--- title: "DUOL (Trans): DAUs keep accelerating; pivot to monetization next year" type: "Topics" locale: "en" url: "https://longbridge.com/en/dolphin/post/44745994.md" description: "Q3 DAU YoY growth accelerated to 29% from 23% in Q2, and Q4 bookings growth guidance is +13%, slightly above prior guidance.However, the second tranche of user growth incentives and the full rollout of Video Call will leave Q4 Adj. EBITDA roughly flat YoY at about $84 mn. Meanwhile, the company sunset 'Max' and plans to ramp monetization next year without adding friction to the free experience." datetime: "2026-11-03T23:40:45.000Z" locales: - [en](https://longbridge.com/en/dolphin/post/44745994.md) - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44745994.md) - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44745994.md) author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)" generator: "portal-rs" --- # DUOL (Trans): DAUs keep accelerating; pivot to monetization next year **Below is Dolphin Research's compiled transcript of** $Duolingo(DUOL.US) **FY26 Q3 earnings call.** **I. Core financial highlights recap** 1\. **Q4 guidance** a. **Growth**: bookings +13% YoY, revenue +14% YoY; the CEO said Q4 bookings are tracking slightly above prior guidance. b. **Profit**: Adj. EBITDA margin of 26.1%, Adj. EBITDA of approx. $84 mn, roughly flat YoY. c. **Cost structure**: Video Call is rolling out to all paid subs, with GPM down slightly vs. Q3. With continued investment and a second bonus accrual, Opex growth will outpace revenue. 2\. **Full-year guidance**: bookings +12% YoY, revenue +18% YoY, Adj. EBITDA margin 26.5%. The initial bookings growth range was 10%–12%, and is now tracking toward the high end. A full 2027 outlook will be provided in Feb. 3\. **User growth bonus**: Q3 Opex includes an $11 mn bonus accrual to reward the team for re-accelerating user growth. Accrual began this quarter as the target looks achievable. The second tranche is embedded in Q4 guidance, with full payout expected in early 2027. 4\. **GPM drivers**: GPM was originally expected to be about 200 bps lower this year, but AI costs came in below plan. The main Q3 GPM drivers were revenue beating estimates and lower-than-expected AI costs, unrelated to direct-payment traffic. **II. Earnings call details** **2.1 Key management commentary** 1\. **User growth**: DAU accelerated for the second straight quarter, with gains across most major markets. a. Q3 DAU rose 29% YoY, vs. 23% in Q2. b. 8 of the top 10 countries saw faster DAU growth YoY; Q4 DAU growth is expected to be similar to Q3. 2\. **Annual strategy**: The year began with priorities of 'teaching better' and driving user growth, and acceleration is now evident. a. Q3 bookings, revenue, and profitability all beat expectations. b. The momentum in user acceleration is expected to continue into Q4. 3\. **Monetization**: Monetization is being ramped while preserving the quality of the free experience. a. Most regions delivered double-digit YoY growth in subscription bookings. 4\. **Long-term vision**: The goal is to ultimately reach 1 bn learners, where a larger user base creates more monetization headroom. **2.2 Q&A** *Q: What are the main drivers of user growth this year, and how much did game collaborations contribute?* A: **Growth is primarily from product improvements, driven by better retention.** The Green Machine optimization system ships hundreds of A/B tests weekly. Last quarter’s Street Revival and this quarter’s collaborations with Brawl Stars and Eggy Party also brought users, but management believes most growth came from core app improvements and should persist. Growth has been broad-based across geos and products. *Q: How do users acquired from collaborations differ in retention and monetization?* A: **Collaboration users retain well, by design.** Players must study on Duolingo for multiple consecutive days to claim rewards, helping form a habit. *Q: DAU has rebounded but paid sub growth is still decelerating. Is this mainly due to extended free trials?* A: **Sub counts track revenue and should bottom in a few quarters.** Sub counts approximate the average of revenue and need to digest prior bookings: bookings recover first, then revenue and subs follow. This year, the team is testing DAU-aligned monetization such as extended free trials and Avatar suit IAPs; **Chess DAU just reached 10 mn** with almost no monetization, leaving many low-hanging fruits. *Q: How did testing perform for moving Video Call into Super, and why sunset Max?* A: **Per-call cost fell from about $0.30 to under $0.01, enabling migration into Super.** Savings mainly came from switching to open-source/open-weight models. Tests showed a revenue benefit and roughly 10x broader coverage. After testing the sunsetting of Max, the conclusion was clear, so the rollout began; **near-term bookings impact is minimal to slightly positive**. **Max accounts for a single-digit percentage of paid subs and bookings**. *Q: What churn assumptions are embedded in sunsetting Max? Is migration one-off or at renewal?* A: **Max is no longer sold to new users, with individual Max being moved to Super Family at similar pricing.** By year-end, only the smaller Max Family cohort remains. Tests indicate a net positive overall with manageable near-term impact. *Q: What is the lag between DAU and bookings growth, and can bookings catch up by 2027?* A: No 2027 guidance was provided, only that management feels good about bookings growth. Monetization of new users takes time, sometimes years. With the active base now materially larger and DAU re-accelerating, focus will shift more to monetization. *Q: Is a 72% GPM a reasonable level after full Video Call rollout?* A: No GPM guidance for next year. AI costs have been below plan this year, and Video Call and other AI features will roll out to all paid users by year-end. There is a strong desire to offer AI capabilities to users, balanced with overall plans and profitability. *Q: CURR hit a record high and newer cohorts retain better. Any update?* A: No specific figures, only that CURR continues to improve. CURR is the core metric when optimizing for DAU and retention, and **most DAU growth is attributable to CURR improvements**. The company does not plan to update it quarterly. *Q: How are paid marketing LTV/CAC and channel allocations?* A: **Paid marketing is about 60% of non-comp marketing spend and is profitable.** Related spend is slightly above $100 mn, with room to further improve LTV/CAC. Channels are diversified and actively reallocated based on returns. Paid marketing will not be the main acquisition lever, but it adds in 2026 and **likely continues in 2027**; the CEO called this one of the biggest internal wins of the year. *Q: Any hard data on engagement and retention uplift post Video Call rollout?* A: **Words spoken per user roughly doubled over the past year.** The metric covers all practice, not only Video Call, and usage continues to rise. Users who experience difficulty-tailored Video Calls during free trials show a **higher propensity to subscribe**. *Q: How is adoption of the Math course, and is there a standalone pricing opportunity?* A: **Math is growing roughly in line with Languages, but slower than Chess.** About 1 bn people globally learn math, implying a large TAM; the team hopes Math can match Chess growth velocity. No comment on standalone pricing. *Q: Can game collaborations scale, and what motivates game studios to partner?* A: **Expect a few collaborations per year, with next year focused on in-house events.** Most partners approach Duolingo, and the company won’t do hundreds. There is no cash changing hands; both sides cross-promote and gain users. Supercell re-skinned Brawl Stars with a Duolingo theme for weeks, received positive community feedback, and saw higher engagement and revenue. Next year’s in-house events include Avatar suits and more. *Q: Do collaborations and performance marketing reinforce each other?* A: **They are related and mutually reinforcing.** Performance ads are largely gaming ads, and the company’s performance marketing also benefits during collaborations. The exact relationship remains under study. *Q: How is AI changing the experimentation process, and how much more throughput is possible?* A: **Experiments in the past month were up about 90% YoY, with headcount up only about 10%.** AI deserves much of the credit: AI coding speed gains took time to realize, but idea-to-launch latency is now markedly shorter. AI excels at small changes while larger ones remain error-prone, aligning well with Duolingo’s high-volume, small-scope A/B testing model; **next year will see far more experiments than this year**, supporting both DAU and monetization growth. *Q: Any change in how shorter attention spans affect frequency and retention?* A: No related data disclosed. A dedicated team has been formed to improve engagement. *Q: How will you balance monetization and product friction in 2027 to avoid a repeat of 2025?* A: **There are strict internal red lines to avoid harming the free experience.** Multiple monetization levers that do not hurt the free user experience have been identified this year. *Q: Why migrate Max users to Super Family instead of individual, and how do you prevent cannibalization via sharing?* A: **Pricing is similar, avoiding a one-off bookings drop.** Family plans, while better value, have **very high retention** and renew as long as one member uses the product. After migration, features are the same and users can invite up to five others. A/B tests have run for some time with **no material cannibalization**. The CFO said this reflects confidence in market potential. *Q: Outlook on Lite tier penetration next year, and the ads business today?* A: **Lite tests are middling so far and the outlook is uncertain; ads exited the quarter with solid growth.** Lite is cheaper, ad-supported, and very small in scale, likely offered only in lower-GDP markets. The ads team and direct sales are now in place, shifting from low-end ad networks to high-quality brand ads, which takes time; **ads were roughly flat in Q3**, with a good outlook for next year. *Q: As monetization ramps again into 2027, what will be different and what are the guardrails?* A: **Continue at scale with A/B testing, with the only guardrail being no added friction to the free experience.** Levers include IAPs, better ads, and nudges to subscribe. With Video Call moved into Super and cheaper per call, it can be surfaced more often; Max was priced on the high side. *Q: Does avoiding friction mean conversion rates can’t improve?* A: **Conversion should improve to some extent, though the magnitude is uncertain.** Extending free trials has already lifted conversion and **is currently rolled out to only part of users, with further expansion planned**. There will be no user squeezing. The CFO noted bookings growth re-accelerated QoQ this quarter; conversion at the top-of-funnel for new users is multiples of that for existing users. *Q: Is AI’s biggest contribution internal efficiency or new products?* A: **Both; next year’s A/B test volume will be multiples of this year’s.** Simple tests can now be shipped by PMs and designers via internal tooling, but engineering is still needed at scale. Whether the multiplier is 3x or 10x is unclear. On product, the team will build speaking features beyond Video Call, and launch new subjects like Chess faster. *Q: Does sunsetting Max affect GPM beyond what’s implied in Q4 guidance?* A: **Q4 guidance implies GPM down about 100 bps QoQ, including the Max migration.** AI spend efficiency will keep improving while more AI features are deployed, balancing with profitability. *Q: Any change in off-app direct payment opportunities?* A: **Direct payment won’t be a major driver next year, with no plans for a full switch.** Some regions and users can already pay directly within the app. It is used mainly for performance marketing, leading users to the website first and then to download the app. *Q: Did the record Q3 GPM benefit from a higher share of direct traffic?* A: **No, it was mainly revenue beating estimates and lower-than-expected AI costs.** *Q: Can 1:1 AI tutoring become core over time, and how would the business model evolve?* A: **It should reach the level of 1:1 human tutoring within a few years, while remaining gamified.** Four to five years ago, research showed Duolingo matched classrooms; it is now better than classrooms but below top human tutors. Most people won’t attend human tutoring even for free, and **the biggest failure mode is user churn**. Therefore, AI will adapt to learners in real time within a gamified framework, like a human tutor. The CFO added that user scale and learning data are core advantages. *Q: If monetization will be a bigger focus next year, what does that mean internally?* A: **Continue to prioritize user growth, with more monetization-related A/B tests.** In 2026, the company’s core objective is DAU acceleration with a company-wide bonus attached. Monetization will also be a key focus next year. *Q: After Chess success, what is the timeline and subject focus for new courses?* A: No plan or timeline disclosed for new courses. With AI, Chess took about nine months from inception to launch, and **it can now be even faster**. *Q: With DAU up ~30%, how is per-DAU inference cost trending?* A: **Per-DAU cost has dropped significantly, outpacing growth in AI usage.** Open-weight models match frontier models for most user scenarios, with frontier models needed only in a minority of cases, driving sizable savings. *Q: At what inference cost level will you offer AI features to free users?* A: **Next year, the company will spend several million dollars on real-time inference for free users.** The scale will be modest, with the aim of opening some AI features to free users. *Q: How do you source compute, and will you build in-house?* A: **Currently rely mainly on inference service providers, with internal efforts also underway.** Choices will be made based on business needs. *Q: How will you support a larger experimentation footprint next year, and how will monetization tests progress?* A: **Next year’s A/B test count will be multiples of this year’s, shifting bottlenecks to ideas and process.** Tests span user growth, monetization, and pedagogy. New bottlenecks include an insufficient pipeline of testable ideas and ownership for fixing bugs introduced by non-engineer changes; processes are being refined. \ **Risk disclosure and statement:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [09999.HK](https://longbridge.com/en/quote/09999.HK.md) - [NTES.US](https://longbridge.com/en/quote/NTES.US.md) - [DUOL.US](https://longbridge.com/en/quote/DUOL.US.md) ## Comments (1) - **yuhaohao · 2026-11-04T08:46:37.000Z**: Collect feed --- > **Disclaimer: This article is for reference only and does not constitute any investment advice.**