Claim and evidence
Duolingo learners who reach a streak of just 7 days are 3.6 times more likely to complete their course.Duolingo · blog.duolingo.com · 2022-01-31
Primary
The company or founder describes the tactic and its result themselves: own blog, talk, filing.
Learners at a 7-day streak completed courses at a much higher rate (correlation, company-reported).
Working · 2024-12 Duolingo's retention lead described 600+ experiments on streaks as of late 2024; correlation figures are not causal.
What the numbers show
more likely to complete their course for learners who reach a seven-day streak (Duolingo, correlation, company-reported)
blog.duolingo.commore likely to use the app the next day at a seven-day streak (Duolingo, correlation, company-reported)
blog.duolingo.comWhy it works
Early on each added day feels like progress; later the fear of losing a long count keeps people returning. The counter turns a vague intention into a visible daily commitment that is cheap to maintain.
How to run it
- Define the smallest daily action that counts
- Show the count prominently and celebrate milestones
- Send a reminder before the day ends when the count is at risk
- Track the share of daily actives holding a streak of 7+ days
- Test mechanics that lower the cost of extending it
Before / after
Generalised wireframe of the pattern, not a copy of any company's screen.
Ways to run it
Design options to test, not promised results.
- Option ASmallest action counts
Define the daily action that extends the streak as small as possible.
- Option BMilestone celebrations
Celebrate streak milestones prominently so each new level feels earned.
- Option CEnd-of-day reminder
Send a reminder late in the day only when the streak is at risk.
- Option DFriend streaks
Let friends keep a streak together, as Snapchat Streaks do with snaps sent back and forth.
Timeline
- 12017 first documented: Duolingo growth team blog
- 22021 Duolingo — Another post states learners with a 7-day streak are more likely to use the app next day.
- 32022 Duolingo — Reported that reaching a 7-day streak is associated with finishing a course; the post ties the streak to habit-formation research.
- 42024 Snapchat — a16z cites Snapchat Streaks (send snaps back and forth for days to earn a flame emoji) as the classic consumer streak.
- 52024-12 status checked: Working
Dated examples
- 2024
Snapchat
a16z cites Snapchat Streaks (send snaps back and forth for days to earn a flame emoji) as the classic consumer streak.
“encouraging them to send snaps back and forth with their friends for at least 3 days straight to earn a flame emoji.”
✓ quote checked word-for-word on the pagea16z.com · 2024-05-07
- 2022
Duolingo
Reported that reaching a 7-day streak is associated with finishing a course; the post ties the streak to habit-formation research.
Learners at a 7-day streak completed courses at a much higher rate (correlation, company-reported).
“Duolingo learners who reach a streak of just 7 days are 3.6 times more likely to complete their course.”
✓ quote checked word-for-word on the pageblog.duolingo.com · 2022-01-31
- 2021
Duolingo
Another post states learners with a 7-day streak are more likely to use the app next day.
“Our data shows that learners who reach a streak of 7 days are 2.4x more likely to use Duolingo the next day.”
✓ quote checked word-for-word on the pageblog.duolingo.com · 2021-01-08
Does it still work?
Duolingo's retention lead described 600+ experiments on streaks as of late 2024; correlation figures are not causal. source
Risks
- Streak-to-outcome stats are correlational (motivated users build streaks)
- Losing a long streak can make users quit
- Counter pressure can feel manipulative
Sources
- blog.duolingo.com
“Duolingo learners who reach a streak of just 7 days are 3.6 times more likely to complete their course.”
✓ quote checked word-for-word on the page - blog.duolingo.com
“Our data shows that learners who reach a streak of 7 days are 2.4x more likely to use Duolingo the next day.”
✓ quote checked word-for-word on the page - a16z.com
Change log
- 2026-11-20 — harvest v0.1 (retention)