Someone else already ran your experiment, and published the result.
Bottom lines
-
Growth halved, and management named itself as the cause.
Year-on-year daily active user growth held at 40% or better every quarter from Q2 2022 to Q2 2025, fell to 30% by Q4 2025, and is guided to roughly 20% for 2026. The CEO attributed part of the deceleration directly to the company's own increased focus on monetisation. Public companies rarely concede that, which suggests the internal data left little room to argue. [1][2]
-
The friction that monetised was the friction that cost users.
Adding friction to the free tier was described as the fastest route to monetisation — and coincided with the slowdown. Only around 10% of monthly active users pay. The company is now deliberately moving AI features back into the free experience and accepting slower bookings growth to do it. [2][3]
-
The cost of the reversal has been quantified and accepted.
2026 guidance is bookings growth of 10–12% against revenue growth of 15–18%, with adjusted EBITDA margin around 25% and margins dipping in the first half. Roughly $50m of bookings is being forgone deliberately. A $400m share buyback was authorised alongside it. [4]
-
The arithmetic of the target is the thing to press on.
Reaching 100 million daily users from 20% growth in 2026 requires re-acceleration to roughly 26% average in 2027 and 2028. Growth plans that require next year to be better than this year, after this year was made deliberately worse, deserve a specific question about what changes. [5]
The chart to put on the slide
Fig. 1 · Duolingo daily active user growth, year on year, actual and guided
Source [1][2][5]. Accessed 28.08.26. The final bar is the average annual growth implied by the stated 100m target, not company guidance, and is labelled as required rather than forecast.
The marked bar is the finding. The trough is a decision, not a shock — which is the entire argument management is making. The bar to its right is the one that has to be defended, because nothing in the guidance explains how it happens.
Recommended actions
| Action | Effect | Effort | Owner |
|---|---|---|---|
| Audit your own free tier for friction added in the last 24 months, and date each change against your acquisition curve. The largest consumer subscription business in the category found the two were connected. | High | Medium | Product |
| Separate monetisation that adds value from monetisation that removes it. Ad load and paywalled core features are the second kind, and the second kind compounds against word of mouth. | High | Low | Founder |
| If AI features are behind your top tier because inference was expensive, re-test that placement. Duolingo is moving them down as costs fall. | Medium | Medium | Product |
| Ask the re-acceleration question of your own plan. If next year's number requires a rate you have never achieved after a year you made deliberately slower, the plan needs a mechanism, not a target. | High | Low | Board |
Position and scale, for context
| Measure | Value | Period |
|---|---|---|
| Daily active users | 56.5m | Q1 2026 |
| Monthly active users | 137.8m | Q1 2026 |
| Paid subscribers | 12.5m | Q1 2026 |
| Revenue | $292.0m | Q1 2026, up 27% YoY |
| Revenue | $298.4m | Q2 2026 |
| Share of MAU paying | ~10% | Stated, Q4 2025 call |
| US share of revenue | 38% | 2025; no other country above 10% |
One counter-signal worth carrying into the discussion: monthly active users declined quarter on quarter from roughly 135m to 133m between Q3 and Q4 2025. A single quarter is not a trend, but it sits awkwardly against a strategy whose entire premise is user growth.
Declared coverage gaps
No cohort data is public
The link between free-tier friction and churn is management's own attribution. It is credible and unusually candid, but it is not independently evidenced here.
Metric definitions shift
Bookings, revenue and DAU are guided on different bases and different periods. Figures in section 04 are as reported and not restated onto a common basis.
The 26% figure is derived
It is the arithmetic implied by a stated medium-term target, not a company forecast, and is labelled accordingly on the chart.
Competitive substitution is not sized
General-purpose AI assistants are widely named as a substitution risk. No attempt has been made here to quantify that, and any figure claiming to would be worth distrusting.
Sources and verification
| Key | Source | Accessed | Verified |
|---|---|---|---|
| [1] | Duolingo shareholder letter and Q4 2025 earnings coverage, PYMNTS, February 2026 — DAU growth by period, 2026 guidance, CEO attribution of the slowdown. | 28.08.26 | — |
| [2] | CX Dive, Duolingo Q4 2025 earnings call coverage, February 2026 — Q4 2025 DAU, revenue growth, share of users paying, friction commentary. | 28.08.26 | — |
| [3] | TipRanks, Duolingo earnings analysis, March 2026 — 2026 investment-year framing, advertising yield strategy. | 28.08.26 | — |
| [4] | Duolingo Q4 2025 earnings call highlights — 2026 bookings, revenue and margin guidance, buyback authorisation, tier strategy. | 28.08.26 | — |
| [5] | Class Central, Duolingo 2026 strategy analysis, May 2026 — re-acceleration arithmetic to 100m DAU, free-tier changes. | 28.08.26 | — |
| [6] | Company metrics summary, StockCounterparts, July 2026 — Q1 2026 DAU, MAU, paid subscribers, revenue and geographic concentration. | 28.08.26 | — |