The company and program
| Client | A seed-stage language-learning app for adults. |
| Stage | Seed |
| Industry | Education apps |
| Timeline | One 90-day growth sprint, July 13 – October 11, 2026. Day-30 retention read through November 3 |
| Channels | iOS and Android app, push, email |
| Team | A dedicated Kiko growth lead plus agents for event audits, cohort reports, experiment write-ups and first drafts of reminder copy. Client side: one mobile engineer, one lesson designer and the product lead |
The approach
The streak counts real lessons only: a lesson completed on each of the first 7 days, install day included. Streak freezes don't count. We kept 10% of new learners on the old experience for the whole sprint.
| Test | Hypothesis | Comparison | Decision rule |
|---|---|---|---|
| Short first lesson | An 11-minute first lesson loses people before they feel any progress | A 4-minute first lesson ending in a real exchange with the AI tutor, vs. the 11-minute lesson | Ship if first-lesson completion rises 5 points or more and the day-7 streak doesn't fall |
| Daily goal | A small default goal is easier to repeat | A goal picker (3, 5 or 10 minutes) with 3 minutes as the default, vs. a fixed 10-minute goal | Ship if the day-7 streak rises at least 2 points, with trial starts within 7 days flat or up |
| Reminder timing | A reminder at the learner's own lesson time works better than a fixed evening slot | "Same time tomorrow?" opt-in after lesson 1, with reminders at that time, vs. a 7 p.m. reminder asked for at signup | Same rule |
| Rescue lesson | Learners who are about to miss a day will do something short | A 2-minute review that counts as the day's lesson, offered at 8 p.m. to learners who haven't practiced, vs. a full lesson only | Same rule |
What we changed
- Weeks 1–2: audit and baseline. We fixed lesson-completion events and pulled June's funnel. Only 64% finished the first lesson on install day, and 34% were still on a streak by day 2.
- Weeks 3–4: short first lesson. Shipped after a 2-week read. Install-day completion jumped first.
- Weeks 4–5: daily goal picker. The paywall-on-day-3 test ran alongside and was reverted.
- Weeks 5–6: reminders at the learner's own time. Opt-in moved to after lesson 1.
- Weeks 7–8: rescue lesson. The 2-minute review shipped to everyone in the treated group.
- Weeks 9–12: hold and measure. No new changes. We read 7-day streaks and 30-day retention for each weekly cohort, then held the sprint review.
The results
Before: 26,800 new learners in June 2026. After: 24,300 treated learners from sprint weeks 9–12 (September 7 – October 4). The holdout reached a 7-day streak 22.6% of the time, so the lift is almost all incremental.
| Metric | Before | After | Change |
|---|---|---|---|
| Day-7 streak rate | 22% | 39% | +17 pts |
| First lesson completed on install day | 64% | 81% | +17 pts |
| Came back the day after install | 41% | 55% | +14 pts |
| Reminder opt-in | 47% | 66% | +19 pts |
| Day-30 retention | 9% | 14% | +5 pts |
| Trial started within 7 days | 6.2% | 7.4% | +1.2 pts |
What didn't work
- Week-one leaderboards. No change in the day-7 streak. We turned them off for new learners.
- "Your streak is at risk" emails. Unsubscribes rose 2.4 times. The push version stayed, and the email was cut.
- Moving the paywall to day 3. We hoped it would protect the habit. Trial starts fell and the streak didn't move, so we reverted it in week 5.
What we learned
- Shrink the first lesson until almost everyone finishes it on install day.
- Ask learners when they practice, then remind them at that time.
- Count only real lessons in the streak metric, so a freeze or a badge can't flatter the result.
Frequently asked questions
How much does a retention sprint for a language app cost?
There was no creator or media spend, so the only cost was Kiko's 90-day sprint, priced from $10,000 a month. The client's mobile engineer and lesson designer built the four changes; Kiko ran the audit, the tests and the holdout read.
How long until the streak rate moved?
Install-day completion moved within the first 2-week read of the short lesson in weeks 3–4. The four changes were all live by week 8, and weeks 9–12 were held with no new changes so the 39% could be read cleanly.
Would this work for a Series A test-prep app or a kids' learning app?
The short first lesson, small default goal and own-time reminder transfer to any daily-practice app. Test-prep has a fixed exam date, so key reminders off it. The client served adults only; a kids' app needs its own consent rules.
Related case studies and guides
Same industry, other angles:
- Creator marketing: Day-7 retention up from 18% to 31% for creator-sourced users
- Paid ads: Cost per paid learner down 44% in 10 weeks on TikTok and YouTube
- Full growth system: grew paid learners 2.9x in 9 months