Figuring out why lapsed learners give up on winning their streak back, and what to do about it. Summer 2026.
Duolingo: Earning back the streak
If you’ve ever lost a Duolingo streak, you know the feeling. Earnback is Duolingo’s way of giving it back: when a streak breaks for over 48 hours and a learner repairs it with a streak freeze or gems, they can win it back by completing a few back-to-back sessions in one sitting.
This summer, I joined the Current User Retention team (Growth pillar) as a Thrive Associate Product Manager intern. My problem space: only 70% of users who started Earnback were finishing it. I wanted to know why.
what did I achieve?
Found where learners were giving up. I spent a lot of time in Duolingo’s internal metrics dashboard building funnels, and one pattern kept showing up: no matter how long the streak, the drop-off happens in the very first session, where 60% of quits occur. Digging deeper, about 70% of learners who got a question wrong in that first session quit, compared to about 15% in a regular lesson.
Narrowed 11 ideas down to one. I wrote a one-pager, a roadmap, and multiple specs. Duolingo framed Earnback as a tough challenge, but these are lapsed learners trying to come back, so I recommended making it feel more encouraging and delightful. My final spec adjusts the lesson difficulty, leaving out hard mode, new patterns and the trickiest exercise types.
Got it approved. I presented the proposal at a mini product review with senior leadership and CEO Luis von Ahn, and it was approved. I left the team with the full roadmap and the feature queued for implementation.
Supported the wider team. Beyond Earnback, I ran experiment analyses for other features on the Current User Retention team, digging into A/B test results to help evaluate how each change affected learners.
thoughts and learnings
I learned what a PM really does. Making data-backed decisions, weighing tradeoffs, and prioritizing ruthlessly. Going from 11 ideas to one taught me how much of the job is deciding what not to build.
Great products take a lot of people. I worked with engineers on my team and on the Learning pillar, product designers, learning experience designers and senior PMs. Every single conversation shaped the final recommendation.
The best solution isn’t always the flashiest one. Sometimes it’s the one the data keeps pointing to.
As a Duolingo user myself, getting to work on the feature that helps learners find their way back was such a full-circle moment! It’s a product I love, and I got to help make it a little more forgiving. 🦉