Duolingo's Endless Cycle - Part 2

MS Paint but with AI

In partnership with

The Daily Newsletter for Intellectually Curious Readers

If you're frustrated by one-sided reporting, our 5-minute newsletter is the missing piece. We sift through 100+ sources to bring you comprehensive, unbiased news—free from political agendas. Stay informed with factual coverage on the topics that matter.

Hey guys, it’s Kushagra! Back with part 2 on Duolingo's crazy growth.

This one’s all about stats, tools, and how folks use and keep the product.

Duolingo is good at grabbing data, like for A/B tests, but they weren’t digging deep.

After seeing how Zynga and MyFitnessPal did it, they realized, ‘Why not use Duo’s data to find that one key stat and make big moves?’

From Zynga and MyFitnessPal, they learned to split users by how much they stuck around. Zynga had three big stats for this:

  • CURR (Current User Retention Rate): Will an active user return if they used the app in the last 2 weeks?

  • NURR (New User Retention Rate): Will a new user come back the next week?

  • RURR (Reactivated User Retention Rate): Will a user who left but came back stick around the next week?

Then MyFitnessPal added a fourth:

  • SURR (Resurrected User Retention Rate): Will a user who has gone for a long time return?

They took these at Duo and leveled up, aiming for our North Star state.

With a data whiz and an engineering lead from the Growth Team, we built a new model—switching from weekly to daily views and adding more stats to crack the code.

Data Model

The groups show how much users vibe with the app.

Each user lands in just one group per day.

These groups are MECE (mutual solo, full set), so they cover all Duo users. The arrows track how folks switch groups (like CURR, NURR, RURR, and SURR), but now we check daily, not weekly.

The groups + arrows make a near-closed loop—new users are the only gap.

The top four groups add up to DAU (Daily Active Users):

  • New users – First day ever on the app.

  • Current users – Used today + at least once in the last 6 days.

  • Reactivated users – First day back after 7-29 days.

  • Resurrected users – First day back after 30+ days.

The next three groups are inactive today but have different levels of chill:

  • At-risk WAU – Used in the last 6 days but not today.

  • At-risk MAU – Used in the last 23 days but not this week.

  • Dormant users – Gone for 31+ days.

These groups give us DAU, WAU, and MAU.

We see how moves impact DAU growth by tweaking the arrows (aka rates). These rates are the levers teams pull to boost users.

So, we started tracking daily to see how the groups + rates changed over time. With this data, we could predict trends and test which moves gave us the biggest DAU spike.

We ran a test: bumped one rate up by 2% per quarter for 3 years while keeping the rest steady.

What did we see? Even tiny 2% shifts had a huge impact on DAU + MAU!

User Growth Metric

The Current Users group is the real MVP ‘cause folks who stay keep the streak alive.

Closer look

Betting on CURR was a game-changer. It stacked up over time, making a huge impact. Hard to move, but when it did, it hit hard. That’s when we knew—CURR was the key.

We stopped chasing new-user growth. That was wild for a team that had always put new users first.

Atlasmoth Designs doesn’t just create visuals—they leave a lasting impression. Let’s craft a distinctive identity that captivates and fuels your success.

Fixing the leaderboard

They had to start from scratch as the CURR never moved.

I still believed gamification was the way. Our Gardenscapes-style fail didn’t mean game tactics don’t work—it just meant that one sucked.

New plan? Leaderboards.
Duo had a basic friends-only leaderboard, but it wasn’t working. From Zynga, they learned that users compete with matters way more than just seeing friends’ scores.

In FarmVille 2, they created leagues—Bronze, Silver, and Gold—so players weren’t just fighting for the top each week; they were leveling up. This kept players hooked.

We figured the same would work for Duo—mix progress + competition and watch retention soar.

Leaderboard

Not all of FarmVille 2’s leaderboard tricks worked for Duo—we had to tweak it for our vibe.

In FarmVille 2, you had to do extra tasks to climb up. We skipped that. More steps would just make learning harder. So, we kept it chill—users joined auto and moved up just by sticking to their studies. Fun, but easy. Nailed it.

This blew up (in a good way). Learn time +17%. Super users tripled. We didn’t have CURR data yet, but D1 and D7 went way up. It was clear—leaderboards worked. And it was the Retention Team’s first big win!

Push notifications

The team wanted more ways to keep folks hooked. So, they looked at push notifications. The duo had seen some wins here before, but the hype stalled.

Groupon’s CEO said: They started with 1 email a day… then 5 a day. It worked—for a while. But users got sick of it, and boom—emails stopped working.

They did not want that to happen, so they made one rule: no spam and no extra pings unless the CEO said yes.

With that in mind, the team went all in—tested times, copies, images, and more. That made DAU grow year after year.

A meme about Duolingo’s “pushiness” that went viral in 2019

Streaks

The Retention Team was still hunting for ways to boost growth. Then, their APM found a gem—users with a 10-day streak were way less likely to drop off. Some of it? Luck. But still, it was too good to ignore.

Streaks are simple—just count the days a user sticks with Duo.

First big win? The streak-saver notification. If a user was about to break their streak, they hit them with a late-night ping. It worked.

After that, Duo kept stacking Ws—calendar views, sick animations, streak freezes, and rewards.

Streaks

Streaks are now one of the best ways Duo keeps users hooked. People flex their streaks all the time.

Why do streaks work? The longer you go, the harder it is to stop. Every day you come back, you care a little more about keeping it alive. That’s huge for DAU.

Big wins don’t always come from wild new ideas. Fast tweaks on what already works can be just as huge.

They knew CURR would hit a cap. So, they decided to build more squads to try new things—new countries, social stuff, more courses, influencers, schools, ads, and going viral on TikTok.

In 4 years, CURR went +21%, daily churn -40%, and DAU 4.5x’d. Now, half of DAU has a 7+ day streak, meaning they’re more active, and more likely to stay, share, and sub to Super Duo. This hyped up their IPO.

Hope this sparks some ideas for you. If you try any of this, tweak it to fit your thing. Don’t just copy—trust your gut. Happy grinding! 🚀

Missed part 1? Go check it out!

AI transforms Microsoft Paint—Copilot redefines creativity.🧠

This track gave me a serious boost—check out ‘Camel’ by Jungle By Night🎵

A rare interview with TSMC’s Morris Chang—untold tech history, thanks to Jensen Huang.📱

"Simplicity in design is clarity in thought—both make life more beautiful."

Thank you for fluttering by and reading our newsletter! If you found it as captivating and engaging as my vibrant wings, please share it with your friends and colleagues. Help us spread the excitement and keep the gamified energy soaring!

Reply

or to participate.