Oh my gosh. I got to say hi to Don. Oh, is it too late?
I think we're starting.
Oh, we're starting. Okay. Don't put it on the transcripts. Missed you, Don.
All right. In the interest of time, I think we're going to keep going. I know people are moving from room to room, but it's my pleasure, for the first time ever, to have Liftoff Mobile here at the conference. Jeremy, thanks so much for being here.
Yeah, thank you so much for having us. It is an honor to be here.
So fresh off your IPO just a couple of months ago, and your first Communacopia.
Yes. Happy to be here.
Welcome to the palace. I am going to read a quick safe harbor and dust off my legal degree. Just a reminder before we dive in, some of what management discusses today may include forward-looking statements which are subject to risks and uncertainties that could cause actual results to differ materially. Please refer to the company's SEC filings for more detail on those risk factors. With that, let us get into it. For those who do not know the Liftoff story, w hy don't you take a step back before we take a step forward and just talk a little bit about the origin story of the company and your path to this point?
Yeah, happy to do that. The story starts in, I think, everyone's desk or pocket, if you think about where your mobile phone is right now. We're spending on average about three hours a day on those phones in the app ecosystem. That was really the genesis of the business, that this is now where we spend our time, and it makes sense for an ecosystem to follow that time. What's really fascinating is, 13 years later, I think we're more excited about the promise of what this ecosystem can provide us. When you think about the specific problem that Liftoff has set out to solve, it's really about unifying a business with a consumer. The entire business is predicated on helping businesses find the right consumer, and we do that through advanced machine learning, a broadly distributed software development kit. We do that all in app.
Now, even though this business has reached some scale, and we're sitting here public today, and there's a few other public names that we share the stage with, this is a very early market.
Yeah.
It's really early. We're really excited for the years to come.
Let's stick with that theme for a minute. There's typically most investors I talk to think about this as a gaming opportunity and a non-gaming opportunity. Just frame for us how you think about the market opportunity across those two sort of ways of thinking about the end state.
Yeah. This is a really important question, because gaming specifically was the first highly evolved mousetrap in app. When we first started using our mobile phones, we can remember the early games that we played, and games did a great job of finding an economically viable way to acquire users and monetize their users in this app experience or container. Liftoff was founded to do was to help any app business across all verticals acquire their consumers in an economically viable way. What do I mean by that? If you're a travel app and you're optimizing for those who download your app, book a hotel, and stay in the hotel, you want to optimize against that event. Liftoff helps you find those users.
If you are a ride-hailing app that wants to optimize against those who are booking their first ride, and in San Francisco, I think we commonly use those apps. If you're a game that's looking to optimize for day seven return on ad spend. This heterogeneous mix of consumers became part of the fabric that was Liftoff, and that continues to be a strength of the business today. We work with every single vertical. We're gathering over $160 billion of purchase data annually to help inform our models. Against the backdrop of this app ecosystem that is still very early, we are there to help each business acquire those consumers in their app and monetize those experiences. Gaming has really been the first mover on a lot of the tactics that create a highly effective mobile business.
Now that intelligence is accruing to all kinds of verticals that have figured out ways to really viably build their businesses through performance user acquisition outside of the walled gardens in the mobile device.
Yeah. Let's stick with that theme, because you've talked a lot about how there's a monetization gap out there, right? There's traditional channels by which app developers reach potential scope for users. There's the app economy more broadly, and what you're trying to drive as a market opportunity. Talk a little bit about the efforts you're putting in place to close some of that monetization gap.
Yeah. It's probably the most profound metric to describe the state of our space and the opportunity ahead of us in the years to come. Right now, the third-party app ecosystem is monetizing at about $0.07 per user-hour. So we spend an hour on an app, $0.07 an hour of economics is being developed through advertising. So if you compare that to other screens, for instance, TV. TV had a 75-year head start, give or take. It is about five and a half, six times as efficient. Now, we know that TV advertising, sometimes when the commercials come on, we walk over and we make our kids a snack, or we run to the restroom, or we look at our mobile phone.
It's monetizing at about six times the density, so it shows you how much more there is to go in terms of bringing this advertising ecosystem into the experiences where we spend so much time. You also have these generational tailwinds. For those of us who have children, Gen Z is spending about three times as much time in the app ecosystem as baby boomers are. The propensity for those younger generations to engage in commerce through their phones is amplified asymmetrically. We're here to help really build out the density of that ecosystem. There's still so many different types of experiences that haven't come to the app economy to build their user base. It's just a matter of time.
Okay.
Yeah.
What we've talked about so far, you've got sort of this user acquisition strategy, you've got this app economy opportunity. Where I find most investors get very bogged down is when you start talking about the tech.
And what differentiates. So your platform, why don't you talk to us a little bit about Cortex and what you're building and what you're scaling and how the tech that can capitalize on this gap and drive some of this inefficiency out of the ecosystem, is built and scaled, and where that sits competitively to others.
Absolutely. I'm going to have to use an acronym that many of you—
It's okay.
—probably haven't heard yet.
It's a tech conference, we're okay with acronyms.
It's just two letters. AI plays a role here.
Okay. We've just got to check how you spell that.
Yeah. I imagine that maybe in your first session today it came up a few times. Our business has benefited quite tremendously from the move from what we would've called linear regression to neural nets.
Yeah.
The acronym for neural nets is AI. What does that mean? We are a recommendation engine as an ad choosing platform. We help choose the right advertisement for a user, and we help price that user. What we did in Q4 of 2023 is we launched our answer to AI, our neural net-backed recommendation engine called Cortex. That product, the best analogy that I can think of with our day-to-day is a spreadsheet. You imagine that you had four columns in your spreadsheet, and whenever you wanted to add a fifth column, you had to actually go in and hard code that new variable in. That's what LR was like. Now, you imagine you have a workbook with many more columns, many tabs, and they can interact with each other and select the data. That's the move to neural nets.
What's happened to us since then? We've grown 11 quarters in a row sequentially at an average of about 8% per quarter. The benefit of Cortex to our overall operating model, it's tough to square, but I think that empirical evidence helps dimensionalize it a bit. It's also very exciting where it can go. I'm going to try and keep everybody here with me. We know the names Meta, Google, those who have really been on the frontier of machine learning and helped compress the cost and the complexity of decisions with recommendation engines. That information on best practices, techniques, how to evolve Cortex or our model is out there for us. Now, we work against both our own knowledge as well as well-researched academic material to help evolve our system, and it can evolve in a few ways. We actually now benefit from recursive learning or self-reinforcement learning.
Yeah.
You hear about that term a lot these days. Our model gets smarter on its own.
Yeah.
We also can add new data to the model that may sit within our stores. We also can really recalibrate the weights of different features or produce derivatives of features that we couldn't have before. The roadmap ahead of us is vast. The market itself, this is really an interesting point, AI itself in the hands of our customers is also producing a benefit for us. We benefit two ways. We have Cortex. Our customers now are benefiting from LLMs. Many more apps have been created in the last year than prior years. That's pretty well-documented. We also have different kinds of user journeys that are now being created by our customers that couldn't be created before, that are much more intelligent. All of that capacity and intensity that customers are bringing to apps, that flows into the data that we get.
Yeah.
We're learning, our customers are learning from user engagement, and it's a pretty beneficial cycle that we're just starting to touch on as LLMs start to get more and more popularized out there and we're really excited about what's to come.
That's really helpful, and that sort of frames, I think, the road ahead in terms of how to capture some of that inefficiency. I think the other debate that I typically have with investors is how visible and linear—
Yeah.
—this will all be. When you think about where you want to take the platform from today to three years from now, how should investors think about how much of this is linear versus how much of it can be nonlinear?
Yeah, it's a great question, and my CFO is sitting in the first row. I have to take it. When you think about the way in which we structure our model, not the machine learning model, but the financial model, what we've articulated is this is a very high growth industry. The numbers that we subscribe to show something like an $80 billion TAM turning into approximately $130 billion TAM in five years. The CAGR is there about 13%-14%. You've got a market that is just growing, and we certainly are benefiting from being early, building a really nice foundation in this market. But that's an underlying piece of our model. Then you have the ongoing benefits longitudinally—
Yeah.
—over multiple quarters of recursive learning of the model just getting smarter on its own without intervention, without me or you touching it. What happens inside our building at Liftoff is everything that sits on top of that.
Yeah.
It's did we release a new column in our model, if you will, a new feature as we did in Q2. Did we create a new derivative field? This is what machine learning engineers and data engineers work, they exist to do, what we exist to do at our company. Everything from design doc to it propagating through our system and finding more high-value users for our customers. For our business, I think when you look out multiple years, it's really important to think about the market we're after. I think this helps distinguish us from really any other named business out there that does what we do. Our market is the entire app ecosystem. Gaming is one vertical for us, and gaming is important. A lot of the innovation that was born in this ecosystem came from gaming.
You see these verticals emerging in the last few years. Prediction markets, OSB is now coming to town. You have financially oriented apps that previously were really focused on durable contracts are now engaging in this, let's call it the prediction market craze. You also see this happening in e-commerce. Collectibles are coming to the market. When we think about what's possible in five years, I would really think about where can consumer taste go across the app experience. As hardware evolves, there may be some announcements from some companies we know tomorrow around hardware. There's the software we just talked about is really evolving with experiences. The experiences that we are willing to engage with on our phone are certainly becoming more dynamic.
Our objective is to be there for that next business that catches the consumer craze to help that business acquire customers that are ROI positive and will acquire as many as our model allows.
Okay. Last one on this point. When you think about the application of either capital or data, how should we be thinking about the need to scale compute for model performance, data for model performance? What are the variables you're the most focused on scaling to make sure the outputs that come from model performance over the next three years—
Yeah.
—whether linear or non-linear, still get you to the end state that you think is a real possibility for the company?
This is a fun question because I think it helps us paint the brush to see where we sit in the ecosystem and gives you a little bit more context on our operating model and how diligent we are in terms of assessing how accretive a given innovation is. Our business, if you think about how many features we have in our model, just to keep it simple, think tens of millions. We've heard LLMs quote numbers in the many billions.
Yeah.
We've heard Meta disclose where they are in terms of features. Why would you have a different count of features or different orders of magnitude? The problem that we are solving is a very hard problem. It's actually impossible for it to be a one out of one. We are trying to assess the probability of a user engaging with a business, installing an application, and either purchasing or showing some sort of proxy for an ROI positive user, and we are pricing that opportunity. It's a narrow problem that we are solving. Because it's a narrow problem, still a hard problem, we need less features in order to respond to the prompt. Because of that, we don't need to communicate across boxes or across chips.
It's different than the photo that you took of something where you are asking an LLM to interpret a visual and then give you directions. Because of that, we can use chips that are not as competed for in terms of the pricing. The liquidity that's available to us in terms of our cost structure is advantageous. Now, we also have the benefit of determinism on a proprietary level. When we release a model, or I should say, when we make a decision to release a model change, we can see down to the bottom line how accretive that usage of compute is.
Yep.
If the usage of compute that we see as short-term accretive, long-term destructive, or the inverse, we can use that calculus to help determine if we should merge the update. This is a real benefit in B2B recommendation engine-focused business model in the types of chips that we can use and the determinism of our releases. Certainly if adding a zero or an order of magnitude to our feature set we see as accretive, we will do that. We will make the economically rational choice.
Okay.
Yeah.
Moving on from the model and compute, you talked earlier about the mix between gaming and non-gaming. But when you think—
Yeah.
—about your go-to-market strategy and emerging verticals, that's obviously also relatively unique relative to the industry. Can you talk a little bit about standing up your go-to-market strategy and what it means for not only the non-gaming side to continue to grow, but how there could be increased verticalization of non-gaming that you gain exposure to over time?
Yeah, absolutely. Our business, having been founded on the back of this more heterogeneous mix of customers, started establishing not only a machine learning approach that was optimized to heterogeneous outcomes, but also an operating model. We aim as a business to be the first to be able to make a new vertical work. Whether that be prediction markets or LLMs or collectibles or you name it. We've seen that the empirical evidence over the last few years is quite powerful. There are different needs that are quite parameterized in terms of what types of needs they could be, but there are different needs that a given business has as a function of the way in which they engage with the consumer. For instance, as I mentioned earlier, if I'm a hotel booking app, I'd like to KPI against, did that user stay in the hotel?
Well, that data signal is determined after the user actually completes the stay. Our business, we have become really well attuned to the way that reporting structure may vary, the KPI structure may vary, and the creative storytelling may vary based on a given creative or a given product, excuse me. That model scales in a nonlinear fashion. In fact, the way in which our workforce has evolved, leveraging technology, leveraging systems, adapters, as we've discussed in our materials, this is an operating model that is truly native to our business that is very hard to repeat. It is very hard to go and build out an entirely new business that caters to a different type of advertiser, and that's something that we pride ourselves on. I think that should the next vertical emerge, you should expect us to be part of the growth story of that business.
Okay. When we pull this all together, we start the process of pulling it all together. When you've got the growth you're putting up, the market opportunity, the verticalization opportunity as well, how do you prioritize what the key investments are that need to be made in the business when you think over the next 12 - 24 months that are critical to capitalize on this opportunity set?
Yeah. With capital, I think the first question on a macro level is: Where does the money go? Reinvesting in the business is our first choice. Getting to the right amount of leverage is our second choice. We're rapidly delevering as our disclosures would show on our earnings call, we discussed. The next bucket would be returning capital to shareholders, which is certainly a consideration given the high flow through of our business, given how profitable it is. Then your fourth bucket would be M&A, which we have empirical evidence of having done M&A, both tech and team, and transformational. Liftoff with a capital L is a function of a merger between two well-known businesses. Within the reinvest in the business area, the index towards technology is always very strong.
Now, I mentioned how we make decisions that are deterministic in nature as it relates to deploying an extra dollar of compute. We have systematized to a pretty detailed degree, where we see dollars working for us in our favor. As you know, Eric, we have had a private equity investment for the last six years.
Yeah.
We learned to make responsible decisions as a business many years ago. This is our opportunity to show it to the public markets now. We're 90 days in.
Yeah. Okay. Understood. Within the investments you would make into the product, the platform, and the growth, talk to us just a little bit about what you think about some of the more higher yielding return investments might be. Is it platform investments at this point? Is it continuing to scale Cortex and seeing what the output of that is?
It's a fantastic question. I think that when I consider what variable we're investing as the most scarce resource, it's time.
Okay.
It's the time of our team, because the opportunity cost of that time is quite tremendous, and the primary area of focus is always going to be on developing a smarter recommendation engine. What does that mean? Well, there was this step function paradigm shift from our old model, if you will, to Cortex. Now it's about taking intermediate steps within this paradigm of advancing Cortex.
Okay.
How does that investment get deployed? There's time, and there's also, we have a team that has tremendous tenure. We have some of the best experts in the world working on our platform, and we need to retain and engage that team. It's a really strong group solving a very specific problem. That group will then work to build out that architecture of Cortex further. Now, some examples could be, are we going to optimize for an entirely new KPI for an advertiser? Well, that could mean that there's opportunity cost of a team being deployed against that project. We could be testing with customers. These are derivatives of strategic choices that we've made at the highest altitude that we end up executing against.
Okay. Understood. Probably the last big-picture topic I wanted to talk about, which is another output of a lot of investor conversations, is I still think there's a lot of misunderstandings about the competitive landscape. I think a lot of conversations with investors get bogged down into gaming versus non-gaming—
Sure.
—demand versus supply, what's a differentiator here. What is your imprint you want to put on the conversation of what the competitive landscape looks like today, and how you see the industry evolving against the market opportunity and all the inefficiency you see in ads today over the next couple of years?
That's a fantastic question. That would get the "Family Feud" one.
Oh, there you go. Okay.
Yeah. I think that it's a very unintuitive structure of this market, but it's understood by looking at a few things. One, you have a vast supply ecosystem. Meaning that there are more impressions available for advertisers to buy than there are qualified advertising dollars, and I'll get back to that in a second. What that means is, just in Liftoff terms, our SDK is integrated in apps that see about 1.5 billion users a day. The SDK is in, last time that we reported on it, I think 167,000 apps. These are all different businesses across the ecosystem that have implemented ads in a specific way that's proprietary to them based on their monetization goals. You then have advertising dollars that are flowing in from businesses that are looking to acquire qualified customers.
Well, first, that is a pretty scary, big problem as an advertiser. It's not like TV. It's not like the web where there's a single surface. You're looking at millions of different applications that could be delivering your product proposition. That's really challenging. As a result of this, and we talked about the $0.07 per user hour of monetization earlier, you have less demand flowing into the ecosystem than you have qualified supply. At the same time, you have very few businesses that have the qualifications to deliver users to you at scale. What are those qualifications? I think there are two. Does that marketing channel have a ubiquitous software footprint into the inventory where your ads will be shown? Does that marketing partner have a smart recommendation engine that can find users for you that are ROI positive?
If you apply those two filters, you have very few businesses. You also have across N equals about three different underlying data sets that are going into the recommendation models that create very different predictions and desires for the model. You now have, as an advertiser, you have a few different channels that can go and find you users at scale in an economically effective way. Your objective as that advertiser is to acquire as many of those as possible. This is a major distinction between, let's call it Madison Avenue advertising and performance advertising. The customer here is attempting to buy as many consumers as possible that are ROI positive, not attempting to distribute budget across a certain amount of partners that's fixed. The pie is not fixed.
The supply side is vast. Because of that, you have the customer conversation sounding like, "Jeremy, why aren't you able to get me more users? I want you to get me more users." There is no shifting of budget. Effectively, there is a macro or ensemble model that you're deploying as a user acquisition manager. That ensemble model may be Liftoff and app and you and others. It could be a subset of those. You're going to be attempting to get as many users as you can that's ROI positive. I'll get off my soapbox with this last comment. Our business functions a lot like cogs in a P&L. The primary executive customer these days is the CFO. When you think about how that works, the CFO doesn't make a decision to go and use more cloud tomorrow.
Cloud is the lifeblood of the business. You have to be using your technology in order to grow your business. It's just like user acquisition. This is why the winner take all or winner take most sort of paradigm doesn't apply to this ecosystem. You also have this property of it being vastly underpopulated. Gaming versus non-gaming, I think is a different question. What I would say is we, as consumers, we might play a game because you're in between sessions this afternoon. You might use a ride-hailing app, where you're the same person in two different experiences. Because of that, I think it's not too big of a leap to expect that we could advertise different products to you that could be relevant to you as a consumer across those experiences.
As a business, we do not aim to achieve any mix of gaming versus non-gaming. We want to be where consumers' attention is accumulating, and provide high-quality users to those advertisers who are accumulating that attention.
Okay.
Yeah.
To bring it all together, we only have a few minutes left. Over the next couple of years, where do you think the biggest growth opportunities sit inside the company where you're aligning your strategic priorities against those growth opportunities? Then the second part, which is kind of a follow-up to the last one, is just how do you think about where competitive advantage for you as a company will accrue over time?
Awesome questions. The primary area of focus for our business, from day one through day X, will be enhancing the quality of the recommendation engine. In a business like ours, you're getting it wrong most of the time when you're predicting the propensity for that given user to then buy something in the given advertised app. There's just so much runway to improve that the question is really around sequencing the efforts and the tests, and getting smarter and faster with testing. I expect there to be tremendous progress over the years to come with the quality of our recommendation engine, and I think the empirical evidence is in the public domain.
When you look at the innovation curves of some of the businesses we're hearing about day to day that are on the cutting edge, some of the LLMs, the Metas of the world, who have taken the neural net paradigm and advanced it many years into the future, it gives us as much determinism as you can have when you're thinking about where the model can go. That's certainly very exciting, but I'm also really excited about the consumer side of this. The amount of apps that we're seeing created far exceeds that of previous chapters of the app economy. We're seeing the democratization of content creation, and innovation within the container itself, which I think means that the focus will become increasingly on the storytelling and the marketing of those businesses.
Certainly, content creation is important, but when you can push a store update with software, it's quite frictionless, especially now with access to LLMs and some of the more advanced technology. I think we'll see the content creation threshold continue to come down, and we'll see the focus on telling that story in an economically viable way increase on a relative basis. I think what we'll also see is that we, as consumers, will have more specific experiences that are really interesting to us that we'll be able to consume. Hardware is evolving. We've just talked about software. What does that mean for our business? Well, it's very clear to me that the direction of travel is a more heterogeneous experience across the app ecosystem, meaning you're not just talking on the phone and playing games. You're booking a hotel, you're booking a ride, you're banking.
You might be engaging with the NFL in various ways tomorrow evening. Now, there are all kinds of different experiences that we're engaging with. What business is best suited to help those businesses grow in the app ecosystem? I believe that Liftoff is.
Okay. Well, we hope you had a good, enjoyable experience here at your first—
Thank you so much, Eric.
—Communacopia. Hope you come back—
Thank you.
—next year, and thanks so much for the conversation.
Thank you, Eric.
Please join me in thanking Liftoff for being part of the conference. Thank you.