Duolingo Beat Its Own Investment-Year Guide on Every Line, Then Told You the Top of the Funnel Is Flat

Results vs. Consensus

Duolingo reported after the close on May 4. This was the first quarter executed under the strategy announced in November and quantified in February, and the useful question is not whether the numbers were good but whether management did what it said it would do. On the financial commitments, it did, and by a wider margin than guided. On the user commitment, the quarter is inconclusive in a specific and disclosed way.

Year-over-Year Comparison

Sequential Comparison

Revenue and bookings

Revenue of $292.0M grew 27% and beat consensus by 1.2%. Bookings of $308.5M grew 14% and beat the company's own point-estimate guide by $7.0M. Subscription bookings grew 15%, slightly faster than total, so the subscription share of the mix is still rising even during a quarter in which the company is deliberately loosening conversion pressure.

The gap between the two lines has now inverted from a year ago. In Q3 2025 revenue grew 41% against bookings at 33%, with recognition lag working in the company's favor. This quarter revenue grew 27% against bookings at 14%, so revenue is being flattered by bookings taken when growth was much faster. That gap closes mechanically through 2026 and is why management guides revenue growth to moderate through the back half even as bookings growth accelerates.

Margins

Gross margin of 73.0% is the headline surprise. Management guided approximately 71% for the quarter and delivered 73.0%, expanding 190 basis points year over year in a quarter that was supposed to show the beginning of a deliberate three-point step down. The letter attributes the outcome to continued reductions in per-unit AI costs. This is now the third consecutive quarter in which the AI cost drag has come in materially lighter than guided.

Below the gross line, the spending is visible where management said it would be. Sales and marketing grew 47%, by far the fastest of the three expense lines and the clearest financial evidence of the strategy. Research and development grew 18% and general and administrative grew only 7%, so the operating leverage that has characterized this business is intact everywhere except the line management deliberately chose to lift. Total operating expense grew 20% against revenue at 27%, which is why GAAP operating margin still expanded 500 basis points to 15.3%.

Adjusted EBITDA of $83.4M on a 28.6% margin beat the 25.5% guide by 310 basis points and expanded 140 basis points year over year. Sequentially the margin declined 120 basis points from Q4's 29.8%, which is the beginning of the guided compression, but from a much higher base than the guide contemplated.

Users, and the part that did not work

DAUs of 56.5M grew 21% against a 49% comparison that contained the Dead Duo campaign, and rose 7.2% sequentially. Paid subscribers of 12.5M grew 21%. The daily-to-monthly engagement ratio reached 41.0%, up 520 basis points year over year and 140 basis points sequentially, and has now risen in every quarter we have tracked.

Monthly active users were 137.8M, up 5.8% year over year and 3.5% sequentially off Q4's 133.1M. Two things about that figure matter. First, MAU was demoted from the headline metrics table to a footnote this quarter, with the company stating it primarily evaluates engagement using DAUs and treats MAUs as supplementary. Second, and more usefully, the CEO answered a direct question about it without hedging: top-of-funnel growth was about flat in the quarter.

Put together, the entire 21% DAU result was generated by retention and engagement depth, not by acquisition. That is a genuinely good outcome for a quarter and a genuinely insufficient one for a plan whose endpoint is doubling the user base by 2028. The 100 million target requires the funnel to widen; so far it has only densified.

Segment Performance

Duolingo reports as a single segment, but beginning this quarter the letter carries a revenues-by-product-type appendix that it did not carry in 2025. That table is the first clean read on the non-subscription streams in this coverage, and it is worth more attention than it has received.

Subscription at +31% is carrying the entire result, and the two streams that ought to benefit most from a larger free base are shrinking. In-app purchases fell 11% as the energy mechanic laps its rollout, and the English Test fell 6%, the only stream down in dollars. Advertising grew 15%, which is respectable but is the line management has said it intends to professionalize, so it is the one to hold to a higher bar from here. The composition is a reminder that the free-user strategy has not yet produced a non-subscription revenue engine.

The other disclosure that carries information this quarter is product and content velocity, which management quantified for the first time.

Speaking practice becomes the product

The February plan promised to make speaking a core part of the experience for free and paid users. This quarter delivered three separate mechanisms: spoken tokens that let a learner speak the answer to almost any exercise, flashcards that build recall by saying words aloud, and Speaking Adventures that place learners in task-based conversational scenarios drawn from the course curriculum. For paid subscribers, Video Call has more than doubled average words spoken per user over the past year.

"Speaking practice has historically been the hardest thing to do well on a mobile app. This quarter, we made it a bigger part of the experience for free users and paid subscribers." — Luis von Ahn, Co-Founder and CEO

Assessment: This is the single largest product change the company has shipped in the period we have covered, and it went almost entirely to free users. It is also the direct cause of the guided gross-margin step down, since inference cost now attaches to a non-paying base. Whether it works will show up in word of mouth, which is the mechanism with the longest lag and the least measurability of anything management is doing.

Content velocity

The company published 20,500 course units in Q1 against roughly 7,100 per quarter in 2025 and 1,800 per quarter in 2024, and completed Duolingo Score 129 coverage across the nine most-learned languages. That was a February commitment with a stated timeline of a month or two, and it was met.

"In Q1 alone, we published 20.5 thousand course units. To put that in context, that is more than 10 times what we were shipping per quarter just two years ago. AI has fundamentally changed what is possible for us, and I believe we are just scratching the surface." — Luis von Ahn, Co-Founder and CEO

Management also described a change in kind rather than degree: automated content creation lets the company push changes across many courses simultaneously and iterate on quality, which it says is already improving engagement among new users.

Assessment: An eleven-fold increase in content throughput in two years is the most concrete return on AI adoption in the business, and it is on the cost side of the ledger rather than the feature side, which makes it durable. The stated quality controls, AI plus human evaluation followed by small-sample user testing before wider exposure, are the right ones. The claim that quality has increased over the last couple of quarters is not independently verifiable.

Video Call moving into Super

February's boldest commitment was moving Video Call, the flagship feature of the top Max tier, into the roughly ten-times-larger Super tier. That has begun, but narrowly: experiments are running against cohorts of new Super subscribers, not the existing base, and Max is unchanged so far.

"At the moment, we have a number of experiments giving video call to Super subscribers, particularly new Super subscribers. We have not scaled this to all our existing user base. So at the moment, there is no change for Max." — Luis von Ahn, Co-Founder and CEO

Management added that early pricing tests show learners are willing to pay more for Super with Video Call included, without quantifying how much, and said results should be available in a quarter or two.

Assessment: This is running slower than February's language implied, and slower is fine. The pricing signal, that Super with Video Call commands a higher price, is the outcome that makes the whole move accretive rather than dilutive, and it is the first evidence in either direction. The Max cannibalization risk that framed the February debate remains untested because the test has not reached the existing base.

Monetization that is not friction

The February thesis required finding monetization levers that do not tax the free experience. The first one arrived this quarter, and it is unglamorous: longer free trials. The company has historically offered seven days and is finding that one month raises bookings without hurting daily users.

"Historically Duolingo, Inc. has given a seven-day free trial. We are finding that giving longer free trials is really good. It increases bookings, and it is good for the user. For example, a one-month free trial feels great to them." — Luis von Ahn, Co-Founder and CEO

Management flagged that it will likely test a three-month trial, and noted that such a test is only possible because the 2026 operating framework tolerates delaying a quarter of bookings.

Assessment: The first working example of the category management promised, and it validates the framework argument rather than just the tactic. A three-month free trial is precisely the kind of experiment a company managing to a quarterly bookings number cannot run, and the fact that this one can is a concrete benefit of the reset guide that nobody was crediting.

Performance marketing and Asia

Asia remains the fastest-growing region and China the standout, with management describing profitable performance marketing in China specifically and noting that China monetizes about as well as Western Europe. Brand partnerships continue at scale, with Meituan in March following Luckin Coffee, and a McDonald's partnership described as imminent. Management was careful to invert the causality: the partnerships follow the growth rather than causing it.

"I would say the main thing is we have historically underinvested in performance marketing, and we are getting a lot more professional about it this year... We are building the infrastructure for the right attribution to users and the right place after you acquire them—things a company our size should have built years ago, but we kind of ignored it." — Luis von Ahn, Co-Founder and CEO

Assessment: The 47% growth in sales and marketing is being spent partly on infrastructure that does not produce users this year. That is the correct sequencing and the reason the top of funnel was flat in Q1. The admission that a company of this size never built proper attribution is candid and slightly alarming; it also means the marketing dollar productivity from here has an unusually low base to improve from.

Key Topics & Management Commentary

Overall Management Tone: Measured and consistent with February, with none of the reset energy of the prior call and none of the defensiveness of November. Management repeated the full-year framework verbatim, converted the ranges into point estimates to help modelling, and answered the two uncomfortable questions, on flat top of funnel and on whether Max is being cannibalized, without deflecting. The weakest area was forward evidence: on nearly every initiative the honest answer was that it has been two months and the results are not in yet.

1. Every February commitment was met or beaten

The value of this quarter is as a scorecard against a plan that reset the equity three months ago. Bookings, revenue, Adjusted EBITDA, Adjusted EBITDA margin and gross margin all cleared their guides. DAU growth of 21% is at or slightly above the approximately 20% framework. Advanced content shipped on the promised timeline. Video Call began moving to Super. The buyback started.

"Q1 was about execution. We said we were going to prioritize teaching better and changes in growing users, and that is exactly what we did." — Luis von Ahn, Co-Founder and CEO

Assessment: When a company resets guidance sharply, the first quarter under the new frame is the credibility test, and this one passed on every measurable line. The market's response of a 5.6% decline suggests it was grading a different question, which is the right instinct: the plan's payoff is user growth, and user growth is the one line that came from retention rather than acquisition.

2. The Q2 bookings guide

Q2 bookings are guided to $283.5M, growth of 5.8% reported and 4.1% on a constant-currency basis. In February management framed Q2 as running slightly more than three points below the guided 11% Q1, which implies roughly 7.8%. The actual guide is about two points below that framing.

"On bookings, our expected Q2 bookings growth of about 6% reflects a tough comp. The prior-year quarter included the initial rollout of Energy, a price increase on our most popular subscription plan, and exceptional advertising performance." — Gillian Munson, Chief Financial Officer

Management named three specific components of the comparison rather than gesturing at difficulty, and guided approximately three points of bookings acceleration in Q3 with a further rise in Q4.

Assessment: The comparison explanation is verifiable and correct: Q2 2025 was the quarter Energy launched, and bookings grew 41% that quarter. That said, a two-point shortfall against a three-month-old framing is a real negative, and it is the first commitment from the February call that was not met. It is the reason the stock fell on a quarter that beat everything.

3. Gross margin keeps refusing to fall

Guided to approximately 71%, delivered at 73.0%, up 190 basis points year over year. Management still guides approximately 71% for Q2 and roughly 69% by Q4, which is the same destination it named in February with the timeline pushed out a quarter.

"In Q1, gross margin was better than we would have expected and pretty good year-over-year, even with a lot of new AI content, because on a per-unit basis the costs have come down a lot. It goes in waves: costs come down, we adopt more, and we manage that." — Gillian Munson, Chief Financial Officer

Pressed on whether the high-60s level is a floor, management said the 69% target reflects a decision to put more AI into the product rather than a cost constraint, and acknowledged the outcome could move in either direction.

Assessment: The important distinction, and management drew it clearly, is that the guided gross-margin decline is a spending decision, not a cost problem. Per-unit AI costs are falling faster than the company can deploy features. That means the margin guide is a floor management chose rather than a ceiling the market imposed, and it is the third consecutive quarter in which the conservative version of that guide proved too conservative.

4. Top of funnel is flat

Asked directly about monthly-user deceleration, the CEO gave the plainest answer of the call.

"The reality is that top of funnel has been about flat for this quarter, and we would like to accelerate it. We are working on that." — Luis von Ahn, Co-Founder and CEO

The named remedies are marketing in underpenetrated regions and product improvements that should accelerate word of mouth. Management also drew a distinction that matters for how this gets graded: retention is directly measurable and controllable, while word of mouth is free but offers far less granularity of control.

Assessment: This is the honest center of the quarter. Every good number in the print came from the controllable half of the growth equation, and the uncontrollable half did not move. Management is not hiding it, which is worth something, but a 2028 goal that depends on doubling the base cannot be reached on retention alone.

5. Monthly users demoted to a footnote

The summary metrics table dropped its MAU row this quarter. The letter now states that the company primarily evaluates user engagement using DAUs, with MAUs as a supplementary metric, and discloses the figures in a footnote: 130.2M and 137.8M as of March 31, 2025 and 2026. The underlying methodology is unchanged: the letter’s Definitions section still averages monthly MAUs across the measurement period, word for word as in the prior letter.

The substance is fine, arguably better: MAUs actually rose 3.5% sequentially off the Q4 level, reversing the sequential decline we flagged as the leading indicator last quarter. The presentation is the issue. A metric that declined sequentially in Q4 was moved out of the headline table in Q1.

Assessment: The company had the better argument available and did not need the demotion. DAU/MAU at 41.0% and rising every quarter is the correct way to argue that daily engagement is the right lens, and management made exactly that argument in Q&A. Moving the number to a footnote in the quarter after it went down invites the reading management least wants, and it costs the company disclosure credibility it has otherwise earned.

6. The tax line turns

Q1 carried a $12.1M tax provision against a $0.1M benefit a year ago, which is why net income grew 24% while pre-tax income grew 59%. More consequentially, the full-year assumption moved from an approximately 6% effective tax benefit guided in February to an approximately 18% to 20% provision guided now. No explanation was offered in the letter, and no analyst asked.

Assessment: This is a roughly 25-point swing in the full-year GAAP tax assumption within one quarter and it materially changes any earnings-based valuation of the company. It follows naturally from the 2025 valuation-allowance release, which removed the shield, but the magnitude of the revision deserved a sentence and got none. It is the single least-examined disclosure of the last three quarters.

7. The buyback in execution

The company repurchased approximately 514,000 shares for about $50.6M through May 1, roughly 1% of fully diluted shares and, per the letter, more than 100% of 2025 net dilution from equity awards. That leaves roughly $349M of the $400M authorization outstanding.

"We have a $400 million authorization, so we are willing to spend that money... On a buyback, you buy more when the stock is lower and less when it is higher. We will look at where the stock is; we think it is a great time to buy our stock." — Gillian Munson, Chief Financial Officer

Management also noted that guided free cash flow of more than $350M for the year is close to the size of the full authorization, and characterized M&A appetite as ongoing but historically small in scale.

Assessment: A February authorization with roughly an eighth executed by early May, at an average price near $98, is a real commitment rather than an announcement. The explicit statement that the company thinks it is a good time to buy the stock is the strongest signal management gave about its own view of the valuation, and it came unprompted in a quarter when the stock is near multi-year lows.

8. AI on the cost side of the business, not just the product

The CFO flagged something that did not appear in the guide's headline: internal AI costs, in operating expense rather than cost of revenue, have started rising significantly. That sits alongside falling per-unit costs in the product, so the two AI cost lines are moving in opposite directions.

"Our operating expenses: we have started to see some pretty big increases in AI costs internally, and our guidance reflects that. But there are always waves of efficiency that come with AI." — Gillian Munson, Chief Financial Officer

Assessment: Worth tracking and easy to miss. The consumer-facing inference cost is falling and the internal tooling cost is rising, which means the company's total AI bill is less favorable than the gross-margin line alone suggests. It is also the direct enabler of the eleven-fold content throughput increase, so it is buying something measurable.

9. Engineering velocity is measurably up

Asked about AI-driven engineering productivity, management gave an unusually calibrated answer that declined the popular version of the claim.

"We are finding that the number of A/B tests we can run is increasing. We believe that is because of AI usage, particularly in our engineering and product organization. The increase is not huge, but it is the first time we have seen an increase on a per-capita basis in years." — Luis von Ahn, Co-Founder and CEO

Assessment: The first per-capita productivity increase in years at a company that runs hundreds of concurrent A/B tests is a meaningful operational fact stated with appropriate restraint. Management explicitly rejected the ten-thousand-tests-per-engineer framing, which is the sort of calibration that makes the rest of its claims more credible.

10. Math nears completion

The math course now covers roughly grades two through twelve and can explain errors rather than only mark them. Management called adding subjects easier with AI but not trivial, distinguishing a text-based tutor from one with diagrams and interaction. Go-to-market remains the open question, with February's Kumon-style supplemental framing unchanged.

"We have reached a point where our math course covers pretty much all content between grades two and twelve and can actually explain things when you get them wrong." — Luis von Ahn, Co-Founder and CEO

Assessment: The product is close and the distribution is not designed. That ordering is consistent with how chess worked, which reached seven million daily users through the existing base before any acquisition effort. Math has no such natural cross-sell from language learners, so this is the growth engine whose go-to-market genuinely has to be invented.

11. Monetization is under-penetrated, and the number moved

Management repeated the framing from February with an updated figure: roughly 12% of monthly active users are paying subscribers, against the roughly 10% cited on the prior call and a stated comparison of close to 50% at a large peer freemium subscription business.

"We are at the same time under-monetized and over-monetized. Roughly 12% of our monthly active users are paying subscribers. We think that number should be much higher." — Luis von Ahn, Co-Founder and CEO

Assessment: Paid penetration rising from roughly 10% to roughly 12% during a period in which the company is deliberately reducing conversion pressure is a better outcome than the strategy required. It is the quietest good news in the quarter and it directly contradicts the reading that the pivot is a demand problem wearing a strategy costume.

Guidance & Outlook

The shape of the revision is the story. Full-year Adjusted EBITDA went up by $5M above the prior high end and the margin target rose 70 basis points, while full-year bookings and revenue point estimates landed slightly below the prior midpoints though comfortably inside the ranges. Profit is better than planned; the top line is tracking at the low-to-middle of what was promised. That is the profile of a company whose cost assumptions were conservative and whose growth assumptions were not.

Implied Q-over-Q ramp: The $1,280M full-year bookings estimate less Q1's $308.5M and the Q2 guide of $283.5M leaves roughly $688M for the second half. Management guided approximately three points of bookings acceleration in Q3, which puts Q3 near $307M on a growth rate around 9%. That leaves roughly $381M for Q4, implying growth near 13%, a four-point acceleration from Q3 and the fastest of the three remaining quarters, though still just below the 14% Q1 delivered. The entire full-year bookings guide therefore depends on a Q4 that has to be the best growth quarter of 2026, in the quarter that contains the New Year promotion.

Street at: Consensus sat at $288.5M for Q1 revenue and $0.79 for EPS, both cleared. The forward comparison that matters is not against a consensus figure but against management's own February framing, and there the Q2 bookings guide is about two points light while the Q2 margin guide is roughly 150 basis points better.

Guidance style: The new CFO said in February that she intended to narrow the gap between guidance and actuals. On this evidence the intention has not yet taken effect: Adjusted EBITDA beat its guide by 13.3% and gross margin by 200 basis points. The conversion of full-year ranges into point estimates is a real step toward that goal and makes the guide easier to grade, but the cushion in the profit lines is still substantial. The cushion in the bookings line is not.

Analyst Q&A Highlights

What is actually driving daily-user growth

The opening question asked for the composition of DAU growth across performance marketing, word of mouth and other drivers, plus regional detail. The answer identified retention rather than acquisition as the working lever, and pointed to the engagement ratio as the evidence.

Q: "Could you talk to some of the different drivers of DAU growth this quarter and maybe entering Q2? Whether it is performance marketing, word-of-mouth maybe starting to return, or something else? And could you also talk to what regions you are seeing any particular strength or weakness?"
— Wyatt Swanson, D.A. Davidson & Company

A: "Historically, another place where DAU increases is just improvements in retention. That is the work of making the product stickier. That has gone really well. We have been making a lot of changes to the product, some small, some larger, that make retention higher. The way you would see that is an increase in our DAU-to-MAU ratio, which keeps increasing pretty much every quarter, and it increased again this quarter."
— Luis von Ahn, Co-Founder and CEO

Assessment: The answer names retention first and marketing second, and the numbers support the ordering. It also implicitly concedes that acquisition contributed little, which the next exchange made explicit. A 41.0% engagement ratio rising every quarter is a real asset; it is not the asset the 2028 goal requires.

The top of the funnel

The most important exchange of the call went straight at monthly-user deceleration and whether the funnel has to widen for the daily-user plan to work. The answer conceded the point without qualification.

Q: "I would love to talk about MAU growth and top of funnel at large. Help us understand the deceleration there—understood the comp and everything from last year—but how do we think about what has been the deceleration and whether that needs to accelerate to support DAU growth?"
— Andrew Boone, Citizens JMP

A: "The reality is that top of funnel has been about flat for this quarter, and we would like to accelerate it. We are working on that... Historically, the main driver of top of funnel has been word-of-mouth. Word-of-mouth is beautiful because it is free, but we do not have that much control over it in the way we can measure retention."
— Luis von Ahn, Co-Founder and CEO

Assessment: A candid answer to the question that decides whether the strategy works. The distinction drawn, between a controllable retention lever and an uncontrollable word-of-mouth lever, is also an admission that the primary mechanism of the 2028 plan is the one management can least steer. This is the exchange to reread in twelve months.

What AI scaling means for margins

One line of questioning asked both halves of the AI question at once: what has been learned operationally, and what it implies for the long-term margin structure. The CFO's answer separated two cost lines that are moving in opposite directions.

Q: "What have been the key lessons so far in terms of scaling AI, both in terms of the user experience as well as the scale of content for the platform over the last couple of months? And, Gilian, as AI scales on both sides of that equation, how should we think about what that means for margins longer term?"
— Eric Sheridan, Goldman Sachs

A: "You see our gross margin guidance has us landing at about 69% in the fourth quarter, and that assumes we are going to put a lot more of that ingredient in our product. Two, our operating expenses: we have started to see some pretty big increases in AI costs internally, and our guidance reflects that. But there are always waves of efficiency that come with AI."
— Gillian Munson, Chief Financial Officer

Assessment: The disclosure of rising internal AI costs in operating expense was volunteered rather than extracted, and it is not in the guide's headline. The wave framing is right and it cuts both ways: the company has now been on the favorable side of that wave for three consecutive quarters, which is why gross margin keeps beating.

Whether Max is being cannibalized

February's most consequential product decision was moving Video Call down a tier. The question asked whether Max subscribers are already migrating. The answer was that the experiment has barely started.

Q: "Can you discuss the affiliated impact on Max overall? Are you seeing Max subscribers cross back down to Super? How should we think about the product go-to-market for Max now that AI is becoming more available across the broader ecosystem?"
— Bryan Smilek, JPMorgan

A: "In terms of metrics we are not seeing a big difference, except that a cohort of new users is not even seeing Max—they are only seeing Super—and that is one of many experiments."
— Luis von Ahn, Co-Founder and CEO

Assessment: The bounded, staged rollout management promised in February is what is actually happening. The risk that framed the February debate has not yet been run, which means it has neither been realized nor retired. Anyone modelling a Max cannibalization hit in 2026 is modelling something that has not started.

When monetization resumes

A question on what signals would tell management it is on the right path, and when to start monetizing again, produced the most substantive forward disclosure of the call.

Q: "Historically you have a bit of a paradox: you over-monetized historically, and now maybe we are in a duration where you are under-monetizing the user base. Stepping back, what signals are you looking at today that inform you you are on the right path? And more broadly, when would be the right time to start monetizing again?"
— Ralph Schackart, William Blair

A: "This year is for experimenting. We will probably experiment with a three-month free trial. That is something we could not have done without operating with the guidelines we set for this year, because a three-month trial delays bookings by a whole quarter. This operating approach gives us room to do that."
— Luis von Ahn, Co-Founder and CEO

Assessment: The clearest articulation yet of what the reset guide actually bought. A three-month free trial is a bookings deferral that a company managing to quarterly numbers cannot test, and management is explicitly citing the reset as the enabler. It is also a warning: experiments of that shape will keep suppressing reported bookings in the periods they run.

Where gross margin settles after this year

A question on the durability of the guided high-60s gross margin got a clean answer about which direction the uncertainty runs.

Q: "Your guidance implies they are going to phase down to the high 60s in the fourth quarter. Is there a reason to think margins hold at that level? Is there a reason to think they should recover higher? How should we think about the trajectory after this year?"
— Mark Mahaney, Evercore ISI

A: "That is why we have margin guidance down to 69%, which is essentially where we were last quarter too. That implies a lot more AI content, which we think is great long term. It is possible we could optimize more, but we want to be putting that much AI in the product. I think 69% is a good place to think of us exiting the year."
— Gillian Munson, Chief Financial Officer

Assessment: The answer establishes that 69% is a spending target rather than a cost outcome, which is the distinction that determines whether the market should treat it as a permanent impairment or a discretionary reinvestment. It is the latter, and it is reversible at management's option in any quarter it chooses.

Capital allocation and the buyback

A follow-up on the balance sheet asked for the right level of repurchase against other uses of cash. The answer was more specific about valuation than management usually gets.

Q: "Thinking about the balance sheet—you have a lot of cash and high free cash flow. What are your thoughts on the right level of buybacks, and what are some central uses of that cash going forward?"
— Nathaniel Feather, Morgan Stanley

A: "On a buyback, you buy more when the stock is lower and less when it is higher. We will look at where the stock is; we think it is a great time to buy our stock. It is a great way to offset dilution from the last couple of years. With our free cash flow estimates, we are going to generate almost as much cash as that buyback anyway this year."
— Gillian Munson, Chief Financial Officer

Assessment: Management stating on the record that it is a great time to buy the stock, while executing at an average price near $98, is the most direct valuation signal available from inside the company. The observation that the year's free cash flow roughly funds the entire authorization frames the buyback as self-financing rather than balance-sheet-consuming.

What They're NOT Saying

1. Why the full-year tax assumption swung roughly 25 points: February guided an effective tax benefit of approximately 6%; this letter guides a provision of approximately 18% to 20%. The change appears in a single sentence with no explanation, and no analyst raised it. It changes every earnings-based valuation of the company.
2. Any quantification of what "top of funnel about flat" means: no new-user number, no installs figure, no registration growth rate, and no target for what it needs to reach to support the 2028 goal. The most important variable in the plan has no disclosed metric attached to it.
3. What the longer free trials cost in the near term: management said one-month trials increase bookings and may test three months, explicitly noting a three-month trial delays bookings by a quarter. No estimate was given of how much bookings deferral is embedded in the 2026 guide from trial-length testing.
4. How much Super with Video Call can be priced up: management said learners are willing to pay more and declined to say how much, promising results in a quarter or two. That figure determines whether the tier change is accretive or dilutive.
5. Max as a disclosed metric: Q3 2025 gave Max at 9% of subscribers with bookings doubling. Two quarters later, in the middle of a change that directly affects Max, there is still no updated share, renewal rate or ARPU figure.
6. Chess, math and music user numbers: chess was disclosed at approximately 7 million daily users in February. No update this quarter, on any of the three subjects management calls its next growth engines.
7. The Duolingo English Test, on the call: a third consecutive quarter with no mention by management or any analyst. The appendix does disclose the revenue, at $11.3M and down 6% year over year, the only stream to shrink in dollars. What is absent is any explanation of why a business management called a credentialing asset two quarters ago is now declining.
8. How advertising is being grown: the appendix discloses advertising revenue of $20.6M, up 15%, the fastest-growing stream after subscription. Management gave no account of what is driving it, no split between direct and network inventory, and no framing of the Q2 comparison it separately flags as containing exceptional prior-year advertising.
9. Marketing spend composition: sales and marketing grew 47%, the fastest line in the business, and management described building attribution infrastructure alongside a multi-tiered approach. No split between brand, performance and infrastructure was given, so the productivity of the incremental dollar cannot be assessed.
10. Why MAU moved to a footnote in this particular quarter: the letter states the DAU-primary rationale but does not address the timing, which follows the first sequential MAU decline. The footnote also describes the figures as “as of” the period-end dates while the Definitions section still specifies a period average, so the letter is internally loose about its own basis.

Market Reaction

• Pre-print setup: DUOL closed at $110.23 on May 4, down 37.2% year to date against the S&P 500's +5.2% and down 77.3% over the trailing twelve months. Notably, the stock had rallied 14.2% over the trailing thirty days, from $96.54 on April 2, so it entered the print with momentum rather than capitulation. The 52-week closing low of $90.03 was set earlier in the window.
• Next-day session: the stock opened at $108.39 on May 5, 1.7% below the prior close, traded a range of $98.50 to $108.88, and closed at $104.03, down 5.6% or $6.20.
• Volume: 6.0M shares against a 30-day average of 1.7M, or 3.6x normal, materially lighter than the 8.1x on the February print and the 10.1x on the November print.
• Market context: the S&P 500 rose 0.8% on the session, so the full move and then some was idiosyncratic.
• Pattern: unlike either prior reaction, this session had no gap. The stock opened near unchanged and sold off through the day, reaching its low intraday before closing 5.6% above it.

The absence of a gap is the most informative feature. November and February both produced overnight repricings of 25% and 21% respectively, which is what happens when a disclosure invalidates a model. This time the stock opened 1.7% lower and then drifted down through the session, which is what happens when a print is fine and the forward detail is slightly disappointing. The market read the release and the call, not the headline.

Three declining volume multiples across three consecutive reaction days, 10.1x then 8.1x then 3.6x, describe a shareholder base that has largely finished turning over. The holders who could not own a low-growth investment year have sold; the ones left were not surprised by a 5.8% Q2 bookings guide.

The identifiable trigger is the combination of the Q2 bookings guide landing about two points below February's framing and the CEO stating that top-of-funnel growth was flat. Everything else in the release was better than promised. That the stock still fell on a quarter where profit guidance was raised above its prior range says the market is now grading this company on users, which is the correct thing to grade it on and is exactly what management asked for.

Street Perspective

Debate: Is a flat top of funnel a failure of the plan or its expected first quarter?

Bull view: The bull case being made is that the strategy was announced ten weeks before the quarter ended and its primary mechanism is word of mouth, which management has repeatedly described as the slowest-acting lever in the business. Management explicitly told investors in February to expect only modest returns late in the year. A flat funnel in Q1 is precisely what was guided.

Bear view: The bear camp contends that the free experience has already been improved materially, with speaking features shipped to all users and content up eleven-fold, and that if any of it were going to move acquisition it would have shown a first flicker by now. Flat is flat, and the 2028 goal requires the funnel to roughly double.

Our take: The bulls are right on timing and the bears are right that the clock has started. Word of mouth from a better free product plausibly takes two to four quarters to register, so Q1 carries almost no information. What Q1 does establish is the test: if the funnel is still flat by the Q3 print, the strategy is not working on its own stated mechanism, and the 2028 target moves from ambitious to implausible.

Debate: Does the gross-margin beat mean the investment year is cheaper than advertised?

Bull view: The optimists point out that gross margin has now beaten its guide for three consecutive quarters, that full-year Adjusted EBITDA was raised above the prior range on the strength of it, and that management describes the 69% target as a spending choice rather than a cost constraint. If per-unit AI costs keep falling faster than features are deployed, the profit sacrifice is smaller than the February guide implied.

Bear view: The skeptics note the CFO also disclosed that internal AI costs in operating expense are rising significantly, which is not visible in the gross-margin line, and that management has repeatedly said it intends to spend any margin relief on more product rather than let it fall through. A cost tailwind that management is committed to consuming is not an earnings tailwind.

Our take: The bears have the mechanism right and the conclusion wrong. Management does intend to spend the relief, but it raised the full-year Adjusted EBITDA guide by $5M above the prior high end anyway, which means not all of it is being consumed. More importantly, a discretionary margin sacrifice is worth strictly more than an involuntary one, because it can be stopped. That optionality is not priced.

Debate: Does the second-half bookings acceleration hold?

Bull view: The bull argument is that the guided path is arithmetically undemanding: Q3 needs about three points of acceleration off a Q2 depressed by the Energy comparison, and Q4 is seasonally the strongest quarter with the New Year promotion. Management characterized the Q3-to-Q4 progression as typical seasonality rather than a required inflection.

Bear view: The bear response is that the implied Q4 growth rate near 13% has to be the strongest of the three remaining quarters and comes on top of a Q4 2025 that already grew 24%, and that management's own late-year assumptions include some benefit from investments it has told investors not to bank on.

Our take: This is the closest debate and the one that decides the next two prints. Management framed the second half as ordinary seasonality rather than a hoped-for inflection, and the constant-currency figures suggest currency contributes a point or two. The honest position is that the full-year bookings estimate is achievable but has no cushion, and the Q2 print in early August is where it gets tested.

Model Update Needed

Valuation impact: At the $104.03 reaction close and 49.7M estimated diluted shares, the equity is worth roughly $5.2B. Netting $1.25B of cash and short-term investments gives an enterprise value near $3.9B, essentially unchanged from the February note despite a higher share price, because the cash balance grew. That is roughly 3.1x the full-year bookings estimate, 3.3x revenue, 12.6x Adjusted EBITDA and about 11x guided free cash flow. We maintain the $140 target, which applies roughly 4.5x to full-year bookings and 18.4x to Adjusted EBITDA, and implies +35% from the reaction close.

Thesis Scorecard Post-Earnings

Overall: The thesis strengthened on the financial pillars and stalled on the one that matters most. Five of six pillars moved in a favorable direction, including the monetization pillar we downgraded in February, which recovered on paid penetration rising during a deliberate conversion pullback. The exception is the funnel. Management's own framing, that word of mouth is the primary growth driver and the one it can least control, means the central bet has not yet produced any evidence in either direction.

Action: Maintain Outperform. The February upgrade rested on three things: that the cost of the pivot was now quantified, that the price already discounted it, and that management would execute the plan it described. Q1 delivered on all three, and the full-year profit guide went up rather than down. The valuation has not moved, since the share price recovery was matched by cash accumulation, so the enterprise value is where it was in February against a better set of numbers. We are watching one thing between now and the Q3 print: whether the funnel widens. If it does not by the third quarter, the 2028 target is at risk and the rating goes with it.