Hi everyone, I'm Andrew Boone. I cover Internet here at Citizens. I'm happy to be here with Matt, the CFO of Duolingo. Thank you so much for being here and participating in the conference.
Glad to be here.
All right, let's start top of funnel. So, I'd love to start with the question of what's the bigger opportunity: attracting new learners or re-engaging with lapsed learners, and which lever has more visibility for you?
Yeah. Well, I appreciate you starting the conversation with a trick question, so thanks for that. Oh, got a laugh from the outside, that's great. Well, the honest answer is they're both really important, and I can talk about why I think as you evolve our market and we grow, both will become important but a different, geography. So, you think about brand new to the platform users, and you think about returning users. There are plenty of markets around the world where we just don't have the brand affinity, the brand exposure, or the kind of installed base that we do in some other countries. So, for example, in the U.S., Brazil, certain parts of South America, we've been around for a long time, people know us really well, we have a lot of brand recognition.
In those markets, bringing back users, as Luis said on the earnings call, becomes a bigger part of the top of the user funnel. That's really important in those markets, but there are other markets, like for example in Japan where we have a bit less awareness, where, you know, new will be a bigger part of it. It's really both, and I think it's really powerful that we have the opportunity to use both around the world at different times.
Okay, switch to talk about advanced learners. Right, the question is really how do advanced learners compare to casual learners, right? Do they have a greater propensity to subscribe, or do they require different product sets, and how does that change really how you guys evolve the product to be able to match them?
Yeah. I think you've asked Luis this before, and I think his answer is what I'm gonna say here, which is, our, I'm gonna take a bunch of steps back and then I will come to answer you, but the primary driver of our platform is user engagement. You know, that's the lifeblood of what we strive for. It's what we measure. We have a bunch of metrics. We have a bunch of teams driving user engagement. And what we have found over time is that that user engagement set of metrics, instead of teams that are driving those metrics, doesn't. We don't differentiate between casual and advanced. We say, a user is a user. Engage them as much as you can. Get them back the next day as often and as frequently as you can, and that's how we organize the company.
So it's not bifurcated, and Luis does that for a bunch of reasons, but one of the reasons is that he doesn't want to introduce complexity. He doesn't want to fork the app so that you're like, well this works for casual users and this works for advanced. He just wants it to work for everyone. So, if you kind of boil it down, it's just all about engagement for us, and we don't try to differentiate between the two. That said, you know, we have a new tier, Max, the higher price tier. It's got Video Call, it does appeal to certain people who are learning conversational skills. That can at times be a more advanced skill and appeal to more advanced English learners, and that's what we see with Max.
I'm sure we'll get to that, but so if you think about the evolution of Max, you can imagine that that drives, potentially over time, a differentiated engagement around those, that vector that you're talking about, but it is not something we are organizing the company around. We are trying to just make sure that our app is as engaging as possible for everyone around the world. We think that if we do that, the model will work really well, as it has in the past, which is you bring in a user, you engage them frequently, and you get a bunch of opportunities to convert them to paid over time. They stick around for a long time, and our users who become subscribers have typically been on the platform for quite some time. So that's really the model, and it's not differentiated right now based on casual versus advanced.
Understood it's hard to break out the success of Max and the conversion factor of Max.
Yeah.
but if I think about an advanced learner and their willingness to pay, given they may have different goals, right, economic goals.
Yep.
Do you guys see a greater propensity though of conversion for an advanced learner?
Well, there's a lot of advanced learners who've paid us for a long time. So it's not just now. I mean, there's folks across the proficiency scale who have paid us for Super in the past. I think what you're getting at is the fact that with a Video Call and the fact that we're enabling this conversational practice, that over time you could see us further penetrating the English learning market, the advanced English learning market, which is enormous. For those who aren't familiar with it, there's multiple billions of people learning a language around the world, and a huge chunk of those folks are learning English.
When they come to our platform, most English learners do have some proficiency, and with Max and Video Call in particular, we can now address, you know, speaking and keep them engaged with that, and we have a lot more advanced English content. So I think there's a lot of ways for us to win in that more, you know, advanced English market over time, for sure, with Max.
Okay, given the context that the majority of user growth comes from word of mouth, can you talk about the awareness of Duolingo in the U.S. and Western markets where I do think penetration is higher on that subscriber side?
Absolutely, yeah. So, just to remind everyone, the vast majority of our user growth still comes from organic growth, which is awesome, which is why if you look at our financials, you can see continued progress and leverage in our sales and marketing line item over time. We spend more in absolute dollars every year than the year before in marketing, but it's growing over the past several years much slower than revenue, and so we get a lot of leverage out of that while also growing users at just an enormous rate. I think it's, you know, over the past three or four years, we've averaged over 50% year-over-year growth on users. That is because people talk about our app to their friends and family and to random people on the street, and it drives user growth, which is just an incredible gift.
I think what's happened over the past couple years is we've actually accelerated our awareness, and it's gone in different paces in different geographies, but our brand is now way more iconic in 2025 than it was in 2024, than it was in 2023 or 2022. One of the pieces of evidence I have for that is the fact that when our unhinged mascot decided to fake his own death, hundreds of millions, if not billions of people mourned the loss of a, you know, owl, green owl. I don't know what to say, like, it was crazy. My 10-year-old daughter was in tears at the breakfast table, which was not fun, but like, that was a thing. That means that your brand has reached a certain threshold, certainly in those markets you mentioned.
And so, to kind of go around the world geographically to kind of be more specific on those points, actually I already mentioned that like in the U.S. and South America, we are. Our aided and unaided brand awareness is pretty high. People know about us. And so really the story there is bringing people back to the app and engaging them in the app to drive user growth. In more developed kind of Western European markets like Germany and France, for example, not only is there less brand awareness, but there's less category awareness. And so our job is to, you know, make sure everyone knows that this category exists, you can learn English or whatever language on your phone, and that we are the best place to do that, and we're very engaging and fun.
And then there's, you know, places like Japan where, again, I think category awareness and Duolingo awareness is low. So in all those different markets, we're attacking different problems, but with really similar methods, viral unhinged content. We are amplifying influencers that talk about our product, and then those things combine to help word of mouth growth and drive user growth. And in all those markets, we've seen, you know, really very fast user growth over the past couple years.
So let's transition this out of those markets more into the kind of Rest of World, if you will.
Mm-hmm.
Talk about one, what's the opportunity in terms of, there are major geographies that you guys are targeting that may unlock users to subscribers in that waterfall that kind of we all know. And then secondly, talk about the difference in terms of how you have to reach them, given what may be more of that advanced user kind of need.
Right. Yeah, I think that's a really important question. In terms of geography, like where we are right now, we're basically there are users of ours in almost every country in the world. So that's point one, and that sounds great, and I'm very, you know, lucky I get to say that. On the other hand, there are probably about, you know, plus or minus, and maybe you round up, about 2 billion people learning a language around the world. You know, we have over 100 million MAU, which is a sizable number, but relative to the market, very small, so we don't actually think that there's one geography that will unlock something for us. It's there's an enormous amount of growth we can have from a user perspective around the world.
And that'll take different shapes and rates, but we think it's just a global opportunity. A lot of those learners, to your point, are learning English. And when you learn English around the world, you know, a lot of users do come to us with some proficiency in English, and so what we're trying to do is meet them where they're at, get them into the right level on the course, because now we have more advanced course content for our English learners, which we didn't have prior to 2023. That was a, you know, an evolving exercise in 2023 and 2024 to add a bunch of more advanced English learner content. And in fact, in our shareholder letter, we have a chart of the exponential growth in the courses and content we created. It wasn't just English content, but it was all content.
We did that through AI in particular. Like, you know, we still have humans in the loop checking all that content, but AI has really enabled us to scale that really rapidly with minimal additional cost. We can come back to that. When then you need to tell people or help people understand, if you place them into the right place, they have a bunch of content they can go learn. You need to go remind people or expose people to the fact that they can get pretty proficient on Duolingo in English. The answer to that is still going to be word of mouth. Like, that's not a new modality. Like, the best way for an advanced English learner to learn that they can get advanced on Duolingo is one of their friends or family to have done it and told them about it.
And so it's very seamless into our current marketing efforts. We just have to figure out how to amplify the folks who are saying that they're learning really advanced stuff on Duolingo, and we're really good at that. We're really good at taking influencers or user-generated content and amplifying it. And that's the way you will see us, you know, evolve our messaging or our focus, and then people will go tell their friends and family, and we expect to grow that way.
I could argue that that may not be the case though in terms of 2025 because of Video Calls.
Right.
You guys have a brand new product that, again, unlocks a new thing where you do not have the awareness unless you pay for it, right? Is that different? Is that a bad thesis or how do you think about that component of literally a feature of Duolingo that may not have that awareness?
Well, I think people are paying for it around the world. So if you look at the markets where Super is versus Max, for the most part, those markets overlap. Like, they're pretty similar. So I think people around the world are paying for Max, and they will tell their friends about it. On the call, Luis mentioned Japan. You know, Japan is an outlier. There is more share of Max being purchased in Japan than in other places, but I think that's probably going to be an example of where we'll see this dynamic play out.
Okay, let's flip to retention. Big picture. Like, how do I think about the biggest operational lever you have to be able to improve retention and people coming back to the app? It's super broad, but I'm gonna let you take it wherever you wanna take it.
No, I'm gonna take that for a ride because it's a really fun question. I mean, again, the power of our model is that there's two retentions that people care about. There's user retention and there's subscriber retention. The good news for this answer, to keep it brief, is that the biggest lever we have for both of those is improving the product a little bit every day. In fact, it accumulates so much that if you used Duolingo in 2022 and you came back to it today, you would think it was a completely different app. You wouldn't notice it. The characters would be different, the path would be different. And so that's what happens, you know, we talk about this 1% better every day, and people think it's kind of, well, what's the silver bullet?
The silver bullet is that every day accumulates into something fundamentally different, which leads to better user retention over time. And that is the most powerful dynamic we have. It's just all of the testing we do to make things as trivial as haptics or, you know, additional gamifications or Friend Streaks to get it more social. All of these things add up to the product being that much more engaging and keeping users coming back and retained. And then when you talk about, you know, retention of subscribers, all of those things are at play for subscribers. Because if they're using the app consistently and they're coming back, they're more likely to be continue to be a subscriber. And when we talk about Max, I think some of the questions have been how are you gonna drive Max adoption and Max retention?
And it's the same answer, which is there's an enormous amount of vectors that the team has come up with, to make Max a more enjoyable product experience. Like right now, Max has only got Lily. You're only, you know, doing Video Call with Lily, but that doesn't necessarily need to be the case. We're gonna experiment to see if there's other ways to keep people engaged with Max, which again, I think will help not only user retention, but subscriber retention over time.
So the derivation, if I step one level back of these, like, continuous little improvements, though, is that A/B testing? Is that the right way to think about it? Just simplistically?
Yeah. Now that's how we run the business. Everything is an A/B test.
So, where, where I'm going with this is, okay, so simplistic A/B testing. I see this explosion in content that you guys have also had with generative AI. Understood, coding assistants have kind of accelerated the engineering process. Are you guys seeing this explosion of A/B tests as well in terms of the cadence of what you guys can experiment in 2025 versus, say, 2022?
Yeah, absolutely. I wouldn't, you know, I don't know if you graphed it on a curve, it would look exactly the same, but relative to any year in the past, we ran more A/B tests last year than ever before. That's a combination of more humans. It's a combination of, you know, AI-enabled everything. And it's just a manifestation of the fact that our teams just get better at knowing what experiments to run. So it's number of experiments and then likelihood of success for those experiments.
As you do optimize conversion, right? Is there a low-hanging fruit element to this as well? Or is each A/B test, like the impact of A/B tests from 2022 versus 2025, are they incrementally smaller? Or is it kind of that continuous progress and, hey, you guys don't actually know, so they're maybe the even equal size?
Well, there's two, there's a couple of variables in play here. One is the absolute user size, user base. Yeah. And so on a per user basis, I don't know what the answer is, but on an overall impact basis, they're probably bigger on average.
You guys can test faster.
We can test faster for sure. But you know, I think that the broader point is just that whole process is the magic. It's not any one particular metric of like this A/B team, we're doing this A/B test faster or it's more scaled. It's the idea that when we are approaching these questions, we are coming at it from a let's not guess, let's find the data. So I just pontificated about the fact that maybe my favorite character, Falstaff, he's an angry, pouty bear, which says something about me, I suppose. But why not have Falstaff have a video call? Well, we don't need to debate that. Let's put Falstaff in there and see what happens, right? Now, I'm not saying that's an actual experiment we're running, but I'm saying that's the approach. The approach is that way.
And the last thing I'll say is something Luis has said publicly many, many times, which is every time he has thought that there was going to be some dead end, especially in purchase flow dynamics, because we run a bunch of experiments on pricing, how we surface our subscription to users when we surface it, you know. He's been like, well, at some point, you know, we won't see gains. He's every time he's surprised. There just continue to be gains. And it's because the user base turns over, the app is completely different, you know. There's just all these different dynamics. And so I think we have, you know, long room to run with our process. And I think the most important point is it is just a process. It's called the Green Machine, which we just published in our handbook, which I know you all should read.
It's a great insight into actually how we run the business.
All right. Transition to monetization questions. Just big picture again, talk about the early learnings from Max's rollout.
Yeah. The biggest thing for Max, and this happened in early last year, I think end of 2023, early last year, but really accelerated in the back half of the Video Call, but just the fact that there was so much demand for a subscription tier that was twice the price. That's the first thing everyone should just underline. We were able to say, here's what Super costs and Max is 2x that. Would you like to buy it? And a big chunk of people were like, yeah, yeah, I'll buy that. That makes sense. And then, and they were doing that without really necessarily early last year knowing what they're buying. And then we found this feature in Q3 and Q4 Video Call that really in and of itself communicates what the product is.
So when I tell you all, if you all haven't used, if you haven't used Video Call, just the word or the name Video Call, you know what that is. And when you're describing it to your friend, which is like, I have a video chat with a character, that's what it is. It's not this complicated thing we have to describe to sell it. People get it. And so it's picked up on, on then Q3 and Q4. And now it ended the year with 5% of subs as, as Max subs. We talked about how it relates to English learners. The English learners are using Max more than other language learners for us on our platform. They're doing about twice as many sessions as other, as the next highest language, which is great. That's what we thought it might do.
We had the hypothesis that it would help us grow in English learning, in the English learning market, and so far it is, and the rest of it is pretty early days, so when you think about kind of the financial model of it, we haven't optimized who we show it to, when we show it to. We haven't optimized price. We haven't optimized cost. We haven't optimized a lot of it, and yet already we're seeing that we like the free trial conversion rates. We like the retention rates such that we believe it's our highest LTV tier. Now that'll remain to be seen. You know, it depends on retention and conversion over time, but right now it definitely looks like it.
And so we feel really good about the fact that, you know, in a couple quarters we got from basically a standing start to 5% of subs at the highest LTV tier by the end of last year.
Do you think, well, I'm sure you know, but is it more incremental or do you think that it's more of an upsell? The subscriber, the unit subscriber.
Yeah, it's been great actually to see both happen. We like the fact that both are happening. So we're getting brand new to the platform folks who have never subscribed, who then purchased Max for the first time. And there's a good chunk of our Max subscribers who were already Super subscribers and upsold themselves, or we upsold them.
That's not the question I asked.
What was the question?
The incrementality of the conversion.
Yeah, I mean, that one is tough to parse without going into the logistic regressions that we run. So I'll probably spare everyone the details.
All right. You guys implemented localized pricing like two, three years ago.
Right. Yeah, 2022. Yeah.
Yeah. So what's the next big lever for monetization for kind of lower GDP markets?
So a couple things on how we think about pricing. We do, as you, as you mentioned, in 2022, we, that pricing, was really just a normalization relative to purchasing power. So it wasn't like a crazy, sophisticated change. It was just trying to normalize for basically GDP per capita normalization. And then now we run pricing experiments pretty frequently. We used to, you know, prior to 2022, we weren't running them all that much. And now we run them, you know, a decent amount. Certainly if there's like a big FX change, we think about it, but we're running them pretty frequently in general. And those are price point changes. So like, should the
Super subscription be $84.99 or should be $89.99 or $94.99? Trying to figure out what the right price point is. And I would say on a price point, the biggest opportunity is Max.
It's just early. So we don't know what the right price point should be relative to Super. And we'll keep running those experiments and figuring it out. But in terms of your actual question and developed markets, the biggest opportunity for, you know, pricing power is to get them to do Max. I mean, it's really, and it's something that I, I think we say a lot, which is if you look at our ARPU trends, we've guided to positive ARPU this year. You know, the price point is all, those are all interesting and we need to do those as good hygiene. But the bigger shift that's possible is this mix shift from free to Super or from Super to Family or from those things to Max.
The more we do that, the higher ARPU will go and the more likely we are to be monetizing in developed countries, developing countries.
Put that in the context of AI deflation, right? Like COGS should come down. How do we think about the evolution of what Max can be as your COGS limitation kind of falls off?
It's a great question. The teams that are working on Max right now have just, you know, we talked about low-hanging fruit. There's like fruit lying all over the floor for them to go experiment with. They want to, they have all these ideas. They want to go try all these things to see if we can make it more engaging and see if people can convert more. And because they have no shortage of ideas, we've asked them to focus on those things. If you make it engaging, you make the product all that much better, conversion will follow, retention will follow, we'll be able to price it right. Good things will happen. Eventually we'll pick up some of that fruit and we'll ask them also to then figure out ways to make it cost less.
As that happens, you know, Luis has said his goal is to make Duolingo the most engaging education app in the world, make it much like a human tutor as, as possible, or as much like a human tutor as possible. And so as that cost comes down, you then will see his desire, as he said on the earnings call, maybe in response to your question, maybe you figure out ways to put video as, you know, maybe not Video Call as it is right now, but elements of it in different places in the app. And you know, that's pretty interesting. 'Cause again, if the goal is just driving engagement, then this, then as costs come down, just becomes a tool to do that. It's not necessarily just to drive conversion somewhere in the app.
It's really just to make it more engaging on the fundamental belief that the more engaging the app is, the better the model works, the better it monetizes, and the more free cash flow there is.
Okay. Matt, I think we're at time. Thank you so much for the conversation.
Thank you. Great to see you.