DUOL (Trans): Guidance intentionally conservative, but user concerns lead it to forgo near-term monetization. ---

Below is Dolphin Research's Trans of$Duolingo(DUOL.US) FY25 Q4 earnings call; for the earnings analysis, see《DUOL: Guidance Bomb Again — Is the Green Bird now 'Dead Duo'?》

I. Key Takeaways

1. FY26 Guidance

a. Bookings growth of 10%–12%. Full-year outlook.

b. Revenue growth of 15%–18%. Full-year outlook.

c. Adj. EBITDA margin of Approx. 25%. For the year.

2. Key financial logic & trends

a. Adj. EBITDA margin: Q1 at 25.5%, with Q2 down Approx. 300bps on heavier investment. Margins recover in H2, with Q4 the peak for the year.

b. Costs: R&D and S&M to grow faster than revenue. Spend growth to outpace top-line.

c. GPM: expected slightly below historical levels as AI coverage expands and cloud/compute costs rise. Short-term margin pressure from broader AI feature deployment.

d. Capital allocation: BOD authorized a $400 mn buyback. Repurchases to begin under the plan.

3. Q1 FY26 Guidance

a. Bookings up 11%. Q1 guidance.

b. Revenue up 25%. This reflects deferred recognition from prior orders, with growth likely easing to below 15% in H2.

II. Call Details

2.1 Management Commentary

1. Core language biz: back to teaching fundamentals and free experience. a. Teaching quality: In 2026 we plan thousands of A/B tests, leveraging AI to fundamentally change learning. The goal is materially better outcomes and stronger word of mouth.

b. Free experience optimization: We recognize DAU growth slowed in 2025 (now above 50 mn, but decelerating), and will improve the free tier to re-ignite user growth. We aim to reduce friction and drive broader engagement.

2. New growth engines: multi-subject expansion

a. Cross-subject roadmap: Math, music, and chess as the next growth engines. Expanding beyond language learning.

b. Strategy: Diversified content to lift the overall DAU ceiling. This underpins the goal of doubling DAU to 100 mn by 2028.

3. AI strategy: a. Tech leadership: Management believes AI will reshape education. Duolingo intends to lead this transition.

b. Cost structure: We plan to push AI features to a broader base, not just a small group of high-paying users. This will modestly pressure GPM in the near term.

2.2 Q&A

Q: On the strategy pivot, how will AI and innovation support the goal to double DAU to 100 mn by 2028? What are the top product priorities?

A: We are at a unique moment where AI will transform how people learn. I believe AI-driven teaching quality will approach one-on-one tutoring in the next few years while retaining the fun of mobile gaming. AI lets us identify weak spots, answer questions in real time, and reduce confusion; in language learning, our 'Lilly video call' feature is getting increasingly effective at simulating real conversations.

The same applies to other subjects. As noted in our shareholder letter, this year we aim to build the world's best math tutoring app. About 1 bn people study math globally, and we are confident AI can deliver high-quality teaching to this massive audience. This is why now is the right time to accelerate acquisition and scale.

To reach 100 mn DAU by 2028, we will invest in three areas. First, keep improving language teaching quality, our largest and long-standing core. Second, optimize the free experience, as 15 years of data show DAU rises when we reduce friction. Third, cultivate new engines (new subjects); chess has reached 7 mn DAU in under a year, likely making us the No. 2 chess platform globally even without ASO. These new subjects will be key to long-term user growth.

Q: For Video Call, how do you balance between Max and Super tiers? As AI inference costs fall, what is the logic and financial impact of moving it to Super?

A: When we launched AI features, inference costs were high, so we placed them in the top-priced Duolingo Max. We stated we would broaden access as costs fall to serve more learners; Video Call costs have dropped markedly over the past two years, giving us confidence to test it in Super.

This aligns with our A/B testing culture. Outcomes could go two ways: a significant lift to Super retention and sign-ups, boosting revenue, or a bookings hit if Max becomes less compelling. If the bookings impact is too large, we will adjust, e.g., metering Super access (daily call limit) while keeping Max unlimited.

Super has roughly 10x the subscribers of Max. Moving Video Call to Super would 10x the base for conversation practice. We are making this shift from a position of strength, given robust financials, and we do not see a threatening competitor. We prefer to optimize proactively before competition emerges rather than make defensive moves later.

Financially, we have built ample room for experiments into this year's guidance. The outlook contemplates product strategy tests and potential volatility, ensuring we do what is right for customers while maintaining healthy performance.

Q: After removing monetization friction, when will DAU growth re-accelerate? How is Q1 user momentum?

A: We pivoted to user growth last call for two reasons: we saw DAU deceleration in 2025, and we believe AI presents a unique window to press harder rather than accept slower growth. Since last quarter, the deceleration has become clearer in the data, and we now expect FY26 DAU growth of around 20%.

My confidence in AI has strengthened, so we are prioritizing DAU expansion. Our FY26 plan and guidance reflect a growth-first approach.

Q: The FY26 strategy reallocation seems more financially impactful than previously expected. What changed in your view on user growth since last quarter?

A: The guidance does reflect the strategy shift. Q1 actuals are above guidance, showing strong fundamentals, but we remain prudent given tough comps later in the year. We are pushing for the 'larger prize.'

From a financial model perspective: staying the course could lead to ~$1.5 bn revenue and over $400 mn adj. EBITDA in a few years. But at 100 mn DAU, with reasonable monetization and scale, we could reach ~$2.5 bn revenue and $700+ mn EBITDA. This is our company-wide target; the positive model feedback will lag, showing late in 2026 and more strongly in 2027–2028, but it is the right long-term choice.

Q: Beyond the above, is the slowdown due to market saturation or increased competition?

A: We are not worried about saturation. Duolingo holds ~85% of DAU share among language learning apps globally, a ratio that has been stable. On penetration, only ~2% of U.S. internet users use Duolingo daily, 3% in the U.K., and 4% in Germany. If global markets merely reach the U.S. 2% level, DAU would double. Exit surveys show most users stop because they 'gave up learning,' not because they switched to competitors.

Q: When will 'Speaking Adventures' and advanced content for the top nine languages (covering >90% of DAU) roll out?

A: 'Speaking Adventures' blends text and voice interactions and will be available to both free and paid users. It is in testing now and should fully roll out around mid-year.

Advanced content is near-term, not years away. Within one to two months, courses for our top nine languages (covering 90%+ of DAU) will offer content at Duolingo level 129 (advanced). That level matters as it is the threshold for jobs requiring that language. Launch is step one; we will iterate through the year based on data feedback.

Q: TikTok virality seems lower vs. last year. How will you optimize this acquisition vector?

A: Our social team remains best-in-class, and videos still earn millions of views. Last year set an extremely high bar with multiple No. 1 viral hits on TikTok; algorithm shifts make sustaining that dominance harder. We will continue to pursue virality and are satisfied with current output.

Q: How does Video Call impact speaking time? Will moving it to Super lift engagement and bookings?

A: Iteration on Video Call is going well. User 'words spoken per call' are rising steadily. We are improving Lily's guidance, prompting users to ask a question next sentence or use connect words, nudging deeper language output.

Broadening Video Call to more Super users should create more 'brand advocates.' Skeptical users often share achievements after fluent conversations, amplifying authentic word of mouth. This engagement lift should translate into stronger brand momentum and bookings growth.

Q: After improving the free experience, how will monetization evolve? Did you over-monetize in the past, and how will you balance DAU growth with monetization?

A: This is central. We are 'under-monetized' with paid mix at ~10%, while mature freemium models like Spotify can reach ~50%. But we also 'over-monetized' historically by adding friction to force subscriptions, which hurt DAU growth.

Friction works fast (e.g., double ad load could add ~$50k/day next week), but conflicts with long-term DAU goals. So in 2026, we will explore monetization that does not rely on added obstacles.

We will prioritize feature value and emotional drivers, e.g., more sticky features like Video Call and gaming-style IAP such as avatar customization. Duolingo has strong game attributes. Value-based monetization is slower to show results, but with only ~10% conversion today, we are confident this healthier, more sustainable path will pay off.

Q: On chess reaching 7 mn DAU, how much overlaps with main app users? Does this imply language growth is slowing, or are users multi-subject?

A: 7 mn DAU in under a year is remarkable. Most were converted from language learners via in-app channels, which is cheaper than acquiring net-new users externally. Many users now both learn languages and play chess on Duolingo.

This does not cannibalize language growth. Language course growth is broadly in line with company DAU growth and remains our largest core. The pattern is dual participation, not subject substitution.

Brand spillover is emerging. Awareness that Duolingo offers chess is rising, and we are attracting net-new chess-first users. Over time, chess should build its own acquisition engine beyond internal conversion.

Q: On bookings outlook, given uncertainty around Video Call tests in Super, how does the guidance differ from past Super conversion assumptions?

A: Decomposing the move from ~24% bookings growth in Q4 to current guidance, the gap is roughly split in two. Half from continued user growth deceleration noted by Luis; the other half from more cautious monetization as we remove some friction.

As you know, we have a deep A/B testing culture. Guidance is designed to preserve adequate room for experimentation. With strong financials today, it is worth sacrificing some near-term growth to reach a larger long-term target.

Q: How long until you can confirm success via data?

A: It will take time. We need extensive product changes and to observe long-term behavior. We embedded some conservatism in year-end guidance, by which we should start seeing positive effects.

Q: Does FY26 guidance include pricing changes across tiers?

A: We will run many price tests as usual, likely at higher frequency, including both increases and reductions. All options remain open.

The outlook already incorporates expected test outcomes. I cannot assert that Super pricing must rise because of Video Call; we will wait for A/B results. We are data-driven, not arbitrary on pricing.

Q: Does the 100 mn DAU mid-term goal imply sustained high marketing spend and secondary monetization throughout?

A: Marketing efficiency remains high, with spend a small share of revenue. Absolute marketing dollars may rise annually, but we expect stronger scale effects post-2026. Hitting 100 mn DAU will reshape our scale; while exact post-2026 models are hard to pin down, we are optimistic on diversified monetization. Direct ad sales are progressing well, and given our friendly brand, avatar IAP has meaningful potential.

Q: How will you balance CAC and monetization over the next few years?

A: Our model is resilient and scalable, showing significant leverage even at current scale. We are in an investment phase for long-term growth, but this does not require doubling headcount or marketing.

We are investing from a position of profitability and strong cash flow. When we pivot back to monetization growth, we will do so smartly, sustaining core margins and FCF while scaling the business.

Q: Will ads play a smaller role in monetization? How will you grow ad revenue without raising ad load?

A: We are unlikely to increase ad load. Ads currently appear only at lesson end, and that frequency will not rise this year as heavier ad loads conflict with our DAU-first strategy. Historically, we underinvested in ads, relying on minimal third-party networks.

The core strategy is to lift unit ad value. We are shifting to direct deals, which bring two benefits: better ad quality (think Disney vs. low-end game ads) and higher CPMs.

We are also testing frontier approaches. The vision is 'immersive language ads' — if you are learning Spanish, parts of the ad would be in Spanish. Users are more engaged given the teaching angle, boosting advertiser conversion. These are early tests, but the direction is clear: keep ad load flat while materially increasing monetization efficiency per ad.

Q: The market worries AI translators and general chatbots caused the 2025 user slowdown. Why are you confident ruling this out?

A: AI translation is not a concern internally. Large-language pairs like Spanish–English have had near-perfect translation for a decade. User motivations are clear: some learn for niche hobby enjoyment, others to acquire English as a skill. Our exit surveys never show users leaving for translators; the core reason is they 'got busy,' usually distracted by social media, not replaced by translation tools.

Q: What data do you see from efforts to improve UX? When do you expect DAU growth to stabilize?

A: We have moved some Max-tier perks to free users. Engagement and affinity for these features have risen notably, but we have not yet seen explosive DAU growth. This is expected as product logic takes time to propagate. As noted, look for steadier recovery toward year-end, not immediately.

Q: How did the first large renewal cohort for Duolingo Max perform? Did it influence the decision to move Video Call to Super, and what subscription impact do you expect?

A: Max performance is solid; renewals and retention look strong. We are satisfied there. Moving Video Call to Super is an offensive move; as inference costs fall substantially, keeping core functionality locked in Max for a small subset is less logical.

Our philosophy is to offer the best features to the broadest base when costs allow, potentially even free in the future if costs drop further. Our core interest is maximizing users and finding diversified monetization, not extracting high prices from a small group. This is about doing right by users, not defending Max.

Q: In FY26 guidance, what ARPU assumptions did you make around potential downgrades from Max if Video Call moves to Super?

A: We cannot precisely predict A/B outcomes yet, but guidance embeds a reasonable range. We are confident results will land within it, and we have tools to keep financials aligned with outlook.

For example, if testing shows severe Max cannibalization and a negative impact from offering Video Call in Super, we will adjust immediately, such as metering Super (e.g., one call/day) while keeping Max unlimited. We will flex product granularity based on feedback and anchor performance to guidance.

Q: On math, will you distribute through schools? How do you approach acquisition in this large market where motivation differs from language learning?

A: Math is unique, with ~1 bn learners globally, similar in scale to language learning. The key difference: language learners are usually self-motivated, while math learners (K–12) often 'must learn.'

Our first step is product excellence. Internal versions perform very well. While we considered school channels, our current hypothesis is to position as a 'strong supplement' to school education rather than pursue complex district deals, competing more with after-school programs like Kumon.

Longer term, math is attractive because parents are highly willing to pay even if kids are not. This 'parent pays, child uses' model offers substantial monetization potential. It will not move the needle financially this year, but it is a foundation for long-term DAU and revenue.

Q: Will multi-subject (math, music, chess) investment materially improve retention?

A: Paid retention is healthy and stable. Users engaging across multiple subjects tend to have higher retention, and their share is rising, but we are conservative in modeling and do not assume a step-up. Better product also aids acquisition, which feeds back to existing users. We assume retention holds steady, with upside potential.

Q: AI lowers development barriers even as it is your advantage. How do you view new competitor risk?

A: AI speeds development for everyone, including us. There is a misconception that one can 'quickly clone Duolingo' with AI; even if code is fast to write, replicating a decade of edge-case handling and engagement-driving product details can take years.

More importantly, our moat is distribution and free product strategy. Duolingo holds ~85% DAU share among language learning apps globally. New entrants typically rely on performance ads, but with our superior free product and massive scale, it is hard to make free acquisition economics work. If they are forced to add paywalls, growth caps quickly. We have seen hundreds of challengers face distribution bottlenecks and fail to scale.

Q: In the AI era, will Duolingo remain an independent learning app or become a plugin/app layer on LLM platforms? How long is the transition to an 'AI-native' stack?

A: History is instructive: wearables and smart speakers (e.g., Alexa) were once touted as 'the next big thing,' and we even built early watch apps, but beyond fitness there was limited utility. If behavior shifts and apps disappear, we will adapt.

Duolingo is closer to a game; users love the animation and visual engagement, much like playing 'Candy Crush Soda.' Text-only chatbots cannot deliver that emotional, visual immersion. We believe investing in a visually rich independent experience is right.

On becoming 'AI-native,' we have always been tech-driven and aim to adopt AI as fast as possible. Sometimes we move so fast it sparks controversy, but we remain at the forefront of AI in learning. This is not a multi-year rebranding, but continuous evolution across our stack to sustain leadership.

<End>

Risk Disclosure & Statement:Dolphin Research Disclaimer & General Disclosure