DUOL (Trans): Reset year; core KPI is +20% user growth

Below is Dolphin Research's $Duolingo(DUOL.US) FY26 Q1 earnings call trans. For the earnings take, see 'Duo: Crawling Forward — The Turnaround Won't Be Easy'.

I. Key financial highlights recap

1. Full-year guidance unchanged: bookings growth ~10.5% (guide 10%–12%), revenue growth ~16.1% (guide 15%–18%), Adj. EBITDA margin ~25.7% (target ~25%).Full-year FCF expected to exceed $350 mn.

2. Q2 and quarterly cadence: Q2 bookings +~6% YoY (lapping tough comps from Energy launch, price hikes and strong ads), Q2 revenue +~17% YoY.Q2 GPM ~71%, Adj. EBITDA margin ~24%. Bookings growth expected to accelerate in H2 (Q3 +~3ppt, Q4 further), while revenue growth steps down in Q3 then stabilizes in Q4.

3. GPM and AI cost: Q1 GPM beat on materially lower unit AI costs.As AI features continue to permeate the product, GPM is guided to ~69% by year-end. Adj. EBITDA margin to be flat to slightly down in Q3 vs. Q2, then recover to ~27% in Q4.

4. Shareholder returns and balance sheet: Ending cash >$1 bn, no debt.Repurchased 514k shares (~1% of fully diluted shares); $400 mn remaining authorization. FY26 FCF expected to be near the buyback authorization size.

II. Earnings call details

2.1 Management remarks

1. User growth and strategy

a. Q1 DAU +21% YoY, in line. This year's strategic focus is on 'teaching better' and 'growing users'.b. FY26 is framed as a pivotal investment year, with spend within the financial guardrails to build scale and durability for the long term.

2. Language learning product upgrades

a. Speaking practice is now a larger share of the experience: launched Spoken tokens (voice answers across nearly all exercises), Speaking Adventures (real conversational scenes), and Flashcards (read words aloud to reinforce rapid recall).b. Video Call keeps improving, with Avg. spoken word count per user doubling vs. a year ago.c. Content milestones: nine top languages now fully covered to CEFR B2. In Q1 alone, 20,500 course units were released, over 10x a single-quarter output two years ago.

3. AI applications

a. AI is fundamentally changing content production capacity and the potential for personalized teaching.b. We are developing highly personalized, real-time content based on each user's full learning data — not just selecting the next exercise, but generating bespoke content for every user.

2.2 Q&A

Q: What drove Q1 DAU growth? Any regional differences?

A: DAU growth is critical and our top priority this year. We are growing across all regions and have been for years, though at different speeds; Asia is the fastest.Drivers are similar to before — word of mouth is consistently our primary acquisition channel, and most users come to us via referrals. We are increasing performance marketing, but spend remains modest vs. peers of similar scale.Another key driver is improving retention. We've made many product changes to increase stickiness, and the DAU/MAU ratio rises almost every quarter — including this one.

Q: What is your Q2 and full-year DAU growth outlook? Is ~20% still reasonable?

A: All unchanged from last call. We expect ~20% for the full year, with small quarter-to-quarter variations on comps, but the direction is around 20% annually.

Q: Update on activating the top 20% of free-layer actives and gaming-inspired initiatives?

A: On the free tier, we've made it better than two months ago — more features are open to free users, which we believe will foster word of mouth over time.But it's only been two months since the last call, so changes aren't dramatic yet.On gaming inspiration, we've long drawn heavily from games. Our goal is a product that teaches as well as 1:1 human tutoring while being as fun as a game. You'll soon see a cool avatar costumes feature, directly inspired by games. We're also changing how rewards are presented — e.g., card formats for a collectible feel.

Another important finding related to the free tier: longer free trials. Historically we used 7 days; we now find that longer trials (e.g., 1 month) not only lift bookings but also improve UX — users perceive 'I get one month free' as a clear positive.This is a monetization lever that does not conflict with DAU growth.

Q: How are you thinking about slowing MAU growth and the top-of-funnel?

A: We focus mostly on DAU and don't have a dedicated MAU team. That said, we do watch the top-of-funnel — it was broadly flat this quarter and we aim to accelerate.We'll invest more marketing in underpenetrated markets and keep improving teaching efficacy and the free experience to drive word of mouth. Referrals are great because they are free, but they are less controllable and measurable than retention, so we do things we believe help word of mouth even if control granularity is lower than on retention.

Q: As voice experiences scale, how do you keep the product fun?

A: I see why you might feel change, but internally our emphasis on 'fun' is unchanged. Duolingo's secret weapon is incentivizing continuous learning — people say they want to learn but do what is most fun. We devote substantial effort to making learning enjoyable.More speaking and voice do not reduce fun, and our data confirms that. Humans are highly visual, so you'll continue to see rich animation and gamified elements; even with voice, the visual experience remains strong. Our team dedicated to making the app more fun is running at full speed. You'll soon see avatar costumes — I'm wearing a hot-dog suit on Duolingo myself.

Q: Key learnings on AI for UX and content scale? Long-term margin impact?

A: In teaching, AI is the key to reaching 1:1 human tutor-level efficacy. Video Call conversational flow has improved sharply, and Avg. spoken word count per user doubled YoY.On content output, Q1's 20,500 units are roughly last year's full-year output, and we were already using AI last year — we are simply getting more proficient. We are also pushing deeper personalization — models choose and even generate bespoke exercises from your entire learning history, which is the essence of 1:1 tutoring.On costs, there are two dimensions. User-facing AI features lift COGS and weigh on GPM, with Q4 GPM guided to ~69%, reflecting heavier AI infusion in the product. Internal AI operating costs are also rising and are in guidance. AI costs come in waves — costs rise with new features, then teams optimize, then we push again. Q1 GPM beat because unit AI costs fell materially. ~69% is a reasonable year-end level and supports long-term growth, though the tech backdrop evolves and optimization can run faster than expected.

Q: How does Voice impact the Max tier? Will users downgrade from Max to Super?

A: As we said, we plan to add Video Call to the mid-tier Super Duolingo because teaching efficacy is so strong and we want broader access. We are currently testing Video Call for new Super subscribers and haven't rolled it to all existing users.Max remains unchanged for now — it may see price cuts, or Max could have unlimited Video Call while Super has caps; many possibilities exist. So far, no clear KPI differences, except some new cohorts only see Super, not Max — just one of many experiments.

Q: What drives H2 bookings acceleration? Any early DAU improvements?

A: H2 guidance is basically built on ~20% DAU growth, and you may see some early payback from investments, but we are not counting on much. FY26 is about the long term.Q2 bookings lap an exceptionally high base (e.g., last year's Energy launch). From Q2 to Q3 you will see a rebound as comps ease, and DAU will continue to drive bookings.

Q: With faster content production, how do you AB test to optimize engagement? Any impact on expansion into new subjects beyond languages?

A: It took years, but we finally brought the nine major languages to Duolingo Score 129. Someone told me this isn't the end — it's the beginning, because now we can truly scale quality optimization based on user performance.We do see content changes materially impact retention, notably for new users. For example, what you teach in Unit 1 matters — greetings vs. 'mom and dad' have different outcomes, and it's not as simple as one always being better; it's more nuanced. We may even move toward real-time bespoke content for each user — not the next exercise, but generating a sentence two exercises later based on all your data.

On new subjects, each has unique challenges. AI has sped up launches — chess took ~9 months to go live. But if you want math content with charts and interactivity, not ChatGPT-style plain text, it's much harder.AI makes adding new subjects easier, but it's far from trivial. We are very satisfied with existing subjects, especially math — courses now cover grades 2 through 12 and can explain why an answer is wrong.

Q: With ample cash, how do you view buybacks and capital allocation?

A: We repurchased shares during the quarter and have a $400 mn authorization. Broadly, we prioritize operating the biz., and we are stepping up investment this year.On buybacks, we buy more when the stock is lower and less when higher; we believe now is a good time to buy our own stock and offset dilution from recent years. This year's FCF is near the authorization.Another allocation lever is M&A — we consistently scan the market, but historically do small deals that don't strain the balance sheet. Duolingo is highly focused on investing in the core to drive growth.

Q: Does full rollout of advanced content unlock performance marketing? When could you scale spend?

A: It's most important for English learners. One major PM channel is underpenetrated markets, especially in Asia. In several large Asian markets we can acquire profitably — e.g., China, where profitable acquisition is possible and related to advanced English content.More fundamentally, we have historically underinvested in performance marketing, and are now building the infrastructure — proper attribution, proper post-acquisition onboarding paths, etc., things we should have built years ago. We are excited and expect results over the coming months.

Q: After adding Video Call to Super, how elastic is pricing? Any GPM impact?

A: We are testing optimal pricing for Super with Video Call. It's only been 8 weeks, so results are limited.We do know users are willing to pay more for Super with Video Call. How much more will take a quarter or two to determine.We are deliberately operating within a financial range this year to create room for pricing experiments. All this is embedded in guidance; any single quarter may be slightly above or below, but within expectations — no big surprises.

Q: Duolingo seems both under- and over-monetized. How do you know you're on the right path? When will you push monetization again?

A: Great question — it is a paradox. Only ~12% of MAUs are paid subscribers while Spotify is near 50%, so we see substantial room to lift paid penetration — we are under-monetized.At the same time, certain tactics may be overdone, adding too much friction to the free experience — that's the over-monetized part. The core issue: most historical monetization conflicted with DAU growth. Adding free-tier friction drives more payers (good) but also more churn (bad).We are finding and implementing monetization that does not conflict with DAU growth. Such levers exist. For example, longer free trials — we used 7 days historically, but many mature subs businesses use 1 or even 3 months.We have validated that a 1-month free trial lifts revenue without hurting DAU — '1-month free' is simply good for users and doesn't push anyone away. We may also test a 3-month trial. A 3-month trial delays bookings by a full quarter — which is why we set this year's operating framework to create space for such tests. These take longer to show impact than 'add friction to the free tier', but it's the right direction.

Q: After year-end GPM dips to ~69%, does it rebound, hold, or keep falling?

A: Margins on AI-heavy features tend to follow a pattern: higher costs at feature introduction, then optimization over time. Q1 GPM outperformance came from materially lower unit AI costs.Looking ahead, we plan to infuse more AI into the product, hence the guide to ~69% GPM — that means more AI content, which is positive for the long-term business. ~69% is a reasonable year-end exit; executed well, it should underpin solid growth.That said, the tech environment keeps evolving and optimization can run faster than expected, so outcomes can vary both ways.

Q: How does AI-assisted coding affect engineer efficiency and AB test volume? Any impact on long-term headcount?

A: You're right — we run hundreds of AB tests concurrently; it's core to our product philosophy. We have seen the number of AB tests we can run increase, which we believe comes from applying AI across engineering and product.The increase isn't massive, but it's the first time in years we've seen AB test output per capita rise. A year ago, the narrative on Twitter made it seem an engineer could code everything in five seconds and run 10,000 AB tests — that's exaggerated, not reality.Until recently, we and peers of similar scale didn't see true end-to-end speed gains; now we are starting to, with a moderate uplift. I expect this to continue, but not to 10x AB tests per engineer. An increase is good enough, and we welcome it.

Q: How do U.S. vs. Intl DAU growth differ? What does that mean for bookings?

A: U.S. DAU is growing, and almost every country is growing, but the U.S. is indeed slower than many Intl markets; Asia is the fastest.The monetization impact isn't huge — the U.S. monetizes well, but many countries also monetize well. China's monetization level is close to Western Europe (e.g., France) — not as high as the U.S., but already strong — and China is growing quickly. So lower U.S. growth doesn't imply low bookings growth.We aim to lift U.S. YoY growth by teaching better, fostering word of mouth, and increasing paid marketing in the U.S. (historically we did very little paid marketing there).

Q: As AI accelerates content production, how do you manage content quality?

A: The reason content volume isn't expanding faster is precisely our stringent quality control. We evaluate with AI and humans, then test on small cohorts before broad rollout.If you neglect quality, AI-generated content can indeed be poor. Over the past two quarters, spot checks and scoring show quality is actually improving.

Q: Are PM bottlenecks driven by product maturity or other factors? When can you fully ramp?

A: PM bottlenecks are mainly two-fold: first, we are building infrastructure to make PM truly professional. Second, more fundamentally, our free product is so good that many acquired users become very satisfied free users rather than paid — profitable acquisition is possible in some regions, but not in many others.Marketing spend this year goes beyond PM; the team has multi-layered plans and overall spend is increasing. Duolingo has existed 15 years; the product keeps evolving and will continue to — that's not an obstacle to PM.

Q: What lessons from China co-branding (Luckin, Meituan, etc.) can be ported elsewhere?

A: Our brand partnerships in China have been very successful. One reason is that Duolingo's IP and brand are very strong in China, attracting top brands to approach us — we'll soon co-brand with McDonald's in China.Also, Chinese brands (e.g., Luckin) are far more open to frequent co-branding than Western brands — you won't see Starbucks rotating full-store collabs every two weeks, but Luckin does. Our China partnerships and marketing teams are excellent; some lessons can be ported across Asia, though part of it is China-specific.More importantly, China's rapid growth isn't just because co-branding is good — causality is reversed: we grow fast and are seen as a cool brand, so majors come to us. China's core driver is large, growing demand for English learning.

Q: Any visible early signals for H2 bookings acceleration?

A: Sequentially, bookings progress from Q2 to Q3 and Q3 to Q4 is broadly consistent with the past two years. Q1 and Q2 reflect a transition period for new monetization tactics; Q3 and Q4 show more typical seasonality.Our core this year is operating around ~20% DAU growth and growing the biz. on that base, and current progress follows a fairly typical seasonal cadence.

<End of text>

Risk disclosure and statement:Dolphin Research Disclaimer and General Disclosure