Classic. Ready?
Yeah.
All right. Good afternoon. I'm Justin Patterson. I lead the Internet Research Team. I'm really excited to have Matt Skaruppa from Duolingo back with us this year.
Hey, everyone.
All right, Matt. So, you know, to kick things off, 2024 was a really solid year for you, with record subs, record DAUs, record bookings. Talk about just some of the momentum you saw at the end of the last year and some of your top priorities for 2025.
Yeah. No, you're absolutely right. We had a great overall 2024, and it ended on a really high note. Q4 was excellent, and it was better than expected pretty much across the board. We saw user growth go faster than we expected, which led to a bunch of goodness and bookings and whatnot. And then towards the end of the period, we ran a promotion once a year, a New Year's promotion where we ran our only discount of the year. That went really well. In particular, it went really well with Family Plan, which we were surprised by or exceeded our expectations. And then throughout the period, Max did really well. And we had an inkling coming out of Q3 that Max was poised for a nice run because Video Call with Lily really had shown that it was a product feature people were willing to pay for.
And so Max did great. And then, you know, on the cost side, we were able to execute with discipline and get the margins to where we wanted to be. So 2024 ended with a lot of momentum. And that's why in 2025, when we just guided, we guided to over $1 billion of bookings for the year, 27% constant currency growth rate, and highlighted the fact that we want to continue doing what we're doing with the main SKU, Super, continue to see progress in Family Plan, and then make progress with Max being even more engaging, which Luis talked about a lot on the call. We can get into how we're going to do that. And there's just a lot of other low-hanging fruit with Max that we think we can optimize.
Yeah. So let's stick with just Max as a starting point. 5% of paid subs right now. It's still expanding to more regions. It just launched on Android earlier this year. What are really the key steps you need to do to make this a more engaging product and drive that paid sub penetration number higher?
Yeah. So you're exactly right. End of the period, so end of the year with 5% paid subs being on Max. A lot of you will remember we actually launched Max well before last year. And for a while, it had approximately 0% of paid subs. And then video call comes out in Q3. We ramp the video call through Q3 and Q4 such that now at the end of the year, some 70% of some odd DAU could purchase Max. And that ramp really did help throughout the back half of the year. That ramp is pretty much done. It's probably 80-some% at this point. It could go a little bit higher, but the remaining areas are unlikely to be the priority. And so the goal with video call and Max this year is to make video call more engaging.
So right now, you're talking to Lily on the phone, and she's really capable at holding a conversation, but she can be funnier, have better memory, be clever, keep you more engaged in different ways. She can look different. Right now, she is kind of a two-dimensional on a background. We can make that much more engaging and joyful. And so we will plan to do all of that. And that's kind of on the product side. And Luis mentioned some of that on the call. I think on the other side of things is we love the fact that it's twice the price of Super in most places, and there's a lot of demand for it, but we have an optimized price. So we can optimize price.
The other things we can do are, you know, we've had success showing internal house ads to get people to convert to paid at all. And usually, that looked like showing them a Super ad at the right time and right place in their journey. And we haven't optimized how Max shows up for those internal ads yet, what the mix is of Max and Super and all of these things. So basically, that's a long way of saying I think there's a lot of opportunity for us to go drive increased Max penetration.
Yeah, for sure. I think one of the interesting tests there was just even people completing X number of lessons, opening a chest, and having a one or two free day trial of Max in there. So it seems like there's a lot of different ways you can drive that adoption. So you mentioned not really optimizing for price so far, but you also haven't optimized for cost. You've taken this decision, let's just keep proceeding forward, investing in this product, albeit with a little gross margin degradation. So talk about just why that's the right decision right now.
Yeah. So to put that in context, again, we try to manage the platform to increase the platform LTV of the subscribers. We run a bunch of experiments, and we can talk about our Green Machine. We put out a handbook that I don't get a royalty for, but I wish I did. You all should buy the handbook. It's free. You can download it. We talk about all the experiments we do. And if you look at our experiments on monetization in particular, they can basically boil down to, does this experiment, this change we've made in the product, increase the lifetime value of the subs on the product or not, more or less. For Max, being that it's twice the price of Super, you start with a really nice LTV bump if everything else was equal.
And then because there's marginal AI costs, mainly from video call with Lily, as you interact with her, we have to call the generative AI features to respond. The gross margin percentage then drops below Super, but the gross profit dollars per Max subscriber is greater. So in general, we're very comfortable maximizing for LTV because we think it's higher LTV. And I think financial-minded investors should be good with that decision because it also is creating more gross profit dollars, which creates more Adjusted EBITDA dollars or Free Cash Flow dollars. So that's how we're thinking about it now. All the things I mentioned around making Lily more engaging, more fun, all of those product experiments, there's just a huge pile of fruit that's fallen off the tree that our teams want to go tackle.
The trade-off then, because we don't have infinite people, we could run infinite experiments, is: do we have that team trying to make Lily funnier or maybe shorter conversations or longer conversations, figuring out how to best make Lily engaging, or do we tell them to optimize costs? What we've chosen for the first couple of quarters of the year is to focus on that engagement piece so that we can drive more LTV in the platform. That comes from a place of our deep belief is that the cost curves for generative AI will come down and will be fine, not even in the long run, but certainly in the long run, but in the back half of the year.
For sure. I think that's a good segue back to just the broader engagement on the platform because what you just outlined suggests that there should be more engagement, not less over time. You keep making a better product. Yet I think from the investor side, we all see just DAU starting to decelerate against tougher comps and acknowledge that once you go through that transition, DAU growth deceleration, that can be a tough point for companies to navigate. So talk about just what you're seeing today and why you might cross that chasm of just DAU deceleration.
You know, it's a question that makes me chuckle a little bit, Justin, because our Q4 DAU growth was 51%.
I don't think I've ever said anything that makes you chuckle, Matt.
51% is not bad. I think it's really good. We are really glad about it. And I guess in the strictly second derivative nature of the world, yes, we said that Q1 was going to be slower growth, but it was still going to be mid-40s, which is really, really good. So the way we think about it is that our DAU has gone from when we went public, it was 10 million DAU. It is now 40 plus million DAU. So we've scaled it enormously, and we're still growing at mid-40s. We feel really good. That's incredible growth. And it gives us a ton of freedom to continue to just optimize the product around engagement. And if we're successful at that, then I don't think that that's going to be a quote unquote chasm.
That'll be. We feel really strong about the DAU growth that we have and the engagement that we can use to drive it with the product features. And we aren't even talking about the social media features and the fact that our unhinged mascot faked his own death and got over a billion interactions on social media. So there's a whole nother side of this, which isn't just the product engagement. It's the social-first marketing that we're known for.
So, is that going to be the shirt, "fall off my dead Duolingo" one, the resurrected Duo?
Yeah, where's your coffin? I thought there were going to be some coffins around here with Duo.
I did actually think about playing Spanish or Finnish for you, but I wasn't sure about the Spotify royalty rates of that.
No, no one knows about those. Yeah.
Actually, I do like every Spotify, but so it goes, so transitioning back to just pricing, you've mentioned optimization a few times in there, so let's just stick with Max. How do you think about optimization? Because you do have this goal of making Duolingo an affordable product globally within there, helping people really learn, improve education globally, so there's high per capita income countries, low per capita. How do you think about just the long-term pricing philosophy around Max?
Yeah. Well, I can't, maybe just because I'm long-winded, sorry, but I can't just talk about pricing philosophy with Max because we're so lucky that we now have a viable third tier that our pricing strategy is not, it's tough to isolate it.
For sure. We can extend it to Super, Family Plan, and rest.
When you think about our goal and our mission to make this education really available and cost-effective, our free product is still amazing. We have a really strong subscriber conversion that comes from a free product with incredible engagement and not a lot of friction. I mean, you cover other names that have free products that introduce quite a bit of friction to get conversion. Our friction is pretty light. Ad loads, not very much. We're going to keep that. That is one of Luis's goals, is to keep that free product accessible around the world, rich countries, developing countries, it doesn't matter, as a free tier. The pricing opportunity really comes in twofold. One is you can mix shift our subbase from Super to Family Plan or Max. We've done a good job with Family Plan.
Family Plan is now over 20% of our subscribers with limited product feature differentiation. It's mainly a pricing differentiation. But because it's a higher price, that's been helpful to ARPU. And then Max, obviously, we've talked about a lot, is a much higher price product. So as we shift from Super to Max or to Family and Max, that does increase your ARPU over time, which is probably the bigger lever for ARPU increases for us relative to the second most important lever, which is price point experimentation. And we run price point experimentation not infrequently. And what I mean by price point experimentation is like in the U.S., the price is $84.99 or thereabouts. Should it be $89.99? Should it be $94.99? What should the price be?
Or should it be cheaper, $79.99, in order to change the mix between Super, Family, and Max and to maximize the lifetime value of the subscribers on our product? So we're doing both of those things, but the pricing experimentation is, I think, first and foremost meant to drive a mix shift to the higher price plans.
Got it. So we've talked a lot about just the monetization side in there, but I want to back up and just talk about the funnel because you're right. You do have a massive base of MAUs and DAUs in there. And you now have new products out there like Advanced English. How do you view that as just changing the engagement opportunity, bringing people into the top of the funnel and converting over time?
Yeah. So what Justin's referring to is the fact that we've always had intermediate and advanced learners across all languages on the platform. What we did in a concerted way was if you look at the TAM for language learning in the world, first of all, you just have to say it's very, very large. I don't know the exact number of people learning a language in the world, but it's got to be close to 2 billion people. You might say not all of it is addressable, and you can haircut that out wherever you want. You still get to a really big number, way bigger than 40 million DAU. A lot of those folks are learning English, and we didn't have all that much intermediate content for English. We had some, but we thought we could have more.
In 2023, we made a concerted effort to build more. In 2024, what happened was with generative AI, we found a bunch of ways to optimize our content creation internally using Gen AI with humans in the loop. Humans are still involved, but there's a chart in our shareholder letter that shows just kind of an exponential content creation that we've been able to achieve with Gen AI.
The coursework chart.
The course, yeah, exactly. And that has helped us create more advanced content for English learners, the biggest part of our TAM. And when you couple that content with Video Call, you're now just much more able to see a path to Luis's vision of making our app be as good as a human tutor, but also to really get more folks who are trying to learn English to better their lives, better their economic circumstances, to pay us money. By the way, a lot of them are already paying money to someone else, mainly tutors or schools or these types of things. And what we need to do is then get more of that share. And I think with that content and with Video Call, you can see us hopefully, our hope is to accelerate into that TAM over time.
For sure. And there is a notion that because Gen AI did give you that explosive content creation cycle, that others in the market might also benefit from that, which leads to more competition. So I'd love to hear you just expand on the moats that you see in place at Duolingo since to your point, you do have a freemium model, you do have a team that does a lot of A/B tests and a lot of data of seeing how people who are German-speaking learn English and English learning other languages.
Yeah. No, it's a good question. I think one of the moats is our merch. No, I'm just kidding. It's a great shirt.
I mark it myself.
You're excellent at it. I wore a pocket square and you wore a dead Duo. That was a mistake. In terms of the moats, I think that when you look, we don't comment on third-party data sources, and there's a good reason for that. They're kind of unreliable and whatnot. But if you look at any of those and you want to figure out what market share is for MAU or some of them published DAU, we have the vast, vast majority of people learning a language online on our platform, 90% by some estimates, and that's not really changed much. If anything, it kind of ticks up over time. So we feel really good about our position.
Why we're so happy to invest gross margin dollars in an engaging Video Call feature, for example, is because when you have that lead, we feel like we can just compound that lead by doing things like Video Call right now. When it comes to the moats, though, again, my kids, my 10-year-old, when my wife was like at the breakfast table one morning a couple of weeks ago, said, "What are you guys doing? You killed Duo." My wife was unhappy about it, and my 10-year-old got teary-eyed. I say that not because N of two people who are related to someone who works at Duolingo, but there were a lot of people around the world who were mourning a fake mascot. What does that even mean? That means that your brand has become pretty synonymous with language learning, but also iconic.
That has to be one of our moats, just that when people think of bettering themselves through education and specifically language learning, they think of us. They don't think of the category. We kind of are that category. That's a moat. The other thing that's really important is having the content. I mean, I think Luis has directly responded to a question that you had a while back. There's tons of books out there on how to learn language. The moat that's most important is our engagement model, like keeping people coming back every day. Language learning takes continual practice, hopefully daily practice. That's what our app does, drive engagement to get you to come back, to give you those rewards, to make you feel better about yourself. That's certainly a moat.
These are things like, just to put a point on it, the streak feature, your social network on the app, streak with friends, these types of things. So between the brand, our lead, our technical capabilities, our process internally, like the Duolingo way and that handbook, and our engagement model, I feel very confident about the strength of our moats.
For sure. And kind of building on that point, you brought up social. You've actually made a lot of social capabilities within the app in there that I don't think actually get enough attention within there. So as you've made Duolingo more social, what have you really observed among the user base?
Yeah, I mean, the numbers in the shareholder letter, I don't have them off the top of mind, but they're pretty impressive. I mean, the number of, it's something like a third of our DAU, something like that, have a streak with friends, like a real engaged social mechanic that keeps you coming back. My mother-in-law challenges me to streaks or to things all the time, like.
Do you let her win?
I have to, well, I know. No, that's not what competitors do, but I try to be responsive. Anyway, so I think what we've seen there is, again, we measure engagement in a bunch of different ways. Those social features keep people coming back, and they help us drive those engagement measures. And I think you're right. We talk about it in the shareholder letter, and they're probably most visible to folks who are actually using the app and see it and touch and feel it. Because when I just rattle off those statistics and talk about that, it's probably hard to grok. But if you use the app, you'll see that when your friend Justin challenges you to an exercise, you're going to do it.
Sounds good. One more for me that I'll open up to the audience. So as we just look at Duolingo today and judge progress over the course of the year, what's going to define a successful app?
The thing that we are going to measure internally. We run the business based on OKRs. And if you look at our OKRs, they go roughly in order of how you all probably build your models. They are an objective and key result around users. There's a bunch of things implicit in them around conversion that lead to bookings and LTV goals. And then there's a bunch of hiring and other expense goals that lead to a profitability goal. So internally, that's how we will measure success. Externally, it's not just Max. The user growth that we are targeting for this year, we think will be very strong. The overall bookings growth, as I mentioned, is 27% on an FX neutral basis. That comes from Max. Max is materially in that, but there's a bunch of other things in that. Most of that's coming from Super and Family Plan.
And so it's really not just one specific thing. I think Luis is going to judge it not entirely, but one of the metrics in his mind is going to be how many multiples of fun is his interactions with Lily when he does a video call with her every day. He's learning French. He's got some crazy, like 2,000-day streak in French. And he's going to say, at the end of the year, he's going to say, "When I talk to Lily every day, is it 10x, 20x, 100x more fun for me or not?" That is going to be important to him.
You do have that nice dopamine hit of, "I can now see where my progress is for proficiency in a language." You actually have that real-time output back in there.
Yeah, and at some point in the future, we'll talk about how you're talking about the Duolingo score in the app and just how if we are successful, we mentioned 90% user market share. If we're really successful over the course of many years, we will have people in the hallway of a conference like this. They will say, "I speak French." And they're, "Oh, really? You're a Duolingo 50 at French?" And they'll say, "Yeah, no, I'm a Duolingo 65." Because once you do that, then you really have the lingua franca for the lack of a better phrase. And that will redound to a moat or a benefit.
For sure. My toddler's saying, "Hola chica.
Hola chica.
All right. So with that note, I'll see if the audience has any questions.
Not on that last point, but yes. I'm dodging that one.
Me too.
Yes. Yeah.
I'm a loyal user. Alex, I saw you last year here. It's just funny to ask. When you've got a million opportunities, how do you figure out what to go after, and as a user, talking to others, it's like, "Oh, you actually have something, a store and app store or whatever, to have alternative sort of ways of progressing on things that you don't want to invest in." What are your thoughts there? Do you need to get to 100 million users before you do that, or what is the?
We have over 100 million MAU, so we feel good about that. The prioritization in the team, if you take away kind of the G&A functions and just focus on really the core of the people in the company, they're engineers, product managers, and designers mainly. They break out about a third, a third, a third, rough justice. A third of the folks are trying to bring people back every day, the engagement side. We call that the growth team. A third of the folks are trying to make our product more efficacious. They're trying to make it better at teaching you language, call those learning and curriculum. A third of the folks are trying to figure out if those other two groups are doing well, how to monetize the group. Internally, we set goals for each of those. We run them based on metrics.
Again, it really does feel like I'm getting a kickback from our handbook, but our handbook does detail how we run those teams. I think the thing you're asking about, I think that's most important, is Luis has a belief. I don't know that you can draw a mathematical proof around it, but he has a belief that if those three teams aren't making roughly equivalent gains every year, percentage gains or unit gains, that something will not work in the app. So you have to make the app more engaging. You have to make it teach better, and you have to figure out how to monetize it better kind of in parallel. It can't slow down.
And so when we're making trade-offs, the trade-offs usually come in that the monetization team has found a great way to make more money, maybe with ads, because we really only show one or two ads at the end of a lesson. So maybe some genius on the monetization team heard me once say, "Why don't we show 10 ads?" But it really hurts growth. It really hurts engagement. And so anyway, that's how we make those trade-offs is really that trying to balance those three in parallel.
Great. With that, we are out of time. Matt, thank you so much for coming back.