Duolingo (Minutes): Accelerating the advancement of Max, profit impact is only temporary

The following is $Duolingo(DUOL.US) the minutes of the earnings call for Q4 2024, for an interpretation of the earnings report please refer to Can the Green Bird Still Fly Amid the Cooling TikTok Refugee Wave? .

I. Key Information Review of the Earnings Report

Full-year performance for 2024: In the 3.5 years since its IPO, Duolingo has added 30 million daily active users and 80 million monthly active users, with bookings nearly tripling, and an adjusted EBITDA margin of 25.7% for 2024.

Q4 2024 performance: Daily active users reached 40 million, a year-on-year increase of 51%. Total bookings in Q4 increased by 42% year-on-year, and revenue grew by 39% year-on-year, with an adjusted EBITDA margin for the full year improving by about 8 percentage points.

Full-year guidance for 2025: Bookings are expected to grow by 25% year-on-year (midpoint), with a 27% increase at constant exchange rates. Subscription bookings are expected to grow by about 31%, with total bookings expected to exceed $1 billion this year.

Q1 2025 guidance: Bookings are expected to grow by about 28% year-on-year, with a 32% increase at constant exchange rates, primarily due to an expected 35% year-on-year growth in subscription bookings. Daily active user growth is expected to be around 40%.

Margin expectations: The adjusted EBITDA margin is expected to expand by nearly 200 basis points to 27.5% in 2025, with an expected incremental margin for the full year of 30% to 35%. The adjusted EBITDA margin for Q1 is expected to be 25%.

Gross margin impact: The gross margin is expected to have a year-on-year impact of 170 basis points in 2025, with about 300 basis points in the first half, primarily due to prioritizing product innovation to promote the adoption of Duolingo Max. In the second half, gross margins are expected to improve with the reduction in AI costs.

Share dilution: The fully diluted share count at the end of 2024 is expected to be 49.5 million shares, with an estimated dilution of about 1% in 2025.

II. Detailed Content of the Earnings Call

2.1 Key Information from Executive Statements

1. Significant user growth: Since its listing three years ago, Duolingo has added 30 million daily active users and over 80 million monthly active users. By the end of 2024, daily active users reached 40 million, a year-on-year increase of 51%, with new subscription user numbers hitting a record high. In Q4 2024, Duolingo Max subscriptions exceeded expectations, and the family plan showed strong momentum, accounting for 23% of total subscription users, indicating a higher retention rate and lifetime value (LTV) compared to individual plans
2. Product performance is impressive: Since the launch of the video call feature, user engagement for Duolingo Max has significantly increased, currently covering most daily active users, accounting for 5% of total subscribers. The family plan performs strongly, making up 23% of total subscribers, with retention rates and user lifetime value higher than individual plans. The company plans to leverage generative AI to enhance the Video Call experience and accelerate content expansion for language, math, and music courses. AI costs are primarily related to Video Call and will scale with the growth of Max subscribers. The company is currently focused on providing a high-quality experience using the latest AI models rather than cost optimization.

3. Business focus for 2025:

Driving subscription bookings growth: The company aims to drive subscription bookings growth by increasing the number of users, improving subscription conversion rates, and promoting Duolingo Max to more learners globally, conducting hundreds of experiments each quarter.

Leveraging generative AI: The company will utilize generative AI to improve the video call experience, especially for advanced learners. At the same time, AI and automation tools will be used to expand the content of language, math, and music courses more quickly, with a combined daily active user count of 3 million for math and music courses, indicating significant growth potential.

Maintaining investment discipline: The company aims to balance strong revenue growth with steady progress towards long-term profitability, maintaining discipline in investments.

2.2 Q&A

Q: With Max's paid user proportion at 5%, which user groups are experiencing the strongest growth in language or global markets (GL)? How many of these users are new users, and how many have transitioned from Super subscribers?

A: The market sources for Max users are roughly similar to those for Super and Duolingo, with some exceptions, such as a higher proportion of Max subscribers in Japan compared to Super subscribers. In terms of user groups, both new subscribers and upgrades from existing Super subscribers have shown good user adoption, playing a significant role in the growth of Max users to 5%. English learners particularly enjoy the "Video Call with Lily" feature, which is used twice as frequently as by non-English learners, leading to a slightly higher inclination for English learners to purchase Max compared to Super.

Q: Is this year's AI investment mainly focused on Duolingo Max and voice calls, or are there other areas of AI investment? What is the pricing strategy for Max targeting English learners, and will prices be raised in underdeveloped regions? Is the expectation for a higher conversion rate or a focus on sales?

A: The AI investment is primarily reflected in two areas.

First, investing in AI to automate internal company affairs, which reduces costs and accelerates content creation speed, with the amount of generated content increasing tenfold or more compared to two years ago;

Second, for functional experiments or real-time features, such as the video call feature, which will increase expenses. Currently, the focus is not on cost optimization but on quickly developing the best features, as the costs of querying large language models are decreasing monthly, indicating significant optimization potential in the future Currently, Duolingo Max has been launched in most countries, but in some poorer countries, the price is relatively high; for example, $70 per year in India is too expensive. However, the pricing principle is to avoid losses, and it is expected that prices will be lowered in the future without incurring losses.

Q: How broad is the foundation for daily active user (DAU) growth? How does the growth rate in international markets compare to the overall report? The gross profit margin seems to be temporary, taking advantage of market opportunities. Are there any structural factors in the long term that prevent it from recovering to historical profit margin levels?

A: The growth of daily active users is good, with a year-on-year increase of 51% this quarter, and growth is occurring in every region. The growth rates in different regions are not strongly correlated with market maturity; for example, the most mature region, Latin America, has a year-on-year growth of 80%, indicating that markets are far from saturated. The changes in gross profit margin are temporary, with noticeable effects in the first half of the year, and optimization will occur in the second half. It is expected that the gross profit margin will roughly return to past levels in the second half. Although there is a structural difference due to marginal AI costs, it is believed that as costs decrease, the gross profit margin will ultimately be at a good level, and there will still be market profit margin expansion, albeit at a slower growth rate than in previous years.

Q: Based on the current scale, how do you view the use of artificial intelligence to further enhance the content of math and music courses, and how do you consider long-term monetization strategies for these two types of courses beyond family plans?

A: We are very excited about the math and music courses, with approximately 3 million daily active users currently studying these two subjects, and their growth rate is faster than that of language learning courses. It is expected that their share in the overall business will continue to increase. For math courses, AI will play a significant role.

Currently, the content of the math course is roughly at the level of grades 3 to 5, and we hope to cover K-12 and even some college-level math content. Previously, generating math content was slow because new exercises needed to be created for each topic, and there were significant differences in exercises across different topics. However, recent large language models have increased reasoning capabilities, and they are performing quite well in math. It is expected that in about a year, the speed at which we add content to math courses will significantly increase, bringing in more users.

For music courses, a large amount of content will also be added, but the transformative impact of AI may not be as significant as in math courses. Currently, the monetization methods for math and music courses are the same as for language courses, meaning ads will be displayed at the end of the course, and users can pay for a subscription to avoid ads; each mistake results in losing a life, which can also be resolved by purchasing a super membership. For now, it is expected that the monetization methods will remain unchanged, and the revenue from these two courses can be roughly estimated based on the proportion of daily active users.

Q: With a Max penetration rate of 5%, is the penetration rate for all English learning channels roughly the same, and is there a higher level of penetration? Are there any non-English channels that have achieved success with Lily and video call features? The newly released course unit release chart, how does it relate to other key performance indicators in the business (such as user engagement or subscription numbers) in the long term?

A: The Max penetration rate is 5%, and it is not uniform across different countries; it has a certain correlation with the super membership penetration rate, which is generally higher in wealthier countries. The penetration rate for English learners is slightly higher than that for super members; for example, in Japan, which is both wealthy and has a large number of English learners, the penetration rate, although still in single digits, is above 5%. However, it is not only English learners All language learners have a certain penetration rate. Regarding the release of course units, due to the application of artificial intelligence, we have automated the content publishing process, significantly increasing the publishing speed.

The impact on key performance indicators is that, on one hand, new courses have been added, such as Spanish courses for Korean users, which can attract more users; on the other hand, adding content to the advanced sections of existing courses can also attract more advanced learners, thereby increasing the total number of users. Currently, a large amount of content released is aimed at more advanced English learners, and we are very excited about this opportunity.

Q: Did this quarter's marketing activities attract more new users or recall revived users? How is the growth of new and revived users in the Japanese and Latin American markets, and which user growth method will be more effective in the coming quarters?

A: Generally, people rarely stop using Duolingo completely after they stop; most will return after a few months or a year. As the business matures, the proportion of revived users at the top of the funnel will increase, which is related to the operational maturity in each country. For example, in the United States, where operations have been running for a longer time, the proportion of revived users is relatively higher than that of new users; whereas in countries like India, where we have only recently started focusing our efforts, the number of new users still far exceeds that of revived users.

Currently, the number of revived users at the top of the funnel is slightly more than that of new users, but the two numbers are close. Our marketing activities mainly involve social media promotions, such as activities on platforms like TikTok, YouTube, and Instagram, which are effective in attracting both new and revived users, as we inquire about the sources of both new and revived users. However, we also need to improve the retention rate of revived users; currently, the retention rate of revived users is not as good as that of new users, partly because we do not assume that users have forgotten all content after being away for a long time, and there is room for improvement in this area.

Q: The company has undertaken a lot of content-related work; how is this related to generative AI testing? What is the testing speed expected in 2025?

A: The company typically runs A/B tests to improve core metrics, currently focusing on getting more users to subscribe to Max. The test content includes when to promote Max, to whom to promote, whether to promote to subscribed users, the frequency of promotion, and the content of the promotion. Some very effective experimental directions have been identified, such as using video call advertisements to attract people to subscribe to Max, as showcasing what video calls look like is very appealing. The number of tests conducted by the company in 2024 is more than ever before, and we believe that the number of tests in 2025 will far exceed that of 2024, with good testing speed and planning.

Q: As inference costs continue to decline, how will the company return the saved costs to users? What additional gains have been made in converting users from casual use to higher levels of proficiency regarding video calls?

A: The company has three tiers: free, Super, and the highest tier Max. Currently, due to inference costs, especially the costs of video calls, the video call feature can only be placed in the highest tier, but this may not be a long-term situation. As costs decrease, it may be placed in lower tiers, or even partially free. The company will conduct a large number of tests to determine the most beneficial plan for the long-term development of the application In terms of improving user proficiency, the company is very satisfied with the effectiveness of teaching conversations through video calls. Previously, there was no good way to practice conversations, but now there is, and it is very effective.

Q: How specifically is Lily being made more dynamic and interactive? The average subscription revenue from users has shown a turning point; can it be assumed that this revenue will continue to grow due to the increased adoption of Max and family plans? Is the slightly lower advertising revenue and pressure on gross margins due to the company's lack of focus on advertising or are there issues that need to be addressed?

A: Currently, Lily is not yet the best friend of users. The company hopes that users are willing to communicate with Lily, which is a feature that distinguishes the company from other learning products. The specific approach is that every time a user interacts with Lily, she has interesting content to share, remembers what the user said last time, asks the user questions, etc.

The current user engagement data with Lily is good, and it is expected that the engagement data will improve by the end of the year. Regarding average revenue per user (ARPU), it was previously expected to remain flat or increase, and it has actually increased. In the coming years, it is expected that transitioning more users to family plans and Max will lead to a greater increase in ARPU. The full-year guidance assumes that ARPU will maintain positive year-on-year growth. Regarding advertising, the company is a subscription-based business, and most of its efforts are focused on optimizing subscription business experiments.

The poor performance of the advertising business in the fourth quarter is due to the combined impact of advertising volume and revenue per thousand impressions (RPMs). In terms of advertising volume, the company has displayed more Duolingo ads to attract users to purchase Super or Max, so it is expected that there will not be a fundamental improvement in the advertising business, as the company mainly focuses on subscription business.

Q: Where will Duolingo's future user growth come from? Which markets or user groups are the next growth avenues? Which areas currently have low penetration? How can growth be achieved in these areas?

A: Currently, there is user growth in various regions, and the maturity of the region is basically unrelated to the growth rate. Mature regions like Latin America are also among the fastest-growing areas, indicating that the market is far from saturated.

Asia is the region with the lowest penetration, as the business started later, with huge opportunities in Japan, South Korea, India, and China. The company is establishing a marketing system in Asia similar to its marketing approach in the United States, mainly using social media hype, which has been very effective. For example, Duolingo's owl character has done "fake death" marketing in markets outside of Japan, as joking about death is not appropriate in Japan, so there is no "death" setting for it there.

Q: How did the 4-day promotional period perform? Did the performance of the Max plan during the promotional period meet expectations? Is user acceptance of the Max plan higher compared to the Super plan? What was the package combination situation during the promotional period?

A: The overall performance during the promotional period exceeded expectations, with the Max plan and family plan performing better in the last 4 days of the year, representing a comprehensive outperformance. However, the performance for the entire quarter was strong, not just in the last 4 days or during the promotional period; there was strong user growth throughout the quarter, and many regular monetization experiments also achieved results beyond expectations Q: What languages are available in the Duolingo Max tier application? Is there an ideal number of languages? How is the language rollout situation? Which languages are easier or more difficult to add?

A: Currently, the most in-demand major languages are offered in the Max package, such as English, French, Spanish, Japanese, and Chinese. User demand for languages is very unbalanced, with about half of users learning English, nearly 20% learning Spanish, and nearly 10% learning French. These three languages account for about 75% of users, while users of the other approximately 37 languages account for only 25%.

The Max package covers most major languages, including the top 8 languages. Adding non-Romance languages is more difficult, such as Japanese and Chinese, because these languages are harder for people from Western countries to learn. For example, setting up video call features for beginner Chinese learners makes it difficult for them to understand the content; whereas for Spanish, many people can understand some simple expressions even without having studied it. Therefore, when adding features for such languages, the rollout may be delayed or further simplified.

Q: Why is the revenue growth rate expected to decline slightly? What penetration rate might the Max package reach by the end of the year? What are the guidance expectations?

A: The decline in revenue growth rate is a mathematical result of the booking rate calculation; the revenue for any quarter essentially reflects the composite of the booking growth rates from the past four quarters, which is a mathematical outcome of booking guidance. Additionally, the annual booking guidance has a significant foreign exchange impact, which will gradually reflect in the revenue guidance.

Regarding the penetration rate of the Max package, it is currently known to be above 5% and is growing daily, with continued growth throughout the year, but it is uncertain how high it will reach. The company expects a considerable booking volume for the Max package in 2025, but it is still in the early stages, with much work to be done in terms of functional innovation, presentation methods, pricing, and artificial intelligence costs.

Q: How has user interaction with Duolingo changed compared to intermediate English learners? How will a greater focus on intermediate English learners change Duolingo's marketing approach and flexibility for reinvesting in products and innovations?

A: Duolingo relatively recently (about two or three years ago) decided to focus on intermediate English learners; previously, English courses were mostly aimed at beginners. Now, there is content aimed at more advanced English learners, but it takes time for people to start interacting with this content. Duolingo primarily relies on word-of-mouth to grow; in the English learning market, people previously thought Duolingo was suitable for beginners but not for advanced learners because relevant content did not exist before.

Currently, it is necessary to ensure that people are aware of this content, and some marketing efforts will be made to promote it. Duolingo needs to change some marketing strategies, and interaction with content aimed at more advanced English learners is increasing. The number of advanced English learners is growing significantly, and it is expected that there will be more by the end of this year and even more in the following year.

Additionally, this year, absolute spending on marketing is higher than last year, and the sales and marketing expenses still have a leverage effect because their growth rate is lower than the revenue growth rate, leading to strong confidence in the reinvestment rate in marketing Q: So far, how does Max's early user retention rate compare to Super? Considering advanced usage features like video calls, how do you view Max's user retention rate in the medium term?

A: So far, the retention rate looks good, but it is still in the early stages. There are methods to estimate it, and it feels positive. Max is undoubtedly the highest tier in terms of lifetime value (LTV), and the more users that convert to this tier, the better. We hope that as features improve, retention rates will also rise over time.

Q: In 2025, which key investments does the team believe will put some pressure on incremental profit margins? Over the past few years, Duolingo's profit margins have exceeded guidance and performed better. In which areas of the business might it exceed profit margin guidance rather than align with expectations?

A: In less significant expenditure items, a considerable amount of initial investment has gone into sales and marketing, and hiring has also accelerated because we see related opportunities that are helpful. The largest and most significant investment and cost increase that has led to a 200 basis point expansion in profit margins is the marginal cost of Max, which is a key driver of profit margins, indicating that Max will grow well, benefiting both revenue and the ultimate LTV of the platform.

While we believe that Max's costs will decrease over time, the team's current focus is on quickly increasing user adoption at this tier. In terms of profit margins, a potential positive factor is that AI costs may decrease faster. Currently, incremental profit margins are within the long-term profit margin range, and we are confident in balancing profit margins while achieving rapid revenue growth.

Risk Disclosure and Statement: Dolphin Research Disclaimer and General Disclosure