How Duolingo Built a $1BN/Year "Free" App | Sub Club tl;dr

0:00Duolingo's founder famously promised the app would be free forever. So, how did they become the monetization powerhouse they are today? It really started when they hired my guest Cem Kansu to do one job. It's funny if you launch if you go watch and I think these talks are still out there. Our co-founder CEO Luis gave these talks for Duolingo's launch and he said we will never have ads, never in-app purchases and never subscriptions. And those are the exact three things that I built in that order. Basically, when I joined, the mindset shift had moved to Luis's like first thing he told me was like, "We hired you. Your sole job is to figure out how to monetize Duolingo." The core

0:41challenge is a tricky one. You have to monetize, but you can't kill the magic that got you those users in the first place. Cem's answer gets at the core of what makes freemium so powerful, but also so easy to mess up. and it gets at what he calls the premium trap, the overoptimization focused on short-term results that ultimately kills your long-term growth. A lot of premium apps fall into this premium trap where the easiest win you can always do is like, oh yeah, let me take some free just make it paid and immediately all your metrics are green and that does kind of work for 6 months or maybe a year, but eventually it stops working. The other thing

1:18it does is it creates opportunity for someone else to do it for free. If we were to tomorrow wake up and say, "All right, we're gonna charge for all of our great features." Very likely in two years, another great app will do it for free. So, how did they introduce ads without a user revolt? It's easy to get overfocused on negative feedback. I did this quantitative analysis in 2016 because we were at this pivotal moment of like, do we launch ads or do we not launch ads? And we looked at, for example, the number of complaints about why we didn't have the Finnish language on Duolingo was 3x the number of people that are complaining about Duolingo having ads. So that

1:54starts putting it into perspective of like every negative comment as a product builder hits home really hard. Like everything looks like the world is falling apart. That's why it helps to like look at the broad picture of like all of our issues as a product and all products have many issues. Cem had a really helpful way of looking at it. You have to zoom out to put things into context. But with that broader context, you do still have to zoom in on the minutia. Sometimes, Duolingo runs hundreds of A/B tests and some of the smallest changes have had a surprisingly big impact. We went from I think start my free trial. So, I think we said something like

2:30uh subscribe uh that was that way. That's your baseline. And then we tried many things, but one of the big up list was start my free trial. That was very good. And then we tried what did we say? Try for free. That was better. And then we tried try try for 0.0. Uh, weirdly that is the string that I think hacks your brain to really grok that that button is entirely free. 0.0 like two 0 decimals. Now I think that's what's live in the app if I'm not mistaken. And that performed really well. That's such a great reminder that conversion is so much more emotional than it is

3:06logical. But even with hundreds of A/B tests, the hardest decisions come when the data conflicts with your long-term strategy. This is really the art of product management. The hardest part is taking a loss in an experiment is not intuitive. If you only always go by positive revenue metrics, you will burn yourself to the ground. We have this operating principle that we I I love and we repeat a lot is take the long view. Meaning you can make business decisions very differently if your goal is to be a very profitable or a very fast growing company for two years versus if you want to build a company for that will be around in 100 years. Before we wrap up, I want to share

3:46one last story from the episode. It's a classic lesson. Be careful what you build even when users are screaming for it. For years, people begged Duolingo for human tutors. So they built it. a lot of our community doesn't remember or don't know is before video call. Our first attempt at like a big speaking feature was a human-based tutoring product. You would press a button, you would connect with a Spanish tutor on demand. No booking required. Honestly, it was pretty

4:15awesome. It had this one massive problem which is uh no one wanted to do it. It was really scary. Duolingo imagined this super fun experience and we would tell you, "Hey, you are going to connect with a human. You're going to practice Spanish." People would press it. They would dial in. We would see their faces come online and they were like, "Oh [ __ ] get me out of here." Because one, it's really scary to speak in Spanish, like on the on the language that you're learning, and two, there's another real human on the other side. That was extremely jarring. We iterated a

4:48lot to ease you into it, but eventually we gave up because humans just did not want to do that. And actually there's this brings me to another hot take that I do want to mention. Do not take user feedback at face value. For years this is what people asked they asked us to build. Like we're like please build me a system where I can talk to a human tutor. We built it. No one used it. So to recap, protect your free experience. Quantify user feedback. Remember conversion is emotional. And always take the long view. There's so much more in the full episode,

5:20so be sure to check that out. If you enjoyed this, please like and subscribe for more.