Duolingo's Endless Growth Cycle - Part 1

Yo peeps, what’s up? Duolingo’s all the rage run.

Caught a podcast with their CPO and got some fire deets to spill!

In this edition, I’ll spill on the Ls and Ws on Duolingo: launching leaderboards, boosting push notifications, and making the Streaks feature addictive.

This edition is part 1 of the lingo’s growth strategy.

A few months back, Jorge Mazal (the former CPO of Duolingo) discussed how Duo reached 4.5x growth for a product that was already big.

Just a few smart tweaks and a dope growth model flipped the game.

Back in 2017, joined as Head of Product, Duo was killing it as the top ed app with millions of users.

But by mid-2018, growth was meager. DAU was crawling in single digits, and investors were pressing for aggressive moves.

Phase 1: Increasing gamification

Their first move to spark growth was fixing our 'leaky bucket,' aka retention probs.

They focused on keeping users over chasing new ones since our new-user growth was all organic.

Plus, they had a hunch that gamification could level up retention.

Why this felt right?

1. Duo already had fire gamification vibes—home screen progress, streaks, and achievements were working.

2. Top games back then had way better retention than us, so we knew there was room to level up.

They made a quick deck and got just enough of a 'let's go' from the execs to start a new squad—the Gamification Team.

Small problem, though: they had zero clue what new gamification tweaks would click for Duo.

At the time, the squad was hooked on Gardenscapes (think Candy Crush vibes), and that game ended up being our first inspiration. 🎮

They checked out Gardenscapes’ game mechanics but didn’t know what we were hunting for—just that it was stickier than Duo and had some similar vibes.

Both had short sessions (3-min Duo lessons = match-3 levels) and progress bars to show how close users were to finishing.

But Gardenscapes had one thing Duo didn’t: a moves counter. It limited the number of moves per level, adding urgency and scarcity to the game. That was🔑.

We stole the idea and gave Duo users several tries to get answers right before restarting the lesson. Game on.🎯

It took us months to drop the counter feature. But guess what?

Total flop. No boost in retention, no DAU spike, and barely any user feedback.

The worst part? It wrecked the team.

Some wanted to keep tweaking it; others were ready to ditch it. Drama hit hard, and the squad split fast.

The idea got scrapped, and the vibe was straight-up trash.

The only W? They learned a lot about the company’s vibe and how to step up my leadership game—but that’s tea for another time.

Phase 2: Referrals

After our gamification flop, they switched gears and built a new squad: the Acquisition Team, which is all about getting new users.

Uber was crushing it with their referral program, so we stole the vibe.

We made a referral deal where users got a free month of Super Duolingo (back then, Duo Plus) to bring in friends.

Seemed like a solid win to us! 🚀

They launched the referral feature, hoping for a big win. But nah, new users only went up 3%. A W, but not the game-changer we needed.

The team kept grinding, tweaking the program, and trying other ideas, but nothing hit.

That’s when they knew—they had to switch it up and find a new way to crack the growth code.

Thoughts

After two straight fails, they took some time to think about making smarter product moves.

Looking back,

In Gardenscapes, each move feels like a big choice, with obstacles to beat.

But Duo lessons are simple—you know the answer or you don’t.

So, the moves counter felt like an annoying add-on, not fun.

Key differences were missed.

The referral program didn’t work either.

Uber’s rides are always paid, so free rides are a constant hook.

For Duo, our best users already had Super Duolingo, so giving them a free month was pointless. Our strategy ended up excluding the users we needed.

In both cases, they borrowed ideas from other apps but didn’t think through how the context would change things. A

After these failures, I knew I had to be smarter about how we borrowed ideas. Now, when I look at a feature, I ask:

1. Why is this working for them?

2. Will this work for us, and why?

3. What tweaks do we need to make it fit?

They needed to be more thoughtful when copying. If they’d been more careful with their choices, they probably wouldn’t have gone all-in on referrals.

Catch part 2 next week for the final piece of the puzzle.

OpenAI's Operator redefines AI—what if your tasks could handle themselves?🧠

This track gave me a serious boost—check out ‘Atlas Air’ by Massive Attack🎵

AI Revolution: Are we blind to its most transformative possibilities?📱