Mission, Retention and an Owl With Attitude: How Duolingo Grew From 3 Million to 200 Million Users Without Paid Ads

Gotthilf opens by describing the conference her company Latitud had just hosted for more than 5,000 people and 70-plus speakers, with Ben Horowitz doing a fireside focused on the region. She points out that Latin America has 660 million people and a roughly $6 trillion economy, yet is easy to overlook from the United States. As a Brazilian, she says she cares about helping the region's founders get the information and resources she had access to.

Her opening talk used the metaphor of a record with an A-side and a B-side. Public stories about people, companies and countries are usually the highlight reel, she argues, while the failed moments get swept under the rug. She notes that those B-moments are easy to mistake for the end of the road if you do not recognize them as moments.

She applies the same framing to the region. The B-side is familiar: instability, weak governments and financial insecurity. The A-side is that tech makes up only a fraction of GDP compared with other markets, roughly a third of India's share, a tenth of China's and a thirtieth of the United States', which leaves enormous room to grow.

Gotthilf's own B-side includes wanting to be an actress or a dolphin trainer, missing every top school she applied to, and dropping out of Reed College because of severe depression. When she finally graduated, a counselor told her the transcript showed little. She says the lesson stuck: it is not enough to do valuable things, you also have to tell the story and understand what others perceive as valuable.

After applying to a hundred companies, she landed an internship at a digital agency in New York, but the company forgot to file her visa on time. A second employer laid her off, and a third job ended in a firing. At Tumblr in Brazil she went six months unpaid because the company could not wire money, and then she was laid off when Tumblr was sold to Yahoo.

She then started an agency to help US startups grow in Latin America. Duolingo's head of marketing heard from a former Flickr colleague that Tumblr had grown in Brazil, and was pointed to her; she was 26. What began as consulting in Brazil expanded to Chile, Argentina, Mexico and beyond, and eventually became full-time ownership of growth, covering communications, social media and government partnerships before it evolved into an A/B testing engine.

Lenny adds that people underestimate how long a career is, citing a line attributed to Marcus Aurelius that life is long and we use it badly. Gotthilf agrees, recalling that at 26 she thought her career was already over.

On the Mike Bloomberg presidential campaign, Gotthilf says the team spent roughly a million dollars a day on paid ads, though she was not personally spending it. With many experienced people crowding the work on ad targeting, copy and design, she decided not to be another cook in the kitchen. She focused on what happened after the click, a step nobody was watching, and argues that the second or third screen of an experience is often as important as the first.

The spend had a real upside for experimentation. At that scale she could reach statistical significance on a landing page several times a day, compared with about two weeks for an A/B test at Duolingo. In one day she took a page from 3% to 12% conversion. The downside, she notes, is that big budgets tempt teams to lose sight of effectiveness.

The fixes were basic best practices that are easy to overlook. Make pages mobile-optimized with the core message and button above the fold, and write for people who skim by cutting copy aggressively. Make the headline and button work together so that someone who reads only those two still gets it, and keep images from distracting from the message.

She also stresses emotion. Because the campaign was arguing that Covid deaths could be avoided with a different president, darker, more menacing imagery performed better than a ridiculous-looking image of the opponent. Her point is to consider how people will feel when they see a page and whether that feeling leads to action.

Lenny notes that his most popular newsletter post is about Duolingo's growth story and asks why it succeeded when consumer subscription apps so rarely do. Gotthilf says almost nobody believed in it at first, since education was seen as unprofitable and Rosetta Stone already existed. Her first factor is an obsession with the mission: learning English can double or triple income potential in markets like Latin America, and the belief that education should be free shaped marketing, product design and hiring.

The second factor is staying lean and avoiding paid acquisition. Her mandate had no budget, and Duolingo did not begin monetizing until roughly year three, so there was no lifetime value against which to justify a customer acquisition cost. She also warns that cheap acquisition dries up as low-hanging fruit runs out, and that paid growth becomes an addiction that is hard to switch off once investors and experiments depend on it.

Third is retention, which she frames simply as whether the product provides real value. If you do not pay attention to who finds it most useful early on, you can convince yourself with acquisition metrics and fail later. She still sees a role for paid ads in testing messaging, images and value propositions, but argues that if you find your ideal user you should figure out how to reach them organically.

Her fourth item, which she calls three-B, is an obsession with the product itself. Founders who clone Duolingo's design without engineers and PMs studying the data every day are, in her analogy, like students who copy the best student's answers and then cannot solve the next problem. The rigor of questioning hypotheses and statistical significance was part of Duolingo's culture from day one.

Gotthilf says more than half of Duolingo's A/B tests failed. Her best-known mistake involved badges, which appeared in most top games and which her team wanted to try. Because the experiments were ranked by expected return, she held the idea back for six months, then approved a minimal version where users simply got a badge for signing up.

That test showed no results, since nobody is proud of signing up and there was nothing to collect or show off, and the team moved on for about eight more months. When they finally looked back, they realized they had never dogfooded their own experiments; she came from marketing and did not know the term. Badges eventually became a trove of opportunity that moved almost all of the company's metrics, including some unexpected ones.

She lists other flops: a social feature called DuoDuels, a Duolingo for Schools product that did not take off, and a China launch that drew a million downloads on day one before the app was blocked and rated one star. In India the team learned that many people set their phones to English, so the app offered learners English-based courses when they wanted to learn English. Lenny shares a parallel: he added Hindi subtitles to his podcast and was told Indian tech listeners prefer English.

Lenny argues that PMs often over-rely on user research instead of using the product and trusting their instincts. Gotthilf agrees, and adds that it is especially hard for young people, women and minorities to trust their instincts. She suggests acting as if you know the answer, and notes that when she makes mistakes as a user of Latitud's banking product, those mistakes probably reveal what other people struggle with too.

Asked for tactics, Gotthilf first points out that A/B testing and product-led growth count as organic growth because they cost no ad money. She urges even early startups to have people on the product team who can evaluate data well, noting that at that stage the opportunities are 20 to 30 percent gains rather than 5 percent. Her second tactic is brand and PR, anchored in a mission that most people can support.

A mission about giving everyone access to great education resonates far more than asking someone to download an app, she says. At Duolingo the owl, the colors and the tone of notifications and emails all tied back to how people felt about the brand. That, in turn, makes people talk about it and gives reporters a human story, unlike the pitch that an app grew 20% last month.

In practice the mission was embedded in communication. The early team learned that behind-the-scenes photos on Facebook drew no interest, so they kept a social calendar with one weekly post on effectiveness and one on the mission. A payment screen that said paying for Duolingo makes language learning accessible to millions improved conversion, which suggests the mission can appear almost anywhere.

Gotthilf loves copywriting and argues that communication is about getting the listener to receive, understand and remember a message. At Duolingo the team asked of every piece of copy whether any other company could have written it. The voice was quirky and a little funny, which made people giggle or do a double take, and she now uses the same approach in Latitud's newsletter with self-deprecating humor.

She walks through examples. Passive-aggressive notifications, including a message sent after five days of inactivity saying the app would stop sending reminders, became a meme. A New Year's email featured Sad Duo, and the team A/B tested the size of the puddle and the number of tears. When fans started making their own versions, the company chose to run with it instead of treating it as a PR disaster.

Lenny recalls the viral meme where a Duolingo reminder was followed by a home security alert. Gotthilf says leaning in was not an easy decision because many colleagues worried about brand image. She says it was a bet that a unique voice means being part of something bigger and connecting with people beyond teaching a word a day, and Lenny ties it to advice to take risks.

On Duolingo's TikTok presence, she credits the people running it, but argues the real foundation is the brand DNA and culture. Without executive-level conversations about who Duo is and how he talks, and a willingness to risk rubbing people the wrong way, such an account would not be possible. Lenny adds that hiring one clever social media person is not enough if the culture does not support that approach.

On internationalization, Gotthilf takes what she calls a controversial position: outside regulated industries, humans are more alike than different. Cultural differences exist, but she sees many of them as the final 5% of understanding people. Focusing on them too early risks making too many changes at once, learning little and adding code and organizational complexity.

Each time she launched in a new country she heard that people there were different and that the approach would not work, whether the claim was about the color green, owls or free products. Duolingo ignored most of it, and when an A/B test or marketing experiment worked in one place the team tried it everywhere. The main exceptions were China, where certain words and phrases had to be avoided, and India.

Her practical argument is about speed. If you maintain separate versions of the app for Mexico, China and India, every new hypothesis must be tested and shipped in each of them, which is costly in code, process, people and especially time. Lenny shares the line that life is maintenance, so everything you add must be maintained forever, and Gotthilf adds that startups should chase the 80 before the 20.

Gotthilf says she does not see her move to Latin America as a pivot, since she has simply moved along the stack from operator to founder and investor. Mentoring entrepreneurs through Endeavor in Japan and Brazil convinced her that the region's tech opportunity is huge, in part because so much that is already digital in the US has not been built there. She also saw that one-to-one calls did not scale, and that tips she took for granted amazed brilliant founders who lacked Silicon Valley's flow of information.

She describes structural obstacles beyond knowledge. Incorporating a company in Brazil takes among the longest times in the world, and investors often prefer Delaware entities because of the liability risk of Brazilian employment. Founders then have to work out foreign exchange and how to bring money home. She says that using her privilege to create a lasting impact where she came from is why this matters so much to her.

Latitud's pitch is an operating system for early-stage Latin American startups. It addresses the loneliness of being a founder, the weak English-speaking connection to Silicon Valley, and the lack of capital and legal infrastructure. It includes a free fellowship with about 1,500 entrepreneurs admitted across four cohorts a year, a fund that has backed 100 startups with a second fund of around $25 million being raised, and paid products such as a dashboard and a bank for incorporation and compliance.

Joining the fellowship is free and Latitud takes equity only when it invests, with products paid for separately. Customers are founders across Latin America as well as US founders building for the region.

Asked what Latin American founders do uniquely well, Gotthilf says they tend to be scrappy, resourceful and resilient because of how they grew up. Beyond that she does not claim a specific strength, but says the region is wide open in almost every vertical. Fintech is especially hot because so many people lack bank accounts and credit, and B2B opportunities abound among small businesses still working on paper.

She argues that solutions built for emerging markets can travel to other emerging markets, which is most of the world. Lenny links this to WhatsApp, and she mentions Leadsales, a Latitud investment that turns WhatsApp into a CRM, something less relevant in the US. Emerging markets also skip steps, as in markets that went straight to mobile, and she is curious how that plays out with AI.

On what to watch, she names portfolio company Pomelo, a fintech platform for fintechs, but mostly urges people to pay attention to the region. Venture money, experienced talent returning from unicorns and IPOs, and unsolved problems have come together. Her evaluation criteria are the same as anywhere: a founder with an unfair advantage, access to top talent and a TAM you actually believe, and she is especially excited about health and education.

In the lightning round, Gotthilf recommends *The Design of Everyday Things*, which showed her that design is about usability and not only beauty, and Viktor Frankl's *Man's Search for Meaning*. Her favorite shows include *The Rehearsal* and *White Lotus*. She says her favorite interview question is why a candidate wants to work at the company, followed by what they are world-class at and how they know.

She says she is curious about microbiome and gut health tests and calls out Levels, from her friend Sam Corcos. She argues that accessible health devices could shift power away from doctors as the sole authority. She says she is going through a hard time with long Covid and loss of taste and smell, which is why her motto "This too shall pass" feels especially relevant.

Her second motto is "fake it till you make it," which she defines as not lying but saying yes to opportunities and standing as your best self. She tells of saying yes when Duolingo asked her to be head of growth before she knew what growth was, and reading Andrew Chen's blog to find out. She also shares her love of the Ship of Theseus allegory as a way to think about identity as a person, company or country changes.