How Duolingo Aims to Diversify Beyond Language Learning

BRIAN KENNY: In my formative years, I had the good fortune to attend schools that required me to take a foreign language, which was not the norm in all schools. Over eight years of middle school and high school, I completed four years of French, four years of Spanish, and one year of Latin for good measure. And after dozens, perhaps even hundreds of hours, conjugating verbs and memorizing vocabulary, I can say with confidence that I hardly remember a thing. Mon Dieu, you say. Well, I’m in good company because less than 20% of people who learn a language in high school retain much of anything beyond five years. It’s hard to learn a new language and even harder to maintain proficiency if you’re not speaking it regularly. But what if it were fun? What if you actually looked forward to language lessons?

Today on Cold Call we welcome Professor Jeffrey Rayport and Nicole Keller to discuss the case, “Duolingo: On a Streak.” I’m your host Brian Kenny and you’re listening to Cold Call on the HBR podcast network. Jeffrey Rayport’s research focuses on growth stage technology ventures and how to scale them. He is an expert in e-commerce and is credited with coining the term “viral marketing.” I didn’t know that, Jeffrey, until I reread your bio today. That’s pretty impressive.

JEFFREY RAYPORT: Back in my misspent youth. What can I say?

BRIAN KENNY: It’s great to have you back on Cold Call. Nicole Keller is assistant director at Harvard Business School’s California Research Center and a co-author of the case and Nicole is also a graduate of Harvard Business School. Great to have you back on the show as well.

NICOLE KELLER: Thank you for having me back. I think the last time I was on Cold Call I was on the very back end of COVID and my voice was very raspy, so it’s nice to be on again today with my real voice.

BRIAN KENNY: I don’t remember that. I just remember your very thoughtful insights that you shared with us that day. And that was about the Angel City Football Club, which was a really fun case to talk about. If people want to go look that one up, it’s still available. And today is about something completely different but I think equally relatable. Duolingo is as hot as it gets in apps. They’ve been around for a while. They have built a tremendous reputation for themselves in the space of teaching people not just how to learn languages, but how to enjoy doing it. I downloaded the app. I have not yet attempted. I’m trying to figure out what language I want to learn. I’m not going to go for French or Spanish. Clearly those didn’t work out for me.

JEFFREY RAYPORT: And you weren’t thinking about English. There’s always the opportunity to brush up those skills.

BRIAN KENNY: I hope you appreciated my “mon dieu” because that was a pretty good French accent even though it’s been a while.

JEFFREY RAYPORT: Oh, it sounded excellent. I was going to say “sacré bleu.”

BRIAN KENNY: You should have. That would’ve been great. Okay, let’s get started. Jeffrey, I’m going to ask you to start by telling us what the central issue is in the case and what your cold call is to start the discussion in class.

JEFFREY RAYPORT: We start the class by asking students about growth. So this is a course called, “Scaling Technology Ventures.” As you said, it’s about growth stage in the tech sector. And of course this is a classic tech business. They do business by fielding a mobile app. Clearly businesses have a lot of ways to grow. They can span geographic footprint, but maybe the most commonplace one and common to our listeners is product or service line diversification. In this case, Duolingo, as you just said, is widely known for language learning. Duolingo can teach you more than 40 languages. And Brian, I must say that despite your educational history even you might be able to learn a language if you put this app to work.

BRIAN KENNY: I’ll put them to the test.

JEFFREY RAYPORT: Absolutely. So we are asking about the specific moves that they’re contemplating into subject matter adjacencies, one being math and the other being music. And asking the question, is this the way to grow? And implicit in that is the question of is there headroom to grow simply as a language learning app, is it necessary to move into adjacent academic subjects to do that? And how credible would moves like that be to their a hundred million monthly active users?

BRIAN KENNY: Nicole, let me turn to you for a second. Duolingo has grown from a free language platform into a $15 billion company. They have a lot of different revenue streams. What are the key strategic decisions that allowed them to make these transformations?

NICOLE KELLER: Well, I think first and foremost, Duolingo was an early adopter of analytical AI. Their AI system is called BirdBrain. Everything at Duolingo has a fun name or character to it. So this is BirdBrain.

JEFFREY RAYPORT: I forgot to bring Duo. I was going to bring you a plush animal so we could have him as inspiration on the table.

BRIAN KENNY: That would’ve been great. For a photo op too.

NICOLE KELLER: Yes. Another time.

BRIAN KENNY: I can still run back to my office.

JEFFREY RAYPORT: It’s all right.

NICOLE KELLER: But their BirdBrain system knows exactly the words you got right yesterday and the grammar that you got right, and the grammar that you got wrong. So it would know that you got la leche right yesterday. It’s not going to review that again with you tomorrow when you go back on the app. And it’s going to know exactly how hard and how easy to make the exercises that you’re given so that it’s very motivating. So if it’s too hard, you’re going to be demotivated. If it’s too easy, you’re going to think it’s boring. So their analytical AI allows them to just find that perfect sweet spot for your learning. So that’s one thing. Another thing they’ve done super well is they’ve really leaned into gamification and they’ve really taken a page out of the gaming apps playbooks and they use experience points and they use leaderboards and they use hearts. You lose hearts if you get something wrong. So watch out for that Brian. And you can earn them back again through various things. But all of that makes the learning whole process very fun. And so that’s another thing that they’ve done very differently and streaks are very important.

BRIAN KENNY: Tell us about the streaks.

NICOLE KELLER: So the streak means how many consecutive days you’ve been on the app. And people are very into their Duolingo streak, so much so that they post it on social media and they tell their friends about it. So streaks are a very big deal. And actually on Duolingo right now, 20% of their users have a streak of over 365 days. So that’s seven million users who haven’t missed a day on the app in 365 days.

BRIAN KENNY: It reminds me of Wordle and Peloton and it’s tapping into the competitiveness that most of us have if you tap deep enough. Do they have psychologists on staff? How do they understand the psychology of these things and how does it factor into the way that they retain people on the app?

JEFFREY RAYPORT: The fascinating thing about Duolingo for Nicole and me is that this is of course the world’s biggest ed tech app. We don’t tend to think of it as ed tech because it’s so incredibly entertaining and there’s a lot of evidence that if you ask who does Duolingo compete with, it’s not necessarily Coursera, Udemy and Khan Academy. It’s TikTok and Instagram. Having said that, despite the fact that it’s education and entertainment, this is a company run by two computer scientists, Luis von Ahn and his student, Severin Hacker no longer a student. They are now CEO and CTO. Are the co-founders of the company. So what’s remarkable is that this issue of how to build engagement is something they have turned into a massive mathematical or quantitative exercise. They use something called a Markov model, which is a very interesting way of representing multiple variables that have dependencies on one another. But the model assumes that the present state is what it is. There’s no need to worry about where it came from and how it got there. The question is how do you move forward? And on that basis, they’re looking for these interesting high leverage relationships between variables. And it’s that focus as technologists and as data analytics guys that they have managed to create this incredible level of engagement. In fact, Severin Hacker, whom we worked with to put together this case, referred to the app as a motivation engine. They have about 15 billion lessons a week that they are analyzing. So it’s a massive amount of visibility into user behavior. And on that basis they can then fine tune it to create this habit formation, which is quite remarkable.

NICOLE KELLER: And just to add to that, they even know what time you were on the app yesterday. So in terms of helping you build the streak, or maybe it’s helping them build the streak, depending on which way you think about it, they will send you a notification right around the same time that you were on the app yesterday to remind you it’s time. If you want to keep your streak up, let’s go. So this AI system is all designed to keep you using the app.

JEFFREY RAYPORT: And Brian, you mentioned the question of what role does psychology play? And what is fascinating is that of course they have psychologists and data scientists on staff. But again, it is so interesting because it’s driven from these metrics that feed the AI. So just an example, depending on who you are and your personality profile and your behavior and whether you’re on a streak or not on a streak, when you continue a lesson, you might get a button to click on that says continue as in go to the next lesson. But you also might get a button that says, commit to my mission. These are obviously two very different ways of influencing behavior. Duo of course is the famous green owl that is the mascot of Duolingo. And in some cases the motivating statement from Duo is that if you don’t complete your lesson, Duo will be mad. Another one be if you don’t complete your lesson, Duo will be sad. And in another one Duo is reminding you to come back tomorrow and keep doing your lessons. In that sense, it is a very sophisticated system that in a sense has automated the psychological understanding of the individual user across a hundred million users to create motivation and a regular steady drumbeat of engagement.

BRIAN KENNY: Yeah. So it sounds like they are delivering on the promise that we’ve heard so much about AI, where AI is going to find ways to serve things to you before you even know you need them. And what you’re describing, I think the case talks about micro-optimizations. Is that the same idea?

JEFFREY RAYPORT: That’s absolutely right. Absolutely right. And on every level. I think it’s interesting that tech companies, when they talk about this micro-optimization, we hear always the phrase AB testing. We always think about it as the e-commerce platform, which has a buy button, and the buy button could be green or it could be red. Which color gets the higher conversion to purchase? In this case, it’s so much more nuanced because it’s not just about the attributes of a digital interface, it’s about the nature of the messaging, the emotional tone, passive versus active, all these things that are attuned to a unique student’s or user’s personality.

BRIAN KENNY: Super impressive. Nicole, I want to talk about their social media, the way that they approach their marketing. Describe a little bit how they go at that and how it allows them to compete on the same level as TikTok.

NICOLE KELLER: Duolingo is known for its fun and quirky personality. So they’ve got the brightly colored graphics and they’ve got their cast of endearing characters and then they’re known for peppering in these almost silly, ridiculous sentences into the exercises that you do. And they’re so fun and silly that people will go ahead and repost those then to their social media account and that gets them visibility. And you know how effective that is when you realize that 80% of Duolingo’s customer acquisition is organic. So you can only imagine the positive economics that come out of that. And then the other thing they’ve done really well is they’ve really leaned into TikTok, and that’s a fun story. They had a social media coordinator who was just 23 years old when she started. Zaria Parvez. And at 23 she’s probably using a lot of TikTok and she thought to herself, Duolingo should probably be on TikTok. Can I give it a try? And one thing we realized in the case is that Duolingo really has a nice culture, a really positive culture about allowing innovation to bubble up from within. And so they said, “Sure. Give it a roll.” So she actually became the persona of Duo the Owl, and she has built Duolingo’s TikTok’s presence to almost 11 million followers. So pretty amazing that she was able to do that. And she continues, and Duolingo continues to do crazy social media antics. And actually the day that Jeffrey taught the case in February was the day that Duo, the owl died and it was posted on X that Duo, the owl has died.

JEFFREY RAYPORT: This was very upsetting to us. It was of course devastating for the students.

NICOLE KELLER: Yeah. And the reason that was posted at that time was that people hadn’t continued their streak. And so Duo was disappointed and Duo died. So even on your phone when you opened it up around that time, Duo the owl had Xs over its eyes because it was dead. And then they took it a step further a few days later and they said that Duo had been hit by a Tesla cyber truck and anyone with information was meant to come forward.

JEFFREY RAYPORT: This became even more upsetting because it wasn’t just Duo the owl that died, but it turned out Luis Von Ahn recorded a video that was released, pushed through social media channels in which he delivered the sad news that not only the cyber truck killed Duo, all the other characters were dead too. And they’re like a dozen of them. Just to make sure that we haven’t upset the listeners of Cold Call all of them have come back to life starting with Duo. So it’s all worth safe.

BRIAN KENNY: Thank goodness.

JEFFREY RAYPORT: Why is this company such a valuable public company? When you can acquire customers and 80% of your acquisition is word of mouth as a result of viral distribution that drives down your customer acquisition cost quite considerably. Our estimate from the case is it’s about two and a half bucks. The average e-commerce site might pay 35 to 60 dollars to acquire a customer, and you have Duolingo acquiring for two and a half dollars for customers that ultimately have a significant amount of economic value to them. And it’s partly because they not only join based on a referral with a higher level of conviction, but they stick around because that’s what streaks are all about. And all these behavioral economics in effect that we’re talking about built into the app.

BRIAN KENNY: They’re not the only firm that’s doing this. There are other language education platforms out there. What are some of the differences? And I guess what I’m wondering is would the criticism maybe be that Duolingo is not taking this seriously enough. This is too gamified and you’re really not going to learn much, but it’s fun to engage with from time to time?

JEFFREY RAYPORT: It’s interesting that you say that. We encountered some limited skepticism in the classroom with the two sections at which we taught the case where a few students were saying, “Listen, I’ve been on a streak.” The app is really fun. I learned a little bit, but it didn’t teach me how to speak the language. And in that sense, my retort was for all of us and your story at the beginning of this podcast, a perfect illustration of it, all of us studied foreign languages in middle school and high school, and how many of us actually came out of those classrooms speaking? That’s a very high bar, and we should talk about that because one of the ways in which they’re using the new flavor of AI, meaning not analytical AI, but generative AI, is actually to cross that line into something that would actually create fluency and language proficiency in a way that I would argue would be very tough to do in a classroom or with conventional software-based learning.

NICOLE KELLER: So they have Lily their video call. So when you finish a lesson and you’ve gotten … And I’m fairly early on in Spanish right now. But it offers up that Lily will do a video call with me, and so we can chat and we can speak in Spanish. And I think the idea there is to obviously really practice the language in a very low-risk way. I am more willing to practice with Lily on my phone than I am with a Spanish-speaking person because I don’t want to mess up. But I’m okay messing up in front of Lily. It’s a chatbot.

BRIAN KENNY: Stakes are low.

NICOLE KELLER: Stakes are low. I actually do think this will help them cross the chasm and help people get fluent. I think this new way of using AI will help them take their language learning to the next step.

JEFFREY RAYPORT: It is worth saying that yes, there are absolute critics of Duolingo to your point, who say, “Listen, if you’re serious about language, you go off and study Rosetta. You do not use a lightweight mobile app that is so incredibly entertaining and even addictive.” I think we would argue as fans of the company that if you go all the way back to the mission that Luis and Severin articulated at the beginning of this story back in 2011 when the company was launched they set out to execute on the following mission, “to develop the best education in the world and make it universally available.” We are looking at this and very impressed by the fact that maybe the future of education has a lot to do with gamification and as opposed to gamification, essentially indicting it for being too frivolous maybe in an attention scarce world where we’re all dealing with too many information streams at any given time, including in classrooms, if students have digital devices. Maybe this is a very interesting human factors experiment on a hundred million humans on how actually to get education done for certain kinds of subjects in a completely modern way.

BRIAN KENNY: Well, that’s a good transition to where they’re trying to go. So at the best education, you didn’t say the best language education, you said education and the case talks about them thinking about pushing into some other areas, music being one, math being the other. If they could find a way to make math fun, I think that would be an enormous breakthrough. But talk a little bit about how do you extend the success that they’ve been able to have with the personality of Duolingo into other educational subjects?

JEFFREY RAYPORT: Well, one thing that’s interesting by way of a screen is in talking with Severin Hacker, one of our questions was could you teach history? Could you teach literature or could you teach somebody to become a neurosurgeon on an app? And he had what seemed like a very credible and persuasive response, which is, look, we set out to create something that went well beyond language instruction. So your point is exactly right, which was this is about creating a better kind of education. But they had no illusions about the idea that this would fit everything. So a couple key conditions. One is that you could learn it on a mobile app. So we talk about, say, learning how to do brain surgery. It’s unlikely that you could do enough reps on a mobile app to figure out it. So clearly that would be impractical.

But the other one was that this is fundamentally about subjects or academic areas where you can learn through interaction or learn by doing. And so rote repetition, series of exercises, stimulus response type educational processes. And so they believe that while history might be off the table as well as neurosurgery, that music and math actually fit those requirements quite well.

BRIAN KENNY: I think there was actually a line in the case where he may have said that those are languages in some ways. I think that’s a really interesting insight because people who are mathematicians have described it in the same way.

JEFFREY RAYPORT: Absolutely. And they’re like languages. They’re skills acquisition. You Brian are a musician and they’re doing things like note identification, ear training, site reading. For all of us who’ve spent time trying to learn musical instruments those are all core fundamental skills that you put together in order to create music. And they believe through rote repetition and interaction with the layer of motivation created by the elements of gamification that it can actually address a learning agenda in a quite serious and substantive way.

NICOLE KELLER: And there’s another language out there of coding. And a lot of people think, is that something that they could also move into? Because coding is very much like one of these subjects and you can probably learn it in a similar way. And so that’s an exciting opportunity for them too. They haven’t tried it yet but they might.

JEFFREY RAYPORT: This is a fascinating company for us to have written this case about for one other big reason we haven’t talked about, which is that many people worry that generative AI is some kind of death star that will crater a lot of businesses and especially ones that look like this, where they’re teaching things where, for example, if Gen AI can do coding that then takes a beautiful, shiny new line of business, as Nicole just said, and puts a considerable cloud of doubt over it. But even go back to language. A lot of people have believed that because of the existence of Google Translate built into our smartphones, if you choose to deploy it, who needs to learn a foreign language again? And this is where it’s interesting, which is that this is a company that has been fueled by generative AI rather than the opposite. I tend to think and share this with the students that to some extent, this is a very dramatic right now “tale of two cities” in the capital markets. And what I mean is if you put Duolingo side by side with another ed tech platform called Chegg, which every one of our MBA students knows because Chegg helped them get through high school, college, maybe even business school with Chegg study and all those tutors and the database of a hundred million answers. It’s a very interesting story. You have Duolingo that went public at a roughly three and a half, $4 billion valuation, and last time we checked prior to the recent market meltdown, it was 70,80, 90 billion. It doubled in value just during the time we were writing the case. Meanwhile, Chegg from the advent in November 2022 of ChatGPT the introduction from an open AI of GPT model 3.5, Chegg is down from its market high by 96%.

BRIAN KENNY: Wow.

JEFFREY RAYPORT: Why is it down? Because there was considerable concern that that database of a hundred million answers was now obsoleted by a Gen AI that could provide answers for everything.

BRIAN KENNY: Yeah. Nicole, let me come back to you because the case references the super app approach that Duolingo is considering. Can you talk a little bit about that and how it factors in?

NICOLE KELLER: Yeah. They did initially launch math as its own app, and then they ultimately decided that they were going to move to a super app model and this was a subject of a lot of debate in Jeffrey’s classroom but ultimately I think they decided that one of the reasons they had to move to a super app was because of discoverability problems. There’s 2.2 million apps on the Apple app store. There’s three million apps on Google Play, so it’s just hard. It’s hard to break through that clutter even for a brand as well known as Duolingo. It’s still hard. It was hard for them to break through that clutter. So that was one reason. Another reason was friction. When you get a new app, you have to download it, you have to set up your username and your password, and then you have to log in and all that takes time and time is friction on one of these mobile apps. If you’re on the main app, those Duolingo users are already signed in ready to go, they can start using it right away. So they decided that was another reason to go with the super app idea. But then also from a cost standpoint. The whole idea was to replicate the gamification features like leaderboards and the streaks and the experience points. And so ultimately, is it better to just build that once and be able to apply that same model to all these different adjacent subjects, or do you want to have to pay for your engineers to keep building that over and over and over again? So from a cost standpoint, I think they decided they would get more economies of scale if they just kept it all in one super app. But I think we talked about in the case and in the class that there are some risks to the whole super app approach because these new subjects are new. It’s taken them 12 years to get the languages to where they are today, and they’re really good. They’re really high quality, but math and music are new, and they will admit they’re probably at that quality standard yet. And so what happens if you get a user who tries it out and then says, “That’s not really what I was expecting,” or, “That’s not up to Duolingo standards,” will they go back? So I think there are legitimate risks to the first time user experience that they’ll have to navigate.

BRIAN KENNY: Are there also brand risks here? How far can you really push? How far should you push? This is probably an existential question for entrepreneurs anyway, and you look at a lot of these ventures. What is reasonable for them to expect to be able to do, or should they just stay in their lane?

JEFFREY RAYPORT: It’s a great question because of course Duolingo has both second language and language built into it. That’s the essence of the name. It reminds me a little bit of the brand constraint that TV guide had in the late ‘90s when they tried to become a guide to the internet, and the problem was their brand name was TV Guide. That was a problem and another existential risk. The flip side of that argument though is that in the same way that people will say, this is the Uber of X to talk about, say, a mobile-based sharing economy app, it is commonplace in the business world to talk about, I’ve got a great concept, it’s the Duolingo of X. And that has become so widespread that Saturday Night Live even did a parody of Duolingo for talking to Children, which was one of their parodies and actually very funny skit. So it may well be that they built enough equity in the name that again, is associated not just about language, but about entertainment, about motivation, about improving yourself, and that has then become something that’s transcended the underlying meanings of either Duo or lingo, And thus is not a brand constraint, and that’s certainly what they’re betting on.

BRIAN KENNY: Nicole, you described as talking about the social media approach and the personality that they’ve developed. It sounds to me like the culture there must be very entrepreneurial, that they give people the freedom to experiment and do these things with the brand. At one point, does the AI perhaps become a tension there because the AI is doing its thing? Does it impede their ability to continue to shape the brand and make it funny and fun and personable?

NICOLE KELLER: Yeah. I think that’s a real challenge for Duolingo. In the past, all of Duolingo was built by experts in learning science and teaching and entertainment who really could ensure that Duolingo maintained its signature style while also being a very effective learning tool. So now that AI is going to be building a lot of the content, Duolingo is really going to have to train it and monitor it so that it retains Duolingo’s voice. And they’re confident that they can do that, but they are putting guardrails on it. If you’re on Lily and you start to go astray and you’re not talking about ordering your coffee at the local cafe, it is going to guide you right back and get you onto appropriate subjects if you are trying to take it somewhere, it shouldn’t be going. You can’t go rogue on Lily. And there’s also a bridge too far. There are some AI features that Duolingo decided were just a bridge too far. One of them was that they were looking at a feature that would clone a user’s voice to create audio of them speaking, and they finally decided that was just simply too creepy, and they decided not to do that.

BRIAN KENNY: Yeah. Yeah. We’ve seen that done in other places, and it is creepy.

JEFFREY RAYPORT: It is very creepy.

BRIAN KENNY: It probably should be against the law. But Jeffrey, you mentioned Google Translate before. Why would anybody go to the effort to learn a language when the technology that allows you to translate is only going to get better and better?

JEFFREY RAYPORT: There is a fascinating answer to that, which is that Duolingo may teach more than 40 languages on that app, but the primary benefit to acquiring a language is not about 40 languages, it’s about English. It turns out that of the eight billion humans on earth, a solid two billion are acquiring a second language. And of that 25% of the human population on this planet, of that 1.4 billion, meaning 70% of them are acquiring English. It’s interesting too. So why would so many people want to learn English? And one of the things that we uncovered in this process is that those who acquire English effectively as a second language increase their earning power by 30 to 50% over the next chapters of their career. If you think about what that really means, if I’m now English proficient in a way I wasn’t three years ago, English proficiency is not showing up for a meeting with you, Brian, holding Google Translate operating on my smartphone. It’s actually me talking.

So what’s interesting is that yes, all of these ways in which AI or generative AI could become a death star for the business seem to have done the opposite, which is that one of the benefits, in addition to what Nicole was describing with being able to have what seemed like natural language conversation with an avatar named Lily and take you beyond just rote classroom learning. Another part of it is that their ability to go into these adjacent subjects without massive cost in the development of content and materials is that they’re using AI to do it. Including one of the concepts they had a few years ago was, gee, people would really do well learning a language if they could listen to a podcast in a language they’re learning. People love podcasts. We know that because you have such a huge visitorship here at Cold Call.

BRIAN KENNY: Did you hear that listeners? You love podcasts.

JEFFREY RAYPORT: Absolutely. Humans are genetically engineered to love podcasts. So they thought, wouldn’t it be great if you’re studying French you’d have a podcast in French. The estimate for how long it would take to build that was about five years plus a very large amount of deployed capital. In the age of generative AI they could do that in a few months. And there’s now something called Duo Radio, which is a podcast in all the languages Duolingo will offer within the app. So it is very interesting to think that some of it is about avatars and being able to do a bunch of very interactive things, but some of it is also back to this issue of how do you make a business organization filled with brilliant human talent even more productive than they already are? And it’s very clear that some of their ability to explore new areas and develop new features is based on the productivity gains of putting Gen AI to work.

BRIAN KENNY: We’ve heard about Khan Academy, we’ve talked about a lot of the competitors that are in the space today. Duolingo is going to have to continue to find ways to innovate, to differentiate themselves from these other groups. Adjacencies might be one of those ways. But did you talk with them at all about how they’re thinking about the future and what that might look like?

NICOLE KELLER: Yeah. If you look at the competitive ed tech landscape, it’s very crowded and you have both subject matter experts like Prodigy for math, Musician for music, and then there’s a whole host of other more general ed tech players, like you mentioned, Coursera and Khan Academy. So on one hand you can look at that and think it’s pretty fragmented. No one’s actually figured this out. Is this even a winnable market? But I think what we’ve talked about today is that Duolingo’s secret sauce is its gamification, which makes learning fun and sticky. And most of the other competitors that we’ve looked at don’t use gamification. The other players are emphasizing learning over fun. And I think Duolingo has this unique approach of focusing on the fun and you learn in the process. So that makes it unique. And I should also note, I think we mentioned it before, but they don’t feel that it’s competing against other language apps like Rosetta Stone. Duolingo really feels like it’s competing for your free time that you spend on apps like Instagram or X. So in a way, they’re not looking at the competition as being other ed tech players, but just other ways you spend your time on your phone.

BRIAN KENNY: That’s a great insight. In fact, we did the Candy Crush case a while back Jeffrey, if you remember. People spend an inordinate amount of time on their phones playing games and things, so this seems to play right into their hands.

JEFFREY RAYPORT: Yes. And it’s interesting that the psychological appeal of getting the same snackable, highly addictive entertainment, but for something that feels worthwhile. It’s a very interesting differentiator for them.

BRIAN KENNY: Yeah. So let me give you the final word in this, Jeffrey. As we think about the challenges and opportunities that lie ahead for Duolingo, what do you think some of the lessons are that other tech companies can take from what they’ve been able to do?

JEFFREY RAYPORT: From my point of view that we’ve learned so much from this case, so I don’t want to minimize it, but maybe it’s two big things. One is as we talked about before, we live in an attention-scarce economy. And so I increasingly think that the challenge for every company that fields digital devices, apps, screens, every company is competing in a world where because attention is scarce, engagement is harder and harder to come by. We know that what creates valuable companies is customers who not only buy once, but come back over and over again. And it strikes me that there is a very profound lesson here about the power that goes beyond gamification of essentially using data, analytics, artificial intelligence, to figure out how to maintain connection with the user. And what Duolingo is demonstrating as laboratory for the world is, I would argue, valuable for any business that is thinking about competing again in any sector of the economy, which has too many competitors, too many offers, and too much choice.

The other is that the other is something we haven’t talked about is an angle on this question of why move into math and music to complement languages. Is of course what they’re doing is they’re creating not a singular educational offer, but a bundle. And it always reminds me of that famous quote from Jim Barksdale, the co-founder of Netscape, one of the early web browsers said, “There are only two ways to make money in business. One is to bundle and the other is to unbundle.” And it’s a very profound statement, which we could devote a separate podcast episode to. But clearly, one of the things that they’re betting on is that this is what in the gaming world is referred to like king as a freemium model of our free to play model, meaning that all of us can use Duolingo without paying for it, but about 8% of their monthly active users choose to pay. The question is when somebody becomes a paying customer, that’s a beautiful thing because that means eight million people are subsidizing 92 million people who get to use it for free, and they’re accessing tiers of services and power tools and the kinds of AI applications we’ve been talking about. Clearly those are a lot of benefits.

But the other benefit is it’s a bundle. And so part of the logic here is that like Netflix, which has a super abundance of programming that comes into a household where on any given day, if there are four or five members of a family, four may have tuned out and maybe watching Hulu, but one of them is following a series on Netflix. And as long as there’s one person in the household who still wants to stay connected, that subscription doesn’t get canceled. And so one of the ways in which Severin Hacker talks about the new multi-subject Duolingo is that the aspiration or the ambition is that it becomes what he calls a forever app. That it is a utility like electricity, heat, whatever we have in our homes. This is a knowledge utility that is sufficiently engaging that actually gets the education job done, and it’s a forever app because there’s always something somewhere on this app that will appeal to someone who’s on a streak living in any given household.

BRIAN KENNY: That’s great. That’s a great vision. Nicole. Jeffrey, Merci beaucoup.

JEFFREY RAYPORT: Merci bien, Brian. Is a pleasure to be here.

NICOLE KELLER: Thank you so much.

BRIAN KENNY: If you enjoy Cold Call, you might like our other podcasts, After Hours, Climate Rising, Deep Purpose, IdeaCast, Managing the Future of Work, Skydeck, Think Big, Buy Small, and Women at Work, find them on Apple, Spotify, or wherever you listen. And if you could take a minute to rate and review us, we’d be grateful. If you have any suggestions or just want to say hello, we want to hear from you, email us at [email protected]. Thanks again for joining us, I’m your host Brian Kenny, and you’ve been listening to Cold Call, an official podcast of Harvard Business School and part of the HBR Podcast Network.