Duolingo’s Scaling Journey: Education Revolution or Engagement Obsession?

This Week’s Focus: Duolingo’s Growth Flywheel—Can It Keep Spinning?

Duolingo has become a leader in language learning by combining gamification, personalization, and accessibility. Its AI system, Birdbrain, analyzes billions of exercises weekly to tailor lessons and improve outcomes—a powerful data flywheel driving both engagement and scalability. But sustaining growth isn’t guaranteed. With 81% of revenue from subscriptions, Duolingo must balance monetization with its free education promise. The platform also struggles to retain advanced learners, as most offerings cater to beginners. Applying the SCALE framework reveals both strengths and risks as Duolingo navigates the tension between reach, revenue, and meaningful long-term learning.

I’m writing this at the airport en route to the MSOM conference in London, where my doctoral student Xufei Liu is presenting our joint work with Ken Moon on using machine learning to enhance learning durability—ensuring people not only obtain knowledge, but also retain it and use it when needed. The project is in collaboration with a language learning app, which is NOT Duolingo!

I know it’s shocking, but there are a few others. The one we used specializes in Chinese, which, unfortunately, Duolingo is not very good at.

But as I write this, I’m wondering if there are any languages that Duolingo is good at….

I’m a Duolingo user myself. In fact, I’m on day 165 of my learning streak in French. But I wouldn’t really encourage anyone to start a conversation with me just yet.

Nevertheless, Duolingo is doing quite well financially.

So maybe the key is not teaching people foreign languages…

When research, my personal life, and scaling intersect, I simply must write about it.

To better analyze how Duolingo achieved this growth, and whether it can sustain momentum, we can apply the “SCALE” framework—evaluating whether the business is Scalable, aware of Constraints, Aligned with its mission and customers, Well-Led, and Efficient in its model. Each of these dimensions sheds light on Duolingo’s readiness to keep growing and the challenges it must navigate.

Creusons plus profondément, i.e., “Let’s dig deeper” in French (according to Google translate).

Scalability: Demand, Supply, and Data Flywheels

Duolingo has rapidly scaled into an edtech leader, reaching 104 million monthly active users (MAU), 34 million daily active users (DAU), and about 8 million paying subscribers by mid-2024. Revenues surged from $71 million in 2019 to $531 million in 2023, reaching approximately $748 million in 2024, with market cap rising from a $3.7 billion IPO valuation in 2021 to nearly $15 billion in 2024.

Duolingo’s business shows clear signs of scalability, evidenced by notable margin expansion. Operating margins improved significantly from -17.6% in 2022 to +8.4% in 2024, indicating enhanced profitability as revenues grew faster than operating costs. Customer acquisition costs (CAC) also declined sharply—from roughly $158 per new daily user in 2021 down to approximately $28 in 2024—as marketing spending decreased from 21% to 10% of revenues. This margin trajectory, supported by growing revenue (from $71 million in 2019 to approximately $748 million in 2024), illustrates Duolingo’s improving efficiency and scalability at the financial level.

On the supply side, Duolingo benefits from economies of scale primarily due to its digital and cloud-based infrastructure. Adding incremental users incurs negligible marginal costs because of efficient scaling in cloud computing, optimized content delivery networks, and streamlined operations. Additionally, as revenues have increased, Duolingo has scaled its R&D and product development efforts without proportional overhead expansion, resulting in a rising adjusted EBITDA per user—another indicator of operational leverage and scalability.

However, Duolingo’s demand-side scalability exhibits certain limitations. Although user growth has been strong, the app’s monetization per user (ARPU) has gradually declined from around $88 in 2021 to approximately $75 in 2024 (see the graph above). This downward trend suggests limited monetization leverage, particularly in emerging markets with lower pricing tiers.

Moreover, while Duolingo uses gamification to drive high engagement (34–35% DAU/MAU), the app provides relatively low switching costs. Many learners leave after achieving basic proficiency or when gamification incentives become less effective, limiting long-term retention and sustained monetization.

Finally, Duolingo demonstrates a meaningful learning curve through its data-driven AI system, Birdbrain. The system continuously improves learning outcomes by analyzing approximately 15 billion exercises per week, refining personalization and lesson effectiveness at scale. As user numbers grow, the accuracy and quality of this AI-driven learning experience improve correspondingly, representing a powerful data-driven flywheel supporting both learning efficacy and platform scalability.

Alignment with Mission and Customer Value

A critical question for any scaling company is whether its growth is aligned with its core mission, and whether the business model remains coherent with delivering value to its customers. Duolingo’s mission is “to develop the best education in the world and make it universally available.”

Initially, the firm positioned itself to deliver free, high-quality language education through a freemium model, significantly lowering barriers to language learning and attracting millions of users who might not have considered traditional, expensive options. With almost 60% of U.S. learners under the age of 30, Duolingo’s accessible, gamified, and bite-sized format successfully aligns with the preferences of casual learners, young adults, educators (via Duolingo for Schools), and universities (through the Duolingo English Test). This broad appeal underscores the company’s effectiveness in meeting diverse user needs while staying true to its core mission of universal accessibility.

However, scaling and monetization pressures introduce challenges to maintaining this alignment. With 81% of revenue coming from subscriptions, Duolingo must balance generating revenue without compromising its free educational promise. While currently managing this well—paid tiers mainly add conveniences like offline access and AI-powered conversational practice—the potential tension between user accessibility and monetization remains. Additionally, Duolingo struggles to retain “serious” learners aiming for advanced fluency, as the app primarily excels at beginner to intermediate levels. Addressing this gap through targeted advanced offerings, such as the higher-priced Duolingo Max subscription featuring GPT-4 conversational practice, may help bridge the divide, ensuring the app continues to deliver meaningful educational value while still serving casual users effectively.

But a critical question remains: Is Duolingo truly revolutionizing education, or is it mainly “gaming” the system to keep us hooked? The app is famous for its gamified approach to language learning—daily goals, streak counts, experience points, leaderboards, funny mascot animations—which some adore and others criticize. The pedagogy behind Duolingo is a blend of genuine learning science and attention-grabbing engagement techniques, making it sometimes difficult to tell which side dominates.

Duolingo’s pedagogical approach integrates established language acquisition methods, notably spaced repetition (which I’ll delve into a little later), where users revisit vocabulary at strategic intervals to reinforce long-term memory. As Klinton Bicknell, Duolingo’s AI chief, emphasizes, this adaptive strategy boosts both learning outcomes and sustained user engagement by minimizing frustration. Additionally, Duolingo employs a multimodal teaching method—combining listening, speaking, reading, and writing exercises—to support comprehensive language skills and to prioritize high-frequency vocabulary.

On the other hand, Duolingo heavily emphasizes user engagement—sometimes so intensely that it risks positioning learning as a secondary benefit. Features like daily streaks, designed to foster habit formation, can lead to compulsive usage patterns where learners prioritize maintaining their streak over meaningful practice. Duolingo also optimizes notifications using AI-driven emotional nudges, such as guilt-inducing messages, a tactic drawn from social media and gaming to maximize re-engagement.

Critics argue this focus reveals that Duolingo prioritizes app usage over deep language acquisition, with one product manager notably describing their goal as creating content that users “have fun interacting with and learn as a byproduct.” This approach captures Duolingo’s strategy of leveraging addictive engagement techniques to sustain user motivation, recognizing that consistent engagement, even if partly driven by entertainment, ultimately supports long-term learning.

Looking forward, Duolingo’s critical challenge is to maintain a careful balance with gamification reinforcing rather than undermining genuine learning. Excessive reliance on engagement tactics risks diluting educational effectiveness. Nevertheless, internal efficacy studies indicate promising outcomes, with Duolingo users performing similarly on standardized reading and listening tests compared to traditional university language students—provided they engage actively and regularly.

Leadership, Culture, and Organization

Duolingo's leadership, spearheaded by founders Luis von Ahn (CEO) and Severin Hacker (CTO), has fostered a product-centric, data-driven culture centered around rapid experimentation and strong mission alignment. The company’s hybrid organizational structure—with cross-functional teams owning specific metrics and driving rapid iterations through extensive A/B testing—enables agility and accountability. Cultural tenets—long-term thinking, relentless quality standards, quick iteration, data-driven decisions, and a playful approach—guide both internal operations and product evolution.

However, Duolingo’s centralized decision-making, notably von Ahn’s hands-on involvement, risks becoming a constraint as the company scales, highlighting concerns related to Conway’s Law—the principle that product structures often mirror organizational structures. We already see early signs of these issues: Duolingo’s app experience increasingly reflects internal divisions, with certain features (e.g., leaderboards, gamification elements, and language curricula) sometimes feeling disjointed or uneven due to fragmented ownership among teams.

For example, I get notified once in a while that I either dropped or “moved up” from a “league.” To be completely honest, I have no clue what this means and why I should care. It’s completely orthogonal to my experience and does nothing to motivate me. And I imagine I’m not alone.

As the organization grows beyond 800 employees, leadership will need to decentralize authority and empower divisional teams to preserve a cohesive user experience and efficient innovation, proactively managing the alignment between organizational structure and product coherence.

Efficiency and a Predictable Path to Profitability

Efficiency is one of the main pillars of the SCALE framework and ties closely to Duolingo’s financial sustainability. A scalable business must not only grow quickly but do so efficiently, i.e., improve margins and create a credible route to long-term profitability.

On this front, Duolingo’s recent performance provides optimism. As noted earlier, in 2024, the company already transitioned to positive operating margins—a rare achievement among consumer-facing edtech companies, many of which struggle to monetize at all. This positive turn was driven by both revenue growth and cost control. Duolingo’s freemium model yields a high gross margin because digital content costs little to replicate; once a course is developed, delivering it to the 10 millionth user is almost free. Thus, gross margins are strong, and the key is covering fixed costs (mainly R&D, personnel, marketing) with volume. The fact that Duolingo’s operating margin swung from -17% to +8% in two years indicates that it’s crossing that threshold of scale where fixed costs are being amortized over a large user/revenue base.

A crucial component of efficiency is Duolingo’s marketing engine. Initially, in the run-up to and just after the IPO, Duolingo spent significant sums on marketing (2021 saw a spike in customer acquisition cost as the company invested in growth). But since then, efficiency gains have been striking: while revenues nearly tripled from 2020 to 2024, marketing costs increased at a slower pace, leading to the drop in marketing as a percentage of revenue from 21% to 10%. In practice, and over time, each new user costs less to acquire because organic channels are doing the heavy lifting. The app’s virality and strong brand recognition (it’s consistently the #1 education app in app stores) means Duolingo doesn’t need Superbowl ads to gain users—its users and the media advertise it for free. This is an inherent advantage of a product that people like to talk about.

The declining CAC gives Duolingo headroom to invest more in product development and to tolerate lower conversion rates, confident that growth doesn’t depend on prohibitively expensive advertising. By contrast, other edtech or app companies hit a wall, pouring more and more money into advertising to sustain growth.

Duoling’s financials highlight that scaling its user base and converting users to higher-value subscriptions like Duolingo Max are key to driving substantial revenue growth without proportional cost increases, reflecting its high fixed-cost, low variable-cost model.

Duolingo’s unit economics demonstrate strong efficiency gains as the company scales. Between 2021 and 2024, revenue per user increased due to new subscription tiers like Super ($80/year) and Max ($168/year), while cost per user notably decreased, particularly in direct marketing and infrastructure. Duolingo benefits from a high contribution margin from paid subscribers, as most incremental revenue (minus app store fees) is profit due to minimal per-user costs. Even modest improvements in conversion rates or ARPU thus significantly boost profitability.

Constraints and Challenges to Continued Growth

Even though Duolingo demonstrates strong scalability, it’s not without constraints. Perhaps the most pressing constraint is monetization: the vast majority of Duolingo’s users use the app for free, so converting even a small fraction to paid plans is critical. As of 2024, only about 8.8% of MAUs were paying subscribers. While Duolingo has grown its paid subscriber count impressively (8 million and rising), this conversion rate underscores a ceiling in how much of its user base is willing to pay, especially in emerging markets.

Moreover, as Duolingo expanded globally, its average revenue per user (ARPU) for subscribers declined to around $75 per year. This is partly due to regional pricing—subscriptions are priced lower in countries with lower incomes—and the introduction of discounted options (e.g., family plans). Another, culture-related challenge is that in certain markets like India or Latin America, consumers are less accustomed to paying for app subscriptions, making conversion an uphill battle. Freemium conversion limitations mean Duolingo must either find ways to boost the paying percentage or continually grow its overall user base. To mitigate this, the company has experimented with geographically tiered pricing (to make subscriptions more affordable in lower-income regions) and with diversifying its offerings—offering “micro-subscriptions” for specific needs (grammar packs, test prep modules, etc.) at lower price points to entice more users to pay. Additionally, Duolingo is expanding B2B channels like the Duolingo English Test for universities and potentially corporate training partnerships, which can open new revenue streams beyond individual learners.

Another significant constraint for Duolingo is content scalability. Traditionally, building and maintaining language courses required extensive human resources—linguists, curriculum designers, and native contributors—which is costly and hard to scale as user demand expands. Maintaining consistent quality and depth, especially for niche languages, has been challenging. To address this, Duolingo is increasingly using AI-assisted content generation to accelerate course creation and has leveraged community volunteers and modular frameworks to reduce costs and enhance scalability without compromising its engaging educational style.

A major challenge Duolingo faces is long-term user retention and engagement. While effective at attracting beginners through gamification, many users lose interest after 3–4 months, either due to learning fatigue or because they reach an intermediate proficiency plateau. Advanced learners frequently leave Duolingo for more rigorous resources, limiting its ability to retain users who might pay for deeper learning experiences. To address this, Duolingo should enhance its advanced content by introducing adaptive, goal-oriented learning paths and strengthening community-driven features that foster motivation and accountability for long-term mastery.

Can We Learn Better? The Science of Spaced Repetition and Optimizing Learning Schedules

With scalability out of the way, let’s revisit Duolingo’s educational mission. Duolingo aims to balance education and engagement. One would claim that language apps could do better on both.

One area where Duolingo’s educational approach particularly stands out—and where future improvements could yield big gains—is spaced repetition.

The concept of spaced repetition is backed by decades of cognitive science: by spacing out review sessions for learned material, each review is timed to refresh memory right before one would forget, thereby strengthening retention more efficiently than cramming.

Duolingo’s Birdbrain AI uses this principle by predicting when a user is likely to forget a word and prompting a review of that word or concept at the right time. In practical terms, if you learned “la pomme” (“apple” in French, which by the way, I wasn’t taught even after 165 days of French) today, the app might show you again tomorrow, then a few days later, then a week later, etc., each time extending the interval as your memory solidifies. This adaptive scheduling is a key reason one can actually learn and not just play—it ensures that knowledge moves from short-term to long-term memory.

However, designing the optimal spaced repetition algorithm is a complex and ongoing area of research. In fact, our recent research sheds new light on how current algorithms might be improved, allowing for better retention while actually spending even less time on learning.

Most spaced repetition systems today (Duolingo’s included) have the goal of maximizing recall while minimizing the number of reviews, using models of memory decay to decide when to test you. There have been several generations of such algorithms: from the classic Leitner system to more advanced formulas like the “half-life regression” model and the latest FSRS (Forecasting Spaced Repetition Schedule) algorithm introduced in 2024. These algorithms typically treat each flashcard or word independently, scheduling reviews based on your history with that item.

Our research found that there’s room to push these algorithms further. Analyzing data from a language learning context, we discovered that students can retain knowledge about 15.8% longer than what the current algorithm expects on average. In other words, the spacing could be slightly more aggressive (longer gaps) without hurting recall as much as current models assume.

This discrepancy was partly attributed to the algorithms treating each recall event as independent, whereas in reality there are interdependencies and “interleaved” learning effects. For example, learning related words or practicing one skill might reinforce another (learning “apple” might also reinforce “fruit” or general memory confidence). Research suggests incorporating these interleaved learning effects into the scheduling model. By recognizing that not all practice is isolated—sometimes reviewing one item helps retain another—the algorithm can schedule reviews more efficiently.

A simple example from everyday life is: teaching someone a mathematical concept often will reinforce related skills. When explaining algebra to a peer, you simultaneously strengthen foundational arithmetic skills and problem-solving strategies. Similarly, effective algorithms should recognize these interconnected skills—leveraging interleaved learning effects to optimize spacing and improve overall retention

Furthermore, our research explores using reinforcement learning (RL) to continually adapt the scheduling policy for each learner. The idea is to personalize the spaced repetition even more granularly, learning the optimal pattern for each individual rather than a one-size-fits-all schedule. Over a long period, an RL-based scheduler might discover, for instance, that one user tends to forget things after 4 days while another can go 7 days, and tailor their review prompts accordingly.

This, of course, is (relatively) easy to implement when developing an educational app. But maybe we should start implementing this at school. Imagine coming to class, and the professor informs you that you, yes you, have a quiz today, earlier than the rest of the class, since based on the algorithm you are about to forget the topic, and it needs to be reinforced.

But maybe not.

Conclusion: Balancing Scale, Engagement, and Efficacy

Duolingo’s success story lies in effectively transforming language learning from a tedious chore into a daily habit for millions worldwide. Leveraging a smart combination of freemium business models, viral growth tactics, and ongoing product experimentation, Duolingo has built a scalable business.

However, while its use of gamification has driven widespread adoption, this same approach raises valid concerns—some critics argue that it prioritizes engagement over depth, creating numerous casual learners but fewer truly proficient speakers. Future strategies aimed at integrating deeper content, advanced curricula, and specialized certification tracks may demonstrate that scale and educational depth are not mutually exclusive.

Ultimately, Duolingo’s continued growth hinges on balancing three key forces: scale, engagement, and learning efficacy.

As the company enters its second decade, the central challenge will be ensuring that its impressive scale and popularity continue to translate into meaningful, lasting educational value rather than merely sustained engagement and screen time.