model to really go target new customers for them. It increases the new customer rate dramatically from what the old settings were. This allows us to now go across the board. We're not really hearing qualitatively anything negative anymore. Quantitatively, we're seeing really good results. Whether the old cohort or the new cohort of customers we're seeing come in, we're seeing really good results. What that leaves us with is one last layer outside of what's gonna end up being just ongoing optimization or improvement of the product, getting the customer to have creative that's adapted for our platform.
The biggest spenders in this category understand that, on average, a creative on our platform is watched by a consumer in e-commerce ads for 35 seconds. That's a lot longer than other channels. In social, the average is seven seconds, and so you have way more time to engage the consumer and try to convert them on a brand offering than you do elsewhere. A lot of the customers that come to our platform, though, in the e-commerce category, don't have the resources to go make AXON-specific advertisements, and so they bring over their social ads. Those tend to be 10-15 seconds in length, and if you could've shown a user a 60-second video clip and you show them a 15-second video clip, there's a loss. It's hard to measure the loss. It's opportunity cost loss.
Their campaign may still work with the 15-second ad, but maybe they spend $10,000 a day on their ROAS goal, and they could've spent $50,000 a day. Very hard to measure, but not that hard to solve anymore. The way we'll solve that is with generative AI-enabled ad creative, where we will help the customer get tools, a model written on top of the frontier models that allow for video creation and static ad creation. That's shortly on the way here, and if we can solve that, what we'll end up doing is equalizing the playing field for these customers in terms of building creative adapted to our platform.
If they're already qualitatively not asking us for more product and we deliver them creative that is tailored for our audience and their performance continues to scale, we think we're gonna be in a really good place to then get to the point where we should be marketing more aggressively, selling our product more aggressively to more customers, opening up the product, and hopefully unlocking years of growth in front of us.
Yeah. I think if you talk to the mega caps, I think they've talked about effectively solving for the two tasks that advertisers have, right? So there's the media campaign management task, and then there's the creatives, which you just touched on. It's surprising to me that people are just kind of bringing over their social campaigns, but I guess that's a short-term fix to a longer-term problem, which you're gonna hope to solve.
It's the easiest way to get going. So in a way, it was seen as a perk. AXON has made it easy to port our creatives over. We don't have to do any work to get a campaign live. We have to integrate. You click a couple buttons.
Yeah.
Draw pixels or Shopify integration is click a couple buttons. Off you go with your other ads, so it felt really good to them. They didn't have to invest more resources in that, not investing more resources. A lot of them also didn't realize how big we are, and if you knew that the opportunity was as big as any other opportunity out there, then you would invest the resources, but it takes a while to build that trust with the customer on the other side.
Yeah.
On the campaign management side, we started from a place post-AXON too where we removed all targeting. The customer can plug in their economic goal, and they can plug in the countries that they wanna target and where they wanna route the click to. That's it. So they didn't have campaign management in our domain, and it felt really good to make the ad experience seamless where they could port things over. And we have no way to tell them, "You spent $10,000. You could've spent $50,000 had you built better creatives." So if you can't measure the opportunity loss, they don't actually believe it. We'll be able to prove it when we're able to help them adapt creative for our platform.
Yeah. And I think you have, you know, 1.6 billion daily active users. So, you know, I think it. I don't know. I don't think anything that I'm hearing from you this morning is a product-related problem. It seems like it's an awareness-related problem. Then that seems like an easier problem to tackle.
Yeah. The business problem is much easier to solve than the product problem. The other thing that Matt touched on is the health of the ecosystem is really important. What drives our success over time is increasing our conversion rate, but also increasing supply. Daily actives are not gonna grow a whole lot. You're in the mobile gaming and mobile ecosystem in general is just mature at this point. You're not gonna have a billion new mobile gamers appear out of nowhere. And so if you assume the daily actives are sort of capped, but engagement is going up. One thing that I'm really excited about is over the next three to five years, if more and more of our impressions shift to these shopping ads or anything that's not gaming, it'll be really healthy for the ecosystem. And I need to unpack the why.
What drives value of platforms that are advertising-based is increase your conversion rate, which increases your CPM, and increase your ad load. More supply equates to more revenue growth. In our case, because we have that MAX marketplace, more supply grows all the players in the ecosystem and obviously the gaming companies. Now, why did I say it's very beneficial to serve more e-commerce ads? Today, when we serve a gaming ad, the ad is a video plus a minigame inside a game. The natural instinct on that is to think, "Oh, the churn risk to the gaming company is high." Like, so that is for sure a fact. When a consumer shifts off to another game, they turn off the original game potentially. More importantly, people love playing those minigames .
There's a certain amount of time that people will spend every single day inside games, and that amount of time isn't going up that dramatically. They love playing the minigame . So the ad experience in a game is much, much longer than a shopping ad that does not have a minigame inside of it. If over the next few years, majority of our ads shift to e-commerce ads, the ad load time will go down per ad served in the MAX auction, which means the ad load overall impressions served supply will go up. I think it can go up dramatically because of this function.
Now, then it would be easy to conclude, "Well, you're cannibalizing your games category, and you need games to be spending as much as they were before." In fact, if this happens, I think gaming will also actually be able to spend more 'cause what'll happen is our system will get more targeted on gaming. We don't serve 1,000 ads a couple years ago when we had talked about a 1% conversion rate on games and do a great job inside games. We're showing users who are playing a game, they love game, game, game, game, game. That's sort of an annoying experience. And the conversion rate's 1%. On the average, this is a couple years ago, predating e-commerce. Now, the model itself on those 1,000 impressions doesn't predict a 1% conversion rate uniform over the 1,000 impressions. Some of those impressions, it might be able to convert at 10%.
Some of those impressions at a 0.1%. So there is a curve in terms of value, and that, that's backed out from that expected conversion rate. If we go to a world where a majority of the impressions become non-gaming, ad load goes up, opportunity goes up, but then on the other hand, gaming might get very targeted. The model may no longer serve the gaming ads when the conversion rate expected is really low, so the ad is annoying. It'll serve the gaming ads when the conversion rate is really high, and it's predicted that the user's ready to go try another game. Gaming has no loss. More ad load allows them more opportunity, actually. They still get the highly likely to convert users, and then e-commerce and other categories fill in the rest.
Health of the ecosystem grows, and we think that'll catalyze growth in ourselves, but also all the other players in the space.
Yeah. So in just unpacking a little bit, so instead of the 30-second minigame ad, which I'm sure all of us have played, right, so you're gonna see an e-commerce ad that might be five seconds, six seconds long. So you can spread that engage that ad time spent watching the ad over many more ads, and that's gonna help you increase the ad load. Is there an opportunity to go back to the publishers and say, "Okay, like, let's increase the ad load because you can stick in another ad here, there"?
So the publishers' ad experiences are pretty well defined at this point. If it's rewarded ad, the user's opting in to watch an ad for currency in the game. If it's in the middle of level, the user, the ad spot is defined. But if the user can play twice the amount of levels because the ads are shorter, then they're gonna see twice the amount of ads, right? That's really the function to understand. So, on average, we've disclosed the e-commerce ads are 35 seconds watched. Gaming with the minigame is gonna be much, much longer than that on average. And so I don't know off the top of my head what the delta is, but it's much higher.
and you can imagine that if you can cut ad time down by a material amount, let's say to half, then you're gonna probably get a lot more ads in your session 'cause you're gonna have more time to actually progress in the game more and see more ads.
Got it. And so what we're talking about is potentially broadening out the list of potential advertisers that any given you know, publisher is gonna have exposure to. So theoretically, their yield on the dollars that they're gonna be able to get as revenue from you should increase as well. So you would have to think that some of them will be putting some of that money back into the market and to help along with what Matt just talked about in terms of helping the overall growth of the game market.
Yeah. I mean, look, that happens programmatically, which is a nice thing. And it's not just us. If there's more ad load, all the players can help them monetize better. If the ad load goes up and the CPM stays constant and twice the amount of impressions for the same number of users, the publisher's making 2X.
Yeah.
That automatically flows into one of our gaming models that's popular is the ad publisher can buy on a return on ad spend basis in our system. It's all real-time and programmatic. So if that model goes, "I'm seeing a lot more ad load. I'm seeing a higher LTV for the customers that are buying on a return on ad spend basis," they would then automatically be able to bid more into the auction. So that's really one of the things that's fueled so much growth inside MAX is our ad model has to take a user who's playing a game for 20 minutes a day in casual and drive them to a game where they'll have higher engagement. That really unlocks expansion of ad load today in the absence of this phenomenon where ads are shifting to shorter ad views.
And then over time, as that happens, that'll automatically help them able to monetize better, bid more for users, get more engagement, grow their user base, and grow their revenue. So it'll be a really healthy dynamic in the ecosystem and one that should compound over the coming quarters and years.
Yeah. We touched on this a little bit, but I mean, you've launched a referral-based, self-serve Ads Manager this quarter. What infrastructure did you need in place for that launch? And what remains, you know, before the full global GA next year?
Yeah. I mean, I always get asked, "What were my expectations?" And, as a CEO, I always think of the worst, not the best, and I don't really set expectations around customer count or growth and revenue. What I cared more about is that as an advertising platform, I think we're the largest advertising platform at this point ever to really open up. I look back on when Meta opened up their Ads Manager, they had $5 billion of revenue. And our ad spend a year ago, roughly, we disclosed $11 billion of ad spend. So very large platform that hadn't opened up. When you open up a platform that's got that much economic opportunity, you really immediately have a fear of scams and frauds and more aggressive advertisers coming in to take advantage of that platform.
And in the early days, traditionally, that's happened to multiple platforms that have rolled out digital advertising, open Ads Managers. What we didn't have a few years ago were automated ways to scan and protect against this problem. One of the things that we invested in was our AXON to be able to scan every single ad creative, every single website being advertised, every single app being advertised, and ensure the quality meets the quality of what we had when we curated the advertisers on the platform. I wasn't sure if this was gonna work 'cause we had to stress test it in the real world. But I was really happy with the result that we have automatically been able to screen and remove anything that would've been a breach of what we want to provide the consumer.
And we think of the consumer on the other side getting an experience early in its life cycle of our e-commerce ads. We don't want them to buy a product that a lot of other consumers have complained about. If that product ships too late or it's a crappy product, that consumer won't come back and buy from us the next time. So that's the way we think about what the ads are that we're delivering to the end consumer. Maintaining that quality high was really critical 'cause that gives us confidence in the next step. The next step is optimizing every step of that conversion funnel. To us, when a customer signs up, they're getting an experience, and it's not a human-curated or human-engaged one.
It's automatic from that point to the point of put a credit card on file, integrate, go live, run ads, and start scaling ads. We don't want to have a human being involved any step of that process. Therefore, we need to see, are they getting a seamless experience? Are they tripping up anywhere? Is there something that they need to actually have addressed through some sort of Q&A bot? And so for us, it was important to start seeing the data through the conversion funnel, start optimizing that conversion funnel, understand the questions that they have so we can write the bots that can respond to those questions, and we can have a customer service bot and a sales bot to be able to make sure that they get a seamless experience. All of that we're building out, and all of it looks pretty effective thus far.
So it gives us a lot of confidence that, like I've said before, first half of next year, we'll be able to open up the platform. Now, what's important about that is we can't just go open up the platform and expect a bunch of customers to come to us. We need to pair it with the sales and marketing effort. And with a lean team, as you all know, we have very limited salespeople and very limited marketing people. The focus there is, can we run paid marketing campaigns to bring advertisers to us? We'll pair that with some brand efforts, but that's really the big catalyst. And we've been in testing on that. I mentioned on the earnings call already. So we're seeing results in marketplace there.
We're optimizing our campaigns just like every other advertiser optimizes their campaigns on our platform, create more ads, optimize your conversion funnel, try to get the, the LTV to cost of user acquisition lines to cross and scale. As we get to a point of confidence on that, we'll know that we can open up, pass a referral state, really start putting money behind marketing, and hopefully start growing advertiser count really quickly. That's something that we're confident we can do. We're, we're all marketers at this company. We've proven we're very good at performance marketing for our clients. We think we can be very good at performance marketing for ourselves. And most likely, that's a much easier problem to solve than the one of building a product that was really good for our customers to begin with.
Yeah. I guess this is probably more of a three-to-five-year perspective here, but I mean, when I first saw AppLovin, I was really reminded. You brought the Meta analogy also. Like, I'm reminded of when the very initial days of Facebook when they rolled out the app install ads and they expanded that Rolodex over the course of time with really effective product development and started touching a greater, I guess, Rolodex of advertisers. You know, why should that not be the expanded opportunity for AppLovin? I mean, they've gone from probably what were maybe thousands of advertisers to 10+ million, right? So why should that not be the case for AppLovin?
I mean, they did a very, very phenomenal job. It's easy to forget in 2011, 2012, there were a lot of questions as to how they can monetize the blank canvas. Social shouldn't be able to drive transactions. How are they gonna compete in advertising? And they realized if the ads become really compelling, they become like content and consumers can discover new products. The value is phenomenal that you can create out of that. We have the same dynamic and view toward the advertising market. If you can help the customer get a new customer, price on revenue, turn them into an arbitrage marketer, you don't need to sell them. They're going to spend. They're going to scale. It's really hard to do that. It's easier to say than actually build the technology that does that.
But if we can do that and we can integrate customers, and if we're in the thousands now to the tens of thousands, hundreds of thousands, eventually millions, what we will have is we will have built a data set that will allow us to get more and more targeted on the advertising. As the advertising gets better, it won't be as annoying as you're someone playing a game and you like the game and you keep seeing gaming ads. The ads will help you discover more content. And if we've got 35 seconds of time times many ads per day to the consumer, we have a lot of seconds a day that the user's seeing our content. If the advertising becomes more like content, the conversion rate's gonna go up dramatically from here.
And if that happens across a broad range of categories, more customers are gonna go, "There's another channel. It's got a billion plus customers that's got maybe the best ad experience in the world today." 'Cause there is, there are very few places where you can capture the user's full attention for over 30 seconds in this ad-rich world. And so you're left in this framework that's really, really powerful. I think the opportunity is really big. And you've seen just on social a lot of our customers in, but the result is so good, they wanna be thought leaders catalyzing growth on our platform. And so if we can pair that volume with our own marketing and bring in tens of thousands to hundreds of thousands of customers, our platform's only gonna get better for the customers we have today.
It should follow a similar trajectory to what Meta has shown as really a catalyst to expand the economy of all these categories they've excelled in.
Awesome. Matt.
Yeah.
Let's put you back on the spot here. I think many other companies here at this conference are talking about investing very heavily into GPUs and compute capacity. Yet, I mean, you've indicated that only about 10% of the revenue growth it flows through to data center and GPU costs. So everything that you guys have planned in the background, the models are gonna scale, e-commerce efforts are gonna expand. So do you expect that ratio to hold? And, you know, how should investors in this room and listening online should be thinking about capital allocation.
Yeah.
And the impact of paid marketing, which Adam just hit on, you know, everything else in between LLM and API usage, etc. There's a lot of things.
Yeah. I mean, I think looking at other companies, we're run very different.
Right?
So looking at data center costs as an example, I mean, our team spends very disciplined. I mean, as they launch model enhancements, they look at the downstream impact of what those do to GPU usage. So as we continue to grow and scale and improve the technology, we're looking at the cost impacts as well. So we don't expect that kind of profile of the 10% incremental cost versus the revenue growth should change materially going forward. And in fact, the team is continually looking for areas where we can even drive improvements in usage and reduce. So we don't expect that overall benchmark should change materially. Outside of that, I mean, the cost structure is relatively straightforward.
Headcount costs, which everyone knows we've run very, very lean, will grow headcount over time. Obviously, we'll need more business development sales folks to address more customers if we have the millions that you're mentioning.
Yeah.
As well as more engineers over time. But we don't think that that should change the overall margin profile of the business. And then Adam's point on performance marketing, that'll be a new cost that we're gonna have, and we're testing now, but it's very small dollars. Over the next year, that'll grow in absolute terms. But because we're doing it in this performance marketing fashion where we're growing it just like our customers are, right, with.
Yeah.
Return on ad spend goals, you should see a short-term impact on absolute dollar growth, but the overall margin profile shouldn't be materially affected.
Yeah.
We think that's kind of where we're at today in the like low 80% EBITDA margin range should be pretty consistent going forward.
Okay. Sounds good. We only have about a minute or so left, so spotlight that to you, Adam. There's been a lot of growth in the mobile gaming sector, advertising space. Everybody's noticed. And are you seeing any impact from any sort of increase in competition or, and I guess more importantly, like, what are the underlying things that could keep you ahead of the competition?
I mean, look, we're always focused on our own execution, and we, we've got a model that works really well inside gaming. We've become the largest demand channel for gaming, really a requirement in the category. We've got a model that's now proving to work really well for e-commerce companies, and hopefully we can become a requirement for those companies. If we can execute on the road in front of us, you'll, it will sound like a broken record. We really do have the core technology that's working and driving performance across multiple categories now, and, and the rest of the, the path for us is just go get more customers. That, that's been solved by a lot of companies. Even companies with weaker ad offerings have hundreds of thousands of customers on their platform. So it's very likely we're gonna be able to solve that.
And as we go get more customers, it adds more data back into our system. More data paired with the power of our models will make us more targeted. That should benefit every customer in our system and every new customer. So we're really excited about the prospects. If we execute forward, this business is gonna be doing a lot better in the next year, two, three, five years than it is today.
Awesome. We'll leave it there. Thank you very much.
Thanks.