Duolingo, Inc. (DUOL) Earnings Call Transcript & Summary

All right. Let's get started. I'm Bryan Smilek, JPMorgan's Internet analyst. We're very pleased to have with us Duolingo's CFO, Matt Skaruppa. Duolingo is a leading mobile-first learning platform, providing its users and subs access to gamified access to courses in 40-plus languages, also math, Music and now chess as of last week. Duolingo's freemium model creates a strong conversion cycle for its subscription offerings. It has 130 million monthly active users, 47 million daily active users and north of 10 million subscribers. We estimate Duolingo will generate north of $1 billion in bookings this year, 28% adjusted EBITDA margins and $360 million in free cash. Matt joined us as CFO in February 2020 and was previous VP of Growth Equity at Goldman, Principal at KKR and a consultant at Bain. So welcome, Matt.

Good to be here.

Awesome. So format today is fireside chat. And with that, let's get started with a series of questions around users. So obviously, as in your financial results, Duolingo continues to drive outsized user growth and especially across its more mature geos. You've effectively expanded the TAM with 80% of your U.S. users not even learning a language before coming to Duolingo. So with that being said, you've guided DAU is expected to grow 40% to 45% in 2Q. So how should we think about the puts and takes, the drivers of DAU growth here, not only in mature markets, but more broadly in some of these emerging markets?

Yes. No, it's a good question and the place to start because our entire economic model is based on the fact that we have this great freemium conversion funnel. So we bring in new users, convert them and all the rest flows down to profit. So it's the right place to start. User growth has been amazing over the past couple of years. So we did guide in Q2 of this year to 40% to 45%. And I think that's really strong growth. I think it's even stronger when you consider that Q2 of last year had near 60% growth. So we're lapping incredible growth and doing that again. The way we do that is something we've talked about a lot. There's a two-pronged approach. We have a social-first media strategy, which really we unlocked in 2021 heading into 2022. So we've had several years of success whereby we create viral content around our app. We have an unhinged green owl that hopefully haunts all of your phones and passively aggressively bugs you to do your lessons. And we try to insert that unhinged owl into social media feeds around the world. That social kind of first strategy also leads to a bunch of user-generated content on social platforms. So that two-pronged approach of our own content being inserted into viral moments around the world and user-generated content creates a really nice brand flywheel for the top of funnel. The other way we do it is that, as you mentioned, we have this great feature, which is that users of Duolingo oftentimes, when they're using us are obviously users. And then even when they stop using us for a period of time, consider themselves Duolingo users. And so we get to resurrect them. We get to bring them back to the app. And when we do, they often find an app that has improved because the app is getting better every day. And so that leads to a nice, again, retention benefit for these resurrected users. You mentioned kind of how that plays out around the world. Resurrected users make up a big portion of our top of funnel all over the place. But in markets where we've had longer presence, more established, contact with users, they make up even more. And we think that's a really strong brand story, which is our ability in mature markets to continue to bring people back to the app and drive top of funnel and then ultimately, DAU growth is one of our superpowers for sure.

Definitely. And on that note, too, around resurrected user growth, whether it be stuff like the app icon change, your unhinged content and marketing as well, too. How do you think about improving retention as well across these mature markets, not only reactivating users but also just getting better at retaining them themselves?

Yes. So I won't go through. We have a very detailed growth model, which is a recursive nonlinear way we think about user retention. We don't have time for that in this setting. But we have a blog post. If you Google Duolingo's growth model, you'll go straight to it. The answer is, we think about retention of users at a bunch of different dimensions. We want them to come back every day. If they don't come back every day, we want to optimize their weekly recurrence. If they don't come back weekly, we want them to optimize their monthly, and so it goes on from there. And the way we're doing that is we have an entire team. If you take our company and look at -- we have 800 or so people and you take out kind of the G&A and marketing and focus on engineering product and design, about 1/3 of those folks spend all their day, every day, running little experiments to nudge people to come back because, again, we believe that you can have the best curriculum in the world, you can be the most efficacious language learning or math or music or chess learning tool; but if people aren't coming back, it doesn't do anything. So a big chunk of those are just experiments to change the color of things, to change the app icon, to change the notifications, to change how often Lily calls you, things like that to drive user retention.

Definitely. And on that note as well, too, 90% of user growth organic, but you're also leaning in, in a very small way, local hires across country market managers as well, particularly across the English learning markets as well. So how do you think about balancing not only unhinged social-first content and marketing, but also more so, hey, we have this efficacy through the Duolingo score, higher proficiency as well to drive that growth across English learners.

Yes. So to take a step back, we don't spend all that much money on paid user acquisition. As you said, the vast majority of our user growth comes organically. In order to do that, we try to put select people in select countries. I don't know the exact number. We have a handful or more of these country marketing managers, and their job is essentially to tailor our social-first unhingedness, our crazy green owl and make it work in their market. That means finding what's going viral or what's the topic du jour in their cultural moments in their markets and making sure we're a part of that conversation. And so that does help us drive growth in countries. We have had a lot of success in English learning countries, Japan, Italy, Germany, these type of places. And one just like real pity example of what that looks like is we had about 1.7 billion views, I think, the week after the Super Bowl when our mascot faked its own death, the guy is crazy, you can't control him. He can't be controlled. In Japan, our country marketing manager reminded us that you don't really joke around about death like that in Japan, so we didn't do it in Japan. So that's just one very specific action that a country marketing manager might take to make sure that our brand is unhinged, but safe and fun in their local markets, but driving user growth.

Definitely. And he did come back to life in case anybody is wondering.

He did. We undeaded Duo. Who knows? The guy is crazy though. He might disappear again. You never know.

Exactly. So with that as well, just digging a bit deeper into English learning, you've said in the past, 46% roughly call it, DAUs are learning English. Yes, if you look at the broader language learning TAM, north of 75% of learners are studying English. So how do you flip that equation go towards 60%, 70% plus of your learners studying English over time? I know you just rolled out 150 more courses at the AA level on the Cypress scale, but curious more as you dig into intermediate and advanced English, how that unlocks that growth.

Yes. I think there's a strategy we have, which is to address what your question is about, which is that the largest part of the TAM for language learning in the world is English learners, not only number of users, but in terms of dollars. And historically, Duolingo has been effective at getting English learning users. I mean, as you said, less than 50%, but a good chunk were learning English. But they didn't think that we had advanced content to help them go from marginal to more advanced proficiency. Over the course of 2023 and 2024, we rectified that. We added a bunch of course, content in intermediate and advanced English learning. We added an English-to-English course. We did a whole bunch of things to help them believe that we have the content needed to help take them to the next level. And then in 2023 and then 2024, we added Max. And Duolingo Max, which is our highest priced AI-powered subscription tier, has a feature that is called Video Call. You can basically talk to one of our characters, Lily. She's the passive aggressive purple haired girl who I'm wearing purple in honor of her hair because she's so effective for us. You can talk to her. And conversation for English learners is really important. And so the combination of adding a bunch of course content, adding Video Call helps us, we believe, unlock English learning around the world. And you can see that Japan is an example we've shown before, we've talked about before, where if you stacked all the countries by super bookings and when you stacked all the countries by Max bookings, Japan plays above its weight in Max. So we think that gives us a nice runway into English learners. And then there's additional things we're going to do. You mentioned the Duolingo score. The Duolingo score is a long-term extension of a product we have called the Duolingo English Test, which measures formally your English proficiency. If we can extend that such that everyone in this room, when they're talking to each other after about how great and charming the CFO of Duolingo is, they can say, I'm the Duolingo 14 or a Duolingo 26 for your English proficiency, we feel like we can lock in an enormous advantage from that as well. Plus when we talk about my viral marketing, we're also trying to insert more efficacy-based influence-based messages in there. So that's kind of our strategy to go attack what remains an enormous untapped set of payers and users for us.

So on that note as well, too, you mentioned Max, 7% of paid subscribers ending 1Q, up 200 basis points, call it, from 4Q. What does success for Max look like in terms of penetration, just financials overall, too? I know you manage towards platform LTV. So can you just walk us through how you go through the trade-offs of Super versus Max when you're trying to optimize for platform LTV longer term?

Yes, absolutely. So for most of Duolingo's history -- well, I guess, technically, for most of Duolingo's history, we had no paid subscription tier. And then late last decade, we added a subscription tier. And then in 2023, we added Max. So Max is the highest subscription tier. It's about twice as expensive as the next subscription tier, which we call Super, and then we have a free tier. The way we think about that is, in general, almost all of our monetization experiments are to drive platform LTV higher, lifetime value of the user. That is an incorporation of free-to-paid conversion. So how many people who come to the platform using the free tier convert. And then it's a price point and then it's a retention curve for people who subscribe. And what we're trying to do and what we found is that the individual Max product because it's priced twice as expensive as the next tier, has a better LTV profile even when you take into account the fact that we have to pay for some AI services to use that tier. And so we've been trying over time to nudge people into the Max tier. It increases platform LTV, but it shows up most quickly in increased bookings and increased ARPU over time. That said, again, this is a platform LTV game. So to the extent that we find ways to make Max even more profitable or make it more higher conversion, we'll take those opportunities over time to drive up LTV overall. What success looks like is higher platform LTV. But also, again, we're a mission-driven company, and we think Max actually teaches conversation better because of the Video Call product. And so we want that to happen as well. And we see evidence of that. English learners, in particular, use Max, use Video Call a lot more than any other language learner does. And we think that's because English learning requires a bunch of conversational practice and the users and subscribers are finding that really effective.

Definitely. And on that note as well, too, I did want to dig a bit deeper. If you think about the incrementality of Max, you have mentioned, right, one thing that has been positively not surprising, but good to see is a good chunk of crossgraders are converting into Max as well. So can you just help us think about what does that cycle look like, somebody cross-grading from Super up to Max versus somebody going completely free to Max effectively overnight?

Yes. When we talk about crossgrade, it's really just an upgrade cycle. So someone is paying us already for a subscription to Super, and we show them the opportunity to upgrade their subscription to Max and they hit the bid. We have seen, as Bryan mentioned, that over the course of having video calls. So really the past several quarters, a good chunk of Max subscribers come brand new to paying subscription. So they were not paying us and then their first ever selection is Max. And then a good chunk have come from -- they were using Super, so they're already paying us a subscription fee and then they upgraded to Max. To us, we are a company that loves to optimize vectors through A/B tests. So if you give us 2 vectors, and that's what we're seeing, the new vector, so brand-new subscription provider coming in and an upgrade, our team goes nuts. They love to then optimize both of those routes. And the fact that we're seeing both of those available to us gives us confidence that Max has got a bunch of opportunity to it. So we're pretty excited about the fact that both things are happening. Over time, I think it's going to vary, probably vary by country as well as to whether or not the majority of new subs in Max are brand new or upgrade. We'll see how that plays out over time. But the team is pretty excited to go run experiments to try to optimize both vectors.

Definitely. And on the other note of monetization as well, too, I think the other key here is the Family Plan, still less than 1/4 of subscribers over time. Can you just talk about how you want to optimize socialization across sub accounts, how big Family Plan could be as a percent of subs overtime, perhaps as you continue to lean in on building out that subaccount cross activity?

Yes, absolutely. Family Plan is a wonderful story for us because, again, if we're optimizing platform LTV, the salient details for you all about Family Plan are that it does cost more than an individual annual plan, but you get up to 5 additional subscribers on the plan. It shows up in our metrics, though, as 1 subscriber. So for you all doing ARPU math in the back of your mind, don't worry, it's just 1 sub. But the retention characteristic is the best part of Family Plan. In that, it's just a higher retaining plan. The more folks you add to your Family Plan, the better the retention is for the Family Plan. And so what we want to do is not only drive people to the Family Plan because it has higher LTV, but help folks, as Bryan mentioned, have a more social connection with those folks on their Family Plan so that they add more folks to that plan and then retain better and further increase LTV. And we've had some success. So Family Plan launched, I believe, in the fall of '21, maybe it was early '22, but we really didn't put a whole lot of marketing behind it. It just grew organically. People would tell their friends, et cetera. And then we started advertising in-app around it, and that grew the penetration as well such that now it's about 1/4 of our subscribers. We think that there's still room to run in terms of driving further penetration. And one of those vectors is going to be, we haven't done any real optimization around Max Family Plan. There is a Max Family Plan. It exists, but we don't talk about it too much to our users and subscribers. So there's vectors like that, that are just one of many, I think, that we can continue to drive that really nice LTV product penetration.

Definitely. And the final equation of, in my view, at least of platform LTV is pricing. So just to hit on -- you recently did increase the price of Super annual plans only for new subscribers. So if you're an existing sub, you didn't get impacted. Just curious, how does that inform your pricing strategy over time? How did the top of funnel respond? And obviously, I assume this would be embedded in your full year guide of 30% to 31% FX-neutral growth. But curious how you think about ARPU as a dynamic within that guide as well?

Yes, absolutely. So you give me a chance to riff -- to wax poetic about our pricing strategy, so I'm going to take it. Our primary pricing lever is not price point optimization, although that's obviously important. And I'll talk a little bit about that. The primary way we're driving ARPU over time is the same conversation I have been having, I guess, with myself up here around platform LTV. We want to drive pricing higher by driving mix shift to higher-priced plans. That's the primary vector by which we increase platform LTV, but then it shows up in ARPU. So if you have more mix on Family Plan, you have more mix on Max, more mix on Family Plan Max, you will naturally see a higher ARPU flow through the business. And again, that's in service of the lifetime value of the platform, not ARPU, but that's -- ARPU is the output. So that's the primary vector. But we also check price points. And what we do a lot is we try to keep our pricing ratio pretty consistent around the world such that the price in any given country of our plans is scaled more or less relative to the U.S. based on the country's GDP per capita relative to the U.S., rough rule of thumb. That changes over time, mainly because of FX changes. And so we try to run experiments to make sure that we can keep those things in line. And then every so often, we go into the market to test whether or not consumers agree with us that we've added so much more value to the plans that we've earned the right to charge more money. So that's what we did. What you're referencing is in February or March of this year, we'd run an experiment that had a higher price on individual annual plans, so individual super plans. That experiment, again, looks at the trade-off between free-to-paid conversion, that increased price, our estimate of retention for those subscribers that see the higher price and then how does that impact the selection of any given subscriber between Max and Super, so mix. Put all that together, run the experiment and if it spits out a positive platform LTV, launch it. In this case, in February and March, it spit out that 10% to 15% increase in new individual super plans was LTV positive. And so we launched it. But you probably all heard Luis talk about how we run pricing experiments relatively frequently, and that's one of the examples of what he means. We test that not infrequently. And we think that in this case, it just made sense to us. The app today for Super looks, feels -- the experience is quite different than it was a year ago and is very different than it was 2 years ago, and we hadn't raised prices for at least 2, 2.5 years. So it worked out that way. And it is definitely 100% embedded in the guide. I was going to make a joke saying, oh, no, I have to go fix something, but I don't think anyone would laugh in here. I won't make that joke. That was a joke.

Awesome. And sticking to the guide too, I mean, look, 30% to 31% FX-neutral growth this year, we obviously laid out the drivers. Going back to the IPO, right, multiple years ago now, but you talked to the North Star as being north of 25% plus annual bookings growth. What gives you the confidence now the platform is clearly very different with Family Plan, with Super, with pricing. So what gives you the confidence that, that multiyear trajectory will look north of 25%? And obviously, not giving anything on the long-term guide, but curious the drivers and building blocks to get there?

Yes. I think if you rewind the clock to the IPO and we kind of said that the 25% long-term number was predicated upon a nice bit of user growth and a bunch of levers we knew that we had to continue to drive conversion. I mean it was a relatively lightly monetized app when we went public. I don't remember the exact number, but I think it was around 5% paid sub until last 12-month MAU number. We're almost double that at this point. So we just knew that there was a nice runway there. From here, it is a different story. We did not expect to grow users as fast as we did for as long as we did. I mean there were 12 or 13 straight quarters in 2022, '23 and '24 of accelerating user growth, which was crazy because the lapping -- the comparables were quite fantastic. So now we have just a much bigger user base that we think, again, is going to pay off over time in monetization. We have a brand-new tier, which we didn't have. We have the Family Plan, which is kind of a tier overlay. And then we have a whole new set of tools around AI that we are really excited about, and I think gives us confidence in the longer term. What I mean by that specifically is we can create content in a way that at the IPO, we couldn't really have imagined. The amount of content we can create per dollar compared to 2021 is quite something different. I mean in the shareholder letter this most recent quarter, we put a chart that shows the exponential growth in content creation. And so all of those things lead me to believe that we have a good line of sight to fantastic long-term growth. That doesn't even include the fact that we've got nascent new subjects that don't monetize really today and that could, in the longer term, monetize. So things like math and music and now chess have big markets, and they are gamified learning experiences that make you feel better about how you're spending your time on your app or on your phone. They're language like. They are a set of content that has rules, and we try to make them fun. And so if you add all that up, you look into the medium term and you feel good, yes, that we're still on track.

Yes. And with that, too, I mean, very strong incrementals, 35% incrementals on the EBITDA implied in your guide at the midpoint. Within that, though, however, you look at the gross margins from Max as well, right? Max gross margins will improve in the back half, but the cost of AI continues to come down. How do you think about Max driving gross margins over time as well to becoming less of a drag and perhaps even accretive over time, but more broadly, not only Max, but AI. You mentioned it building out new subjects, course units published, continuing to grow at exponential rates. So curious how you think about incremental margins in that framework?

Yes. I think to answer that question, I think, to me, it goes back to what are we using AI for and what are we investing in AI. And for the most part, it comes in the form of content creation. So a couple of weeks ago, we announced 148 new courses or new language pairs really. That was not possible before generative AI. We went from having just an amazing team doing what we call content scaling. So this is new courses, new lessons within a course, et cetera, all of that. They basically made it by humans. And then over the course of '23, '24, they went to human computer-aided content creation, and now they're doing more like human review of content creation. So we're spending money on that. And then in Max, we're spending money because when someone uses Video Call, it has a marginal cost. As those costs come down, I think of it as options. So we want to continue to grow. We're at the early stages of language learning, and we're certainly at the early stages of math, music and chess. And so $1 spent investing in those products, hiring engineers, product managers and designers, which we're still -- our hiring plans haven't changed, gives us the chance to go earn more users, earn more subscribers and ultimately build a much bigger business, and that's the priority. And then over time, as those costs come down, you will see us experiment with other ways to do that. So like maybe some of those AI features that are right now in the higher-priced tier go to the lower-priced tier or maybe we add another tier, all in service of driving a bigger business of platform LTV. While we're doing that, though, I think you all should feel confident that even as we've been investing right now, while AI costs are coming down but have not "come down," we guided to a nice incremental margin for the full year, nice free cash flow per share. So right now, we can do both. We can try to invest for growth and maintain incremental margins. And in the future, we'll have the option, if costs come down, to pick what we do, invest or take the profits.

And on that note as well, too, within your 30% to 35% long-term margin guide. I think the one variable where I've been getting a lot of questions is Apple versus Epic. Everything in the App Store is changing. So I wanted to give you the chance to discuss your strategy not only across alternative payments, but I mean, also there is this notion that, look, I mean, alternative payments do come with friction on the user and sub conversion cycle. So how do you marry those 2 potential long-term gross margin gains at the same time that you're getting OpEx leverage with also the potential upside over time, call it, from Apple versus Epic and App Store changes more broadly?

Yes. Well, before we get to the Apple and Epic, you just said something that I think is also important, which is even absent that, we are getting OpEx leverage as well. So we're trying to do what we've always done, which is manage the business prudently on the cost side as well. If Apple or if the gross margin -- cost of revenue goes down, gross margin goes up because of some of these changes, that's upside. The way we're going to think about it is I am a broken record, at least in your note, you will say platform LTV because I've said it 100 times. That's how we're going to evaluate any kind of change in payment flow. We are going to take a look at what is the friction introduced. So in and of itself, friction to Luis is not ideal. We are trying to make the most fun and engaging user experience and then user-to-sub experience. And so friction in and of itself is not what we're looking for. But we'll look at the trade-off between kind of conversion declines from free to paid that come from this increased friction if there is it. We'll trade that off against incremental gross profit dollars that come from a lower cost of revenue. And then we also have to layer in, does this have any impact on subscriber retention? I think it's too early to be seen. And so we're going to run those experiments. I think what's interesting about the current state is that at various times over the past several years, people have experimented with this. They've tried to take people in various different payment flows. I think what's happening right now is that a whole bunch of people are experimenting with this. And so it's not really a moment in time set of experiments. It's going to be an evolution and a conversation with a bunch of consumers about, does this change how they experience all of the apps that they interact with and the payment flows there. We're just one of those many apps. So we're going to be experimenting with this over time. And I don't expect it to be -- we know in a month or 2 months. I think it's going to be a longer-term conversation with consumers to see how their preferences evolve and how that impacts the friction conversion over time.

Definitely. And so we've talked about a lot of the components of OpEx here and the gross margins as well, obviously. But as you look towards to get closer to 35% longer-term margins at the high end, what gives you confidence from an OpEx perspective that you can get there? We've talked about the social-first marketing, but is there anything else as well to keep in mind too, just as you continue to scale?

I just think it's the history of Duolingo is what I think gave us even before generative AI and any of the content scaling benefits we've seen. Being born in a period of time where Luis's quote before we started monetizing was, I am not getting any money from my users, so I'm not paying for it to get any users. That ethos permeated the early days of Duolingo when it was founded. And I think that led to a kind of relatively measured bit of hiring. Even when other companies in the early part of this decade were hiring really rapidly, expanding really rapidly, we didn't. And so I think that kind of ethos still purveys. And that's why I think you're seeing a lot of operating leverage in general, like the G&A side. On the sales and marketing side, it's the core almost religious belief of Luis and Manu, our CMO, that constraints are good for marketing, that you don't need to go nuts on paid user acquisition that you should be able to grow organically and try to do the hard work to do that. And so I think if you layer those 2 things up, then the only thing left is engineering product and design. And we're very comfortable spending money on that and hiring people because that is what drives the innovation that it will power our kind of longer-term growth and success. So that's how we ladder up our OpEx build.

Definitely, it all makes sense. So I wanted to save some time to talk about the new verticals. We talked about, right, chess launching really in the last few days here. It's very new. Third-party articles suggest there's north of 600 million adults globally that play chess regularly. Curious, as you expand into chess, again, it's very early days. How do you build up the product? What is the opportunity? What is the overlap between language learners and chess look like in your view? And I guess, do you think that chess has the opportunity -- do you have the opportunity to expand the chess market in the meaningful way that you did for language learning?

Yes, we hope so. I mean that is the theory of the case. The theory of the case is chess is one of those games that feels like you're learning that when you play chess, you feel like it's time well spent. It's fun, it's engaging, it's challenging, but you're thinking strategically, it's about logic and memorizing opening gambits and all of these things. So it's a fun game that's very in keeping with the ethos of our app, which is time well spent, bettering yourself, learning something new. And I think we can gamify it in a really fun way. When you all take a chance to play the chess game, it's a wonderful and joyous mobile experience when you put the king and checkmate and it explodes, right? It's just like why not? Yes, sure, fine. That's a fun thing to do. So we think it's along the lines of all the things we know how to optimize and work at. It is a big market in and of itself right now, but we are hopeful that we can do for chess or math and music, what we did for language learning, which is just expand the market. There's a bunch of people who might otherwise not be doing those things on their phone. But if we bring it to them in a fun and engaging and joyful way, they will engage with it. The other thing you should know is that it's in the main app, like there's not a separate app. So all of the mechanics, all of the engineering behind our streak, our notifications, all of this is instantaneously leveraged, which we think can help grow users. And ultimately, what we're trying to do with math, music and chess, is figure out to what scale we can get them so that we know that these things are sticky and wonderful products that people love and then monetize them after that. But that's not the near-end priority.

Definitely. So I guess just to wrap it up, I wanted to do a quick word association, so I'll ask you to prompt, you give a one-word answer, whatever comes to mind right away.

Oh god. This is dangerous. All right.

So the first one would be users.

Growing nicely.

Okay. The second one would be English learning?

Growing nicely.

Bookings growth?

Growing nicely.

Duolingo Max. can't say growing nicely.

All right. It is growing nicely. But Lily, passive aggressive Lily.

GenAI and LLMs.

Content scaling.

Marketing.

Organic, yes.

And profitability.

My job.

Awesome. We'll end it there. Thank you.

Thanks.