Duolingo Inc (DUOL) 2022 Q4 法說會逐字稿

(presentation)

（推介會）

We hope you enjoyed that fun TikTok. It's one of the many examples of organic learner-generated content that goes viral every day without any involvement from us. And we love to see the creativity and enthusiasm of creators like this, which continues to drive our brand awareness and word-of-mouth growth. And now let's get on with today's call.

我們希望您喜歡這個有趣的 TikTok。這是眾多由學習者生成的有機內容的例子之一，這些內容每天都在傳播，而無需我們的任何參與。我們喜歡看到像這樣的創作者的創造力和熱情，這將繼續推動我們的品牌知名度和口碑增長。現在讓我們繼續今天的電話會議。

Good afternoon, and welcome to Duolingo's Fourth Quarter and Full Year 2022 Earnings Webcast. (Operator Instructions) Today, after market close, we released our year-end shareholder letter with our Q4 and 2022 results and commentary, which you can find on our IR website at investors.duolingo.com.

下午好，歡迎收看 Duolingo 的第四季度和 2022 年全年收益網絡廣播。 （操作員說明）今天收盤後，我們發布了年終股東信函，其中包含我們的第四季度和 2022 年業績和評論，您可以在我們的 IR 網站 investors.duolingo.com 上找到。

On today's call, we have Luis von Ahn, our Co-Founder and CEO; and Matt Skaruppa, our CFO. They will begin with some brief remarks before opening the call to questions. (Operator Instructions) And please note, this event is being recorded.

在今天的電話會議上，我們有我們的聯合創始人兼首席執行官 Luis von Ahn；和我們的首席財務官 Matt Skaruppa。在開始提問之前，他們將先做一些簡短的評論。 （操作員說明）請注意，正在記錄此事件。

Just a reminder that we'll make forward-looking statements regarding future events and financial performance, which are subject to material risks and uncertainties. Some of these risks have been set forth in our risk factors of our filings with the SEC. These forward-looking statements are based on assumptions that we believe to be reasonable as of today, and we have no obligation to update these statements as a result of new information or future events.

提醒一下，我們將就未來事件和財務業績做出前瞻性陳述，這些陳述受到重大風險和不確定性的影響。我們在向美國證券交易委員會提交的文件中列出了其中一些風險。這些前瞻性陳述基於我們認為截至今日合理的假設，我們沒有義務根據新信息或未來事件更新這些陳述。

Additionally, we will present both GAAP and non-GAAP financial measures on today's call. These non-GAAP measures are not intended to be considered in isolation from, or substitute for or superior to, our GAAP results. And we encourage you to consider all measures when analyzing our performance.

此外，我們將在今天的電話會議上介紹 GAAP 和非 GAAP 財務指標。這些非 GAAP 措施並非旨在孤立於我們的 GAAP 結果，也無意替代或優於我們的 GAAP 結果。我們鼓勵您在分析我們的表現時考慮所有措施。

And now I will turn it over to Luis.

現在我將把它交給路易斯。

Thank you, Debbie, and welcome, everyone. I'm pleased to report that we finished 2022 with a record quarter, achieving our highest-ever number of DAUs and subscribers. We ended Q4 with 16.3 million DAUs, which is 62% year-over-year growth, and we ended 2022 with 4.2 million paying subscribers, which is 67% year-over-year growth. These metrics led 2022 to be our best year ever in terms of bookings and revenue, which grew 46% and 47%, respectively. It was a fantastic year.

謝謝你，黛比，歡迎大家。我很高興地報告，我們以創紀錄的季度結束了 2022 年，實現了有史以來最高的 DAU 和訂閱者數量。截至第四季度，我們的每日活躍用戶數為 1630 萬，同比增長 62%；截至 2022 年，我們的付費用戶為 420 萬，同比增長 67%。這些指標使 2022 年成為我們在預訂和收入方面有史以來最好的一年，分別增長了 46% 和 47%。那是美妙的一年。

Our DAU to MAU ratio, which is an indicator of user engagement, improved to 27% this past quarter, up from about 24% a year before. We have also seen the number of learners with long streaks grow immensely. To remind you, a streak is the number of consecutive days that a user learns with our app. This mechanic has proven to be an efficient tool for encouraging engagement and helping us grow DAUs. 63% of our DAUs now have a streak of 7 days or greater, which is up from 53% last year.

我們的 DAU 與 MAU 比率是用戶參與度的指標，在上個季度從一年前的約 24% 提高到 27%。我們還看到長期學習的學習者數量急劇增加。提醒您，連續天數是用戶使用我們的應用程序學習的連續天數。這種機制已被證明是鼓勵參與和幫助我們增加 DAU 的有效工具。我們 63% 的 DAU 現在連續 7 天或更長時間，高於去年的 53%。

What's even more impressive is that we now have over 3 million DAUs with a streak of over 1 year. That means that more than 3 million people use Duolingo every single day for the last year or longer. It's also impressive that it took us 8 years to reach 1 million DAUs with a year-long streak and only 21 months to get from 1 million to 3 million. This is one example of the compounding effect of our product improvements.

更令人印象深刻的是，我們現在擁有超過 300 萬的 DAU，並且連續超過 1 年。這意味著在過去一年或更長時間裡，每天有超過 300 萬人使用 Duolingo。同樣令人印象深刻的是，我們用了 8 年的時間達到 100 萬日活躍用戶，並且連續一年連續活躍，而從 100 萬到 300 萬隻用了 21 個月。這是我們產品改進的複合效應的一個例子。

Our user growth was strong in every region of the world. We also have users of nearly all ages, of all socioeconomic statuses and who have a variety of motivations to learn, including work, school and travel. The result is a diversified business that has been largely uncorrelated with macro and geographic trends.

我們的用戶增長在世界每個地區都很強勁。我們也有幾乎所有年齡段、所有社會經濟地位的用戶，他們有各種學習動機，包括工作、學校和旅行。結果是多元化的業務在很大程度上與宏觀和地理趨勢無關。

Our investments in monetization have also shown the power of compounding. Over the past 3 years, we've seen conversion increase across nearly all cohorts. Users who are new to the app have converted more readily to our paid subscription and users who have been on the app for longer periods, like multiple years, have also seen an increase in their conversion rates from free to pay. And now I'd like to discuss what you should expect from us this year.

我們在貨幣化方面的投資也顯示了複利的力量。在過去的 3 年裡，我們看到幾乎所有群組的轉化率都在增加。該應用程序的新用戶更容易轉換為我們的付費訂閱，而使用該應用程序時間較長（例如多年）的用戶也發現他們從免費到付費的轉換率有所提高。現在我想討論您今年對我們的期望。

In a lot of ways, we expect 2023 to look similar to 2022. We expect users and bookings to grow nicely. We still have a massive $60 billion addressable market that we have only barely penetrated, and we have strong organic growth on our side. And I want to remind you that currently, our subscribers make up about 8% of our monthly active users. So we have plenty of room to further monetize our current user base.

在很多方面，我們預計 2023 年會與 2022 年相似。我們預計用戶和預訂量會增長良好。我們仍然擁有一個價值 600 億美元的巨大潛在市場，我們只是勉強進入了這個市場，而且我們有強勁的有機增長。我想提醒您，目前，我們的訂戶約佔每月活躍用戶的 8%。因此，我們有足夠的空間進一步從我們當前的用戶群中獲利。

We also have an attractive opportunity to increase bookings from ala carte in-app purchases like we did in 2022. But I do want to point out again that you should not expect any bookings from our math or [legacy] apps this year. As to where we'll invest, the majority of our engineers, designers and product managers will continue to build on the success we've had in growth, efficacy and monetization.

我們也有一個誘人的機會，可以像 2022 年那樣增加單點應用內購買的預訂量。但我確實想再次指出，今年您不應該期望我們的數學或 [舊版] 應用程序有任何預訂。至於我們將在哪裡投資，我們的大多數工程師、設計師和產品經理將繼續以我們在增長、效率和貨幣化方面取得的成功為基礎。

That means that we'll continue to run experiments to increase DAUs and engagement, improve how we teach and optimize how we convert free users to pay. And in our shareholder letter, I discussed the 2 additional areas we're focused on in 2023: the first is generative AI, which will power a higher-tier Duolingo Max subscription as well as help us reduce content creation costs; and the second is improving how we teach and monetize English learners.

這意味著我們將繼續進行實驗以增加 DAU 和參與度，改進我們的教學方式並優化我們將免費用戶轉化為付費用戶的方式。在我們的股東信中，我討論了我們在 2023 年關注的另外兩個領域：第一個是生成 AI，它將支持更高級別的 Duolingo Max 訂閱並幫助我們降低內容創建成本；第二是改進我們教授英語學習者的方式並從中獲利。

I'd like to emphasize that we are excited about these initiatives, but they are still in the early stages. And I also want to touch on profitability. Coming off a strong 2022, we are well positioned to grow users and bookings rapidly while also delivering higher profitability. We plan to continue to compound the returns we've seen from our past R&D investments, so we don't plan to hire as many people as we did last year.

我想強調的是，我們對這些舉措感到興奮，但它們仍處於早期階段。我還想談談盈利能力。在強勁的 2022 年之後，我們有能力快速增加用戶和預訂量，同時提供更高的盈利能力。我們計劃繼續複合我們從過去的研發投資中看到的回報，因此我們不打算像去年那樣僱用那麼多的人。

Just to remind you, we've never gone nuts with hiring. And in fact, we've grown revenue at about twice the rate of head count in the past 4 years. This year, we plan to show operating leverage across all costs, R&D, sales and marketing and G&A. Given those strengths, we're guiding to a 10% to 12% adjusted EBITDA margin for the year, up from about 4% in 2022.

提醒一下，我們從來沒有瘋狂招聘過。事實上，在過去 4 年中，我們的收入增長速度大約是員工人數增長速度的兩倍。今年，我們計劃展示所有成本、研發、銷售和營銷以及 G&A 的運營槓桿。鑑於這些優勢，我們指導今年調整後的 EBITDA 利潤率為 10% 至 12%，高於 2022 年的約 4%。

And with that, I'll turn it over to Matt to talk more about our financial outlook.

有了這個，我會把它交給馬特更多地談論我們的財務前景。

Thanks, Luis. To quickly recap the highlights of our impressive Q4 and full year 2022 results. In the fourth quarter, we delivered 39% bookings growth year-over-year, which was about 46% on a constant currency basis. We saw a 42% revenue growth year-over-year, which was 47% on a constant currency basis. We had a net loss of $13.9 million compared to a net loss of $17.5 million in the year ago quarter. We posted our highest quarterly adjusted EBITDA of $5 million, which was a 5% adjusted EBITDA margin, and we had a free cash flow margin of about 11%.

謝謝，路易斯。快速回顧一下我們令人印象深刻的第四季度和 2022 年全年業績的亮點。第四季度，我們實現了 39% 的預訂量同比增長，按固定匯率計算約為 46%。我們看到收入同比增長 42%，按固定匯率計算為 47%。與去年同期的淨虧損 1750 萬美元相比，我們的淨虧損為 1390 萬美元。我們公佈了 500 萬美元的最高季度調整後 EBITDA，即 5% 的調整後 EBITDA 利潤率，我們的自由現金流利潤率約為 11%。

For the full year, we delivered 46% bookings growth year-over-year, which was about 52% on a constant currency basis. We saw 47% revenue growth, which was about 51% in constant currency. We had a net loss of $59.6 million compared to a net loss of $60.1 million last year. We saw adjusted EBITDA positive $15.5 million compared to an adjusted EBITDA loss of $1 million last year, and we had a free cash flow margin of about 12.5%.

全年，我們實現了 46% 的預訂量同比增長，按固定匯率計算約為 52%。我們看到收入增長了 47%，按固定匯率計算約為 51%。我們的淨虧損為 5960 萬美元，而去年淨虧損為 6010 萬美元。我們看到調整後的 EBITDA 為正 1550 萬美元，而去年調整後的 EBITDA 虧損為 100 萬美元，我們的自由現金流利潤率約為 12.5%。

In 2023, we expect to continue delivering strong bookings growth, and we will do so while making progress towards our long-term profitability goal of an adjusted EBITDA margin of 30% to 35%. For the full year 2023, we are guiding to $530 million to $542 million in total bookings; $486 million to $498 million in revenue; and an adjusted EBITDA margin of 10% to 12%, which translates to about a 32% incremental margin at the midpoint.

到 2023 年，我們預計將繼續實現強勁的預訂量增長，同時我們將在實現調整後 EBITDA 利潤率 30% 至 35% 的長期盈利目標方面取得進展。對於 2023 年全年，我們預計總預訂量為 5.3 億至 5.42 億美元； 4.86 億美元至 4.98 億美元的收入；調整後的 EBITDA 利潤率為 10% 至 12%，相當於中點約 32% 的增量利潤率。

And for Q1 2023, we are guiding to $127.5 million to $130.5 million in total bookings, $111 million to $114 million in revenue and an adjusted EBITDA margin of 9% to 10%. Our guidance assumes current prevailing foreign exchange rates. As a reminder, roughly half of our revenue comes from outside the U.S., so every 1% increase or decrease in the value of the dollar versus our basket of currencies has about a $2 million headwind or tailwind, respectively, on full year total bookings. At the prevailing exchange rates we've used in our guidance, foreign exchange is expected to be a 3-percentage-point headwind to Q1 2023 year-over-year bookings growth.

對於 2023 年第一季度，我們預計總預訂量為 1.275 億美元至 1.305 億美元，收入為 1.11 億美元至 1.14 億美元，調整後的 EBITDA 利潤率為 9% 至 10%。我們的指南假設當前的通行匯率。提醒一下，我們大約一半的收入來自美國以外的地區，因此美元兌我們的一攬子貨幣每升值或貶值 1%，對全年總預訂量的影響分別約為 200 萬美元。按照我們在指南中使用的現行匯率，預計外匯將對 2023 年第一季度的同比預訂量增長產生 3 個百分點的不利影響。

In terms of quarterly cadence of our bookings for the remainder of the year, we expect our year-over-year growth for total bookings in Q2 to be about the same as Q1. We expect Q3 year-over-year growth rate to be slightly higher, and we expect Q4 to be our biggest quarter in terms of dollar bookings. Given our outperformance in Q4 of this year, it will be a slightly lower growth rate.

就今年剩餘時間的預訂季度節奏而言，我們預計第二季度總預訂量的同比增長與第一季度大致相同。我們預計第三季度的同比增長率會略高，我們預計第四季度將成為我們美元預訂量最大的季度。鑑於我們今年第四季度的出色表現，增長率將略有下降。

After pricing in 2022, we lowered pricing -- prices in several countries to get our prices more in line with each country's GDP per capita. This initiative and our intentional mix shift from monthly to annual plans was to increase the overall LTV of our platform, and that drove down subscription revenue per paid sub. But we're done, for the most part, with lowering prices globally. And in fact, based on strong conversion trends, we believe we have the opportunity to raise prices in some places and are experimenting with pricing as we speak.

在 2022 年定價之後，我們降低了定價——幾個國家的價格，以使我們的價格更符合每個國家的人均 GDP。這一舉措和我們有意從月度計劃轉變為年度計劃的組合是為了增加我們平台的整體 LTV，這降低了每個付費訂閱者的訂閱收入。但在很大程度上，我們已經完成了全球降價。事實上，基於強勁的轉化趨勢，我們相信我們有機會在某些地方提高價格，並且正在試驗我們所說的定價。

We have not included any material amount of bookings or revenue from Duolingo Max in our guidance since we've only just started testing it, but we'll provide you with an update when we report our Q1 results.

由於我們才剛剛開始測試 Duolingo Max，因此我們沒有在我們的指南中包含任何實質性的預訂量或收入，但我們會在報告第一季度結果時為您提供更新。

I'd like to emphasize that we are focused on managing the business so that we achieve our full year adjusted EBITDA margin of 10% to 12%. We'll be making spending decisions throughout the year based on our revenue and gross margin trends to ensure that we are in this range for the full year.

我想強調的是，我們專注於管理業務，以便我們實現 10% 至 12% 的全年調整後 EBITDA 利潤率。我們將根據我們的收入和毛利率趨勢做出全年的支出決定，以確保我們全年都在這個範圍內。

Also to remind you, given the seasonality of our revenue, our Q2 and Q3 margins will be slightly lower than Q1, and Q4 will be higher. We will continue to run the business with discipline and prudently manage our expenses. Starting in Q1, we expect to see meaningful leverage in total non-GAAP OpEx compared to the fourth quarter of last year. And in 2023, we expect to achieve about 7 to 9 percentage points of operating leverage in total non-GAAP OpEx. For both periods, the improvement is mostly in sales and marketing and G&A and, to a lesser extent, R&D.

還要提醒您，考慮到我們收入的季節性，我們的 Q2 和 Q3 利潤率將略低於 Q1，而 Q4 會更高。我們將繼續以紀律經營業務，審慎管理開支。從第一季度開始，與去年第四季度相比，我們預計非 GAAP 運營支出總額將出現顯著的槓桿作用。到 2023 年，我們預計非 GAAP 運營支出總額的運營槓桿將達到約 7 到 9 個百分點。在這兩個時期，改進主要體現在銷售和營銷以及 G&A 和較小程度上的研發方面。

We ended the year with approximately 48 million fully diluted shares outstanding using the year-end closing price. In 2023, we expect to end the year with about 2% dilution from equity issued to employees, which is less than the roughly 3% dilution we had in '22.

使用年末收盤價計算，我們在年末擁有約 4800 萬股完全稀釋的已發行股票。到 2023 年，我們預計到年底，向員工發行的股權將稀釋約 2%，這低於我們在 22 年的大約 3% 的稀釋度。

And with that, I'll turn it back to Luis.

有了這個，我會把它轉迴路易斯。

Thank you, Matt. Before we get into Q&A, I'd like to thank our very talented team of Duos, who continue to find new ways to innovate and delight our learners.

謝謝你，馬特。在我們開始問答之前，我要感謝我們非常有才華的 Duos 團隊，他們不斷尋找新的方法來創新和取悅我們的學習者。

And now we would be happy to take your questions. I'm going to turn it back to Debbie to manage the queue, and here we go.

現在我們很樂意回答您的問題。我要把它轉回給 Debbie 來管理隊列，我們開始吧。

Okay. Thanks, Luis. (Operator Instructions) And the first question comes from Ralph Schackart of William Blair.

好的。謝謝，路易斯。 （操作員說明）第一個問題來自威廉布萊爾的拉爾夫沙克特。

First question, the shareholder letter, I think you called it a "Super" end to 2022. I think that may incorporate some of the beginning of the year campaign that you run to attract new users. Just curious, what drove the really strong performance in the year-end? And then how did your New Year's campaign compare to expectations? And I have a follow-up.

第一個問題，股東信，我認為你稱之為 2022 年的“超級”結束。我認為這可能包含你為吸引新用戶而開展的一些年初活動。只是好奇，是什麼推動了年底真正強勁的表現？那麼你的新年活動與預期相比如何？我有一個後續行動。

Thank you. That's a great question. So yes, we had a great Q4, great year-end. The main reason -- there's 2 main reasons. The first one is just our compounding product improvements. I mean we've just been getting better and better at making our products stickier and attracting more users. We have a very high-performing growth team that does that. That's one thing.

謝謝。這是一個很好的問題。所以是的，我們有一個很棒的第四季度，很棒的年底。主要原因——有兩個主要原因。第一個只是我們的複合產品改進。我的意思是，我們在提高產品粘性和吸引更多用戶方面做得越來越好。我們有一個非常高效的增長團隊來做到這一點。那是一回事。

And then another one is some of our marketing campaigns, we just did really well. A good example of that was our year-end review campaign where towards -- around December 14 or so or December 10 to December 14, all the users got a summary of what they did throughout the whole year. And they're expected -- well, they're encouraged to share it. We got a lot of shares, millions and millions of shares. And we hit top trending topics on Twitter, and stuff like that, by just having people share their Duolingo stats over several days. So stuff like that really helped us do the outperformance.

然後另一個是我們的一些營銷活動，我們做得非常好。一個很好的例子就是我們的年終回顧活動，大約在 12 月 14 日左右或 12 月 10 日至 12 月 14 日，所有用戶都對他們全年所做的事情進行了總結。他們被期望——好吧，他們被鼓勵分享它。我們得到了很多股份，數百萬股。我們通過讓人們在幾天內分享他們的 Duolingo 統計數據，在 Twitter 上發布熱門話題，諸如此類。所以像這樣的東西真的幫助我們取得了優異的成績。

In terms of New Year's campaign, we started -- just for reference, we start a New Year's campaign on December 28 usually every year, and it goes all the way to January. So for this quarter, it's about 3 days-ish of campaign, and they performed better than we expected. And the reason for that outperformance just has to do with the fact that every year, we take the learnings from the previous year of the campaign, and we just get a little better. So every year, we run a few A/B tests to try to make that campaign more effective, and we take those learnings for the next year. So that worked out pretty well for us.

在新年活動方面，我們開始——僅供參考，我們通常在每年的 12 月 28 日開始新年活動，一直持續到 1 月。所以對於這個季度，大約有 3 天的活動，他們的表現比我們預期的要好。之所以表現出色，是因為每年我們都會從前一年的競選活動中吸取教訓，而且我們會變得更好一點。因此，每年我們都會進行一些 A/B 測試，以嘗試使該活動更加有效，並且我們會在明年吸取這些經驗教訓。所以這對我們來說效果很好。

Great. Maybe just a quick follow-up. Just in terms of DAUs, I think you reaccelerated for 6 consecutive quarters, which is pretty tough to do. But just maybe some color what's driving the strong acceleration and growth. I think in the prepared remarks, you talked about sort of broad-based. But were there any sort of 1 or 2 regions that's stronger growth than others?

偉大的。也許只是一個快速跟進。就 DAU 而言，我認為你連續 6 個季度重新加速，這很難做到。但也許只是一些顏色推動了強勁的加速和增長。我認為在準備好的發言中，你談到了廣泛的基礎。但是有沒有一兩個地區的增長比其他地區更強勁？

Yes, it's a great question, too. So in terms of regions, it really is growing in every single region in the world. If you want to know stronger regions, the U.S. is particularly strong. Asia is particularly strong, and Western Europe is particularly strong in terms of growing DAUs.

是的，這也是一個很好的問題。因此，就地區而言，它確實在世界上每個地區都在增長。要說強區的話，美國尤其強。亞洲尤其強勁，而西歐在不斷增長的 DAU 方面尤其強勁。

And in terms of why it is, I mean it's again the same 2 reasons. It is continued product improvements that just compound over time. The product is just better and stickier, and then we just get better and better at figuring out what strikes a chord with marketing. So things like our social media campaigns like our TikTok campaigns and our Twitter campaigns and also some of the influencer work that we do.

至於為什麼會這樣，我的意思是這又是同樣的兩個原因。持續的產品改進只會隨著時間的推移而復雜化。產品變得更好、更有粘性，然後我們就越來越善於找出什麼能引起營銷的共鳴。所以像我們的社交媒體活動，比如我們的 TikTok 活動和我們的 Twitter 活動，還有我們所做的一些有影響力的工作。

And some of the stuff, we don't even do. For example, you just saw this video at the beginning of this call. These are entirely organic videos that other people make. It's almost every single day. I don't know if it is a fact that every single day in the world, but almost every single day, there is an organic video that goes viral that mentions Duolingo. So all of that just keeps getting -- keeps growing our users.

還有一些東西，我們甚至不做。例如，您剛剛在通話開始時看到了這個視頻。這些完全是其他人製作的有機視頻。幾乎每一天都是這樣。我不知道這是否是事實，世界上每一天，但幾乎每一天，都有一個提到 Duolingo 的有機視頻走紅。因此，所有這些都在不斷增長——不斷增加我們的用戶。

Thanks, Ralph. And the next question comes from Justin Patterson of KeyBanc.

謝謝，拉爾夫。下一個問題來自 KeyBanc 的 Justin Patterson。

Great. Luis, recently, there was a really interesting post on the Duolingo blog about just using data science to optimize DAUs. I think current user retention rate was that the metric you've been optimized in the past few years. Would love to hear more about just how you're using a lot of these top-down statistical models to really drive DAU growth, link the product team together. And then now that you've gone pretty far with the current metric, how you're thinking about tailoring that to individual cohorts. So more of a bottoms-up approach to take advantage of things like, say, Gen Z, having much shorter attention spans and making sure they're still having healthy engagement.

偉大的。 Luis，最近，Duolingo 博客上有一篇非常有趣的帖子，內容是關於使用數據科學來優化 DAU。我認為當前的用戶保留率是您在過去幾年中優化的指標。很想听到更多關於您如何使用大量這些自上而下的統計模型來真正推動 DAU 增長的信息，將產品團隊聯繫在一起。然後，既然您已經在當前指標方面走得很遠，那麼您是如何考慮針對各個群體量身定制的。因此，更多的是一種自下而上的方法來利用諸如 Z 世代之類的東西，他們的注意力持續時間要短得多，並確保他們仍然保持健康的參與度。

I'm glad you read that blog post. We're very proud of our growth model. Over the years, we've just gotten significantly more sophisticated at analyzing our user base and making changes that improve certain cohorts. In this post, we mentioned what has been -- I would say the single most important metric for this company over the last several years, which is called -- something we call CURR, or current user retention rate. And what that is, it is the rate of people that if they were here today and they were here over the last one other time, at least over the last 7 days, what is the probability they're going to come back tomorrow, that's current user retention rate. That number is around 80% right now, and we've increased it. Over the last 2, 3 years, we've increased that number from about 65% to about 80%, and every 1% increase of CURR increased our DAUs by a lot. And so that model has been -- really helped us, and we keep getting more and more sophisticated.

很高興您閱讀了該博文。我們對我們的增長模式感到非常自豪。多年來，我們在分析我們的用戶群和做出改善某些群體的改變方面變得更加成熟。在這篇文章中，我們提到了——我想說這家公司在過去幾年中最重要的單一指標，它被稱為——我們稱之為 CURR，即當前用戶保留率。那是什麼，如果他們今天在這裡，並且他們上一次在這裡，至少在過去 7 天裡，他們明天回來的概率是多少，那就是當前用戶留存率。這個數字現在大約是 80%，而且我們已經增加了它。在過去的 2、3 年裡，我們將該數字從大約 65% 增加到大約 80%，並且 CURR 每增加 1%，我們的 DAU 就會增加很多。所以這個模型已經——真的幫助了我們，我們變得越來越複雜。

You are right, we're starting to -- this was a model -- at first, when you're just -- before this growth model, we would just treat all our users in one cohort. Then we started getting more sophisticated, and we treat kind of new users, current users, et cetera. Now we're getting even more sophisticated than that. I don't think we have a specific thing for Gen Z that we do. But we are -- relatedly, we are making our sessions shorter and shorter because, like you said, they have a shorter and shorter attention span.

你是對的，我們開始——這是一個模型——起初，當你只是——在這個增長模型之前，我們只會把所有用戶都放在一個隊列中。然後我們開始變得更加複雜，我們對待新用戶、現有用戶等等。現在我們變得比這更複雜。我認為我們沒有為 Z 世代做任何特定的事情。但是我們 - 相關地，我們正在使我們的會議越來越短，因為就像你說的那樣，他們的注意力持續時間越來越短。

Great. The next question comes from Andrew Boone of JMP.

偉大的。下一個問題來自 JMP 的 Andrew Boone。

Great. Can we start off with just any learnings from in-app purchases? I saw that it doubled this last quarter. Can you talk about what successes are there? And then how we should think about that for 2023? And then secondly, on profitability, Matt talked about real contribution margin kind of (inaudible) in the back half of this year. Is there any reason to think that, that steps down as we think about kind of a 3-year plan or something beyond this upcoming year?

偉大的。我們可以從應用內購買中學到的東西開始嗎？我看到它在上個季度翻了一番。你能談談有哪些成功之處嗎？然後我們應該如何考慮 2023 年的情況？其次，關於盈利能力，馬特談到了今年下半年的實際邊際貢獻（聽不清）。是否有任何理由認為，當我們考慮某種 3 年計劃或明年以後的計劃時，它會下降？

I can take the IAP question, Matt, and then you can take the next one. So for in-app purchases, yes, we have a very high-performing team behind this that is just making in-app purchases more and more prominent, more and more successful in the app. As you saw, they've been growing very rapidly. We expect them to continue growing rapidly this year.

我可以回答 IAP 問題，Matt，然後你可以回答下一個問題。所以對於應用內購買，是的，我們背後有一個非常高效的團隊，他們正在讓應用內購買越來越突出，在應用中越來越成功。如您所見，它們一直在快速增長。我們預計他們今年將繼續快速增長。

And in terms of learnings, the main way we do in-app purchases, by the way, is by selling gems. So there's this in-app currency called gems that you can either earn by doing stuff on the app or you can also buy them. And then you can use gems to buy other stuff, kind of power-ups in the app. One of the main power-ups that we use in the app is the timer boosts that help for certain timed lessons. So we have these -- most lessons on Duolingo are not timed, but there are some lessons that you can do -- some practice lessons that are timed. And the idea is that if you're not very fast, you run out of time or you can buy these timer boosts with gems. And those timer boosts get a lot of people to spend on in-app purchases. This is the type of stuff you're going to get to see us do.

在學習方面，順便說一句，我們進行應用內購買的主要方式是出售寶石。所以有一種名為寶石的應用內貨幣，您可以通過在應用上做事來賺取，也可以購買它們。然後你可以使用寶石購買其他東西，比如應用程序中的能量提升。我們在應用程序中使用的主要增強功能之一是計時器增強功能，它有助於某些定時課程。所以我們有這些 - Duolingo 上的大多數課程都沒有定時，但有一些課程你可以做 - 一些練習課程是定時的。這個想法是，如果你不是很快，你就會用完時間，或者你可以用寶石購買這些計時器提升。這些計時器的提升讓很多人花在應用內購買上。這是您將看到我們做的事情。

So in general, for us, for in-app purchases, we're not going to be selling content within our purchases, but mainly we're going to be selling gamification things that people really like. And so far, that strategy has been working pretty well.

所以總的來說，對於我們來說，對於應用內購買，我們不會在購買中銷售內容，但主要是我們將銷售人們真正喜歡的遊戲化東西。到目前為止，該策略一直運作良好。

Great. And then on profitability. So Andrew, the profitability increased in 2022 by about 450 basis points from 2021 on an adjusted EBITDA basis, which was good progress towards our long-term profitability goal, of that 30% to 35%. This year, we're guiding to 600 to 800-basis point improvement over that. And so as we look out kind of into the longer term, our goal hasn't changed. We want to be at a 30% to 35% adjusted EBITDA margin. And between last year and what we're guiding to this year, we think we're showing the right progress towards that goal.

偉大的。然後是盈利能力。因此，安德魯，在調整後的 EBITDA 基礎上，2022 年的盈利能力比 2021 年增加了約 450 個基點，這是朝著我們的長期盈利目標（30% 到 35%）邁進的良好進展。今年，我們正在指導 600 到 800 個基點的改進。因此，當我們從長遠來看時，我們的目標沒有改變。我們希望調整後的 EBITDA 利潤率達到 30% 至 35%。在去年和我們今年的指導之間，我們認為我們正在朝著這個目標展示正確的進展。

Thanks. And then the next question comes from Ryan MacDonald of Needham.

謝謝。然後下一個問題來自 Needham 的 Ryan MacDonald。

Congrats on a strong close to the year. Luis, let me start with you on the Duolingo English test. There was recently a ban in China from the government on online degrees, kind of a Chinese citizen essentially taking a fully online degree from an international university. They're no longer recognizing that degree. And it seems like this is going to start creating sort of a shift in an inflection point, and more Chinese learners looking to go in person overseas. I'm just curious if you've seen any impacts of this thus far on sort of applications or sign-ups for the Duolingo English test and how you think -- or what your expectations for that are in 2023 here.

祝賀今年的收官之戰。路易斯，讓我從你的多鄰國英語測試開始。最近中國政府禁止在線學位，中國公民基本上是從國際大學獲得完全在線學位。他們不再承認那個學位。似乎這將開始在拐點上產生某種轉變，更多的中國學習者希望親自出國。我只是想知道到目前為止，您是否看到這對 Duolingo 英語測試的申請或註冊有任何影響，以及您的想法 - 或者您對 2023 年的期望是什麼。

So well, thank you for the question. We have not seen any impact based on this, and I think there's many reasons. I mean, for one, yes, China is a relatively large chunk of the Duolingo English test takers, but we really have test takers all over the world.

那麼好，謝謝你的提問。我們還沒有看到任何基於此的影響，我認為有很多原因。我的意思是，一方面，是的，中國是 Duolingo 英語應試者中相對較大的一部分，但我們確實在世界各地都有應試者。

The other thing is the reasons for taking the tests are very varied. Some of them is people who are applying for international undergraduate admissions. There's also graduate admissions. There's transnational education. There's all kinds of different pockets. And maybe one pocket has some sort of policy implications to it. But in general, the pockets are pretty healthy. That's just kind of one thing.

另一件事是參加考試的原因多種多樣。其中一些是正在申請國際本科招生的人。還有研究生招生。有跨國教育。有各種不同的口袋。也許一個口袋對它有某種政策影響。但總的來說，口袋很健康。那隻是一回事。

The other thing is we are still -- we are not the leader in this market. We're coming into this market with an online test, and it is a market that usually has had all these off-line tests. So given that we are still in the kind of growth phase of this, we don't -- it doesn't affect us too much when there's something that changes the full volume of it because we're coming in with whatever percentage of tests we have, which is significantly under 25%. We're still growing quite a bit and pretty fast, so this is not something that we're seeing yet.

另一件事是我們仍然 - 我們不是這個市場的領導者。我們通過在線測試進入這個市場，這個市場通常已經進行了所有這些離線測試。因此，鑑於我們仍處於這種增長階段，我們不會 - 當有一些東西改變它的全部體積時，它不會對我們產生太大影響，因為我們正在進行任何百分比的測試我們有，遠遠低於 25%。我們仍在快速增長，所以這還不是我們所看到的。

That's helpful. Appreciate it. And then maybe as a follow-up. You talked a lot about generative AI and the use of that, and I'm very curious around the content creation. You mentioned an increased focus on sort of more advanced English content for non-English speakers or basic English speakers. How quickly do you think that the generative AI can compress your content creation time line? And how do you manage for quality in that? And then for Matt, are you assuming any -- within the leverage that you're showing this year on the profit margin side, are you assuming any leverage from that content creation from generative AI?

這很有幫助。欣賞它。然後可能作為後續行動。你談了很多關於生成 AI 及其使用的內容，我對內容創作非常好奇。你提到越來越關注為非英語使用者或基本英語使用者提供的更高級的英語內容。您認為生成式 AI 可以多快壓縮您的內容創建時間線？您如何管理其中的質量？然後對於馬特，你是否假設任何 - 在你今年在利潤率方面展示的槓桿作用中，你是否假設來自生成人工智能的內容創建的任何槓桿作用？

Yes. Thank you for the question. So we're very excited about generative AI. And just to put things into context, at Duolingo, we've always leaned into AI from the beginning. I mean a lot of our products are built based on AI, et cetera. Over the last couple of years, generative AI has become significantly better, and we're definitely going to lean into this as well.

是的。感謝你的提問。所以我們對生成人工智能感到非常興奮。簡單來說，在 Duolingo，我們從一開始就傾向於 AI。我的意思是我們的很多產品都是基於 AI 等構建的。在過去的幾年裡，生成式 AI 已經變得更好了，我們肯定也會向它傾斜。

You mentioned content creation. That's one of the places where we think that it can help us a lot and speed things up quite a bit. It's a bit early for me to give you a precise estimate of how much you can speed it up, but we are certain -- I mean it's already happening. We're already speeding things and also lowering costs. One of the places where we're using that is to be able to create a much more English learning content. So you'll see us create a lot more of that.

你提到了內容創作。這是我們認為它可以幫助我們很多並大大加快速度的地方之一。現在就準確估計可以加快多少速度對我來說還為時過早，但我們可以肯定——我的意思是它已經在發生了。我們已經在加快速度並降低成本。我們使用它的地方之一是能夠創建更多的英語學習內容。所以你會看到我們創造了更多。

In terms of quality assurance, we're still going to have humans go over it. But the time frame, it used to be the case that all of this was almost entirely generated by humans. By now, it just has to be reviewed by humans, and it's a lot quicker to review than to generate. So in terms of time frame, we're going to get a massive increase in that, and I assume it's the same for cost. So yes, this is something we're very excited about.

在質量保證方面，我們仍然會讓人檢查它。但是時間框架，曾經是所有這些幾乎完全由人類產生的情況。到目前為止，它只需要人工審核，而且審核比生成要快得多。因此，就時間框架而言，我們將大幅增加，我認為成本也是一樣的。所以是的，這是我們非常興奮的事情。

And the other thing I should mention about generative AI, so content creation is one of them, and we're going to definitely use that. But it also is going to help us just create better experiences to our users, kind of more online conversational practice that we're going to be able to do. And that's the type of stuff that we're putting into our new higher tier subscription, Duolingo Max. So this is something we're very excited about. As a company, we've partnered with open AI, and we're going to be working very hard on this.

關於生成人工智能，我應該提到的另一件事是，內容創建就是其中之一，我們肯定會使用它。但它也將幫助我們為用戶創造更好的體驗，這是我們將能夠做的更多在線對話練習。這就是我們在新的更高級別訂閱 Duolingo Max 中加入的內容。所以這是我們非常興奮的事情。作為一家公司，我們已經與開放式 AI 合作，我們將在這方面非常努力。

Yes. And Ryan, Luis answered the question for the guide. I mean we only included in the guide things we have direct line of sight into. It's not -- generative AI is so new. We don't have a ton of line of sight into direct savings yet. But as Luis said, over time, that's probably the way to bet.

是的。瑞安，路易斯回答了指南的問題。我的意思是我們只在指南中包含了我們有直接視線的東西。不是——生成式 AI 太新了。我們對直接儲蓄還沒有太多的看法。但正如路易斯所說，隨著時間的推移，這可能就是下注的方式。

Thanks, Ryan. And the next question comes from Aaron Kessler of Raymond James.

謝謝，瑞安。下一個問題來自 Raymond James 的 Aaron Kessler。

A couple of questions. Maybe for Matt. On the paid conversion rate, it was very strong in Q4. Can you just talk about that? Was that family plan a part of that? And then how should we think about paid conversion in 2023? And then maybe any update on a given test, maybe kind of human tutoring as well. Is there any kind of thoughts on that direction as well, especially as maybe it relates to Duolingo Max, how you're talking about -- thinking about integrating that as well?

幾個問題。也許是為了馬特。在付費轉化率方面，第四季度非常強勁。你能談談嗎？那個家庭計劃是其中的一部分嗎？那麼我們應該如何考慮 2023 年的付費轉化？然後可能是給定測試的任何更新，也可能是某種人工輔導。在這個方向上是否也有任何想法，尤其是它可能與 Duolingo Max 相關，你是如何談論的——考慮整合它嗎？

Sure. Yes. I'll let Luis answer that last part. For the payer penetration, we had mentioned, I think throughout most of last year, so we expected a 1% to 2% increase each quarter year-over-year in that metric. And that's coming from all of the things that Luis has described before, that all the product improvements that compound over time to make our free-to-pay conversion go up, in general. There wasn't any one particular item like family plan or this or that. It was all of them compounding over time.

當然。是的。我會讓路易斯回答最後一部分。對於付款人滲透率，我認為在去年的大部分時間裡我們都提到過，因此我們預計該指標每個季度同比增長 1% 至 2%。這來自 Luis 之前描述的所有事情，即隨著時間的推移，所有產品改進都會使我們的免費付費轉化率上升，一般來說。沒有任何一個特定的項目，比如家庭計劃或這個或那個。隨著時間的推移，它們都在復合。

I think going forward, as we mentioned on, I think, the Q2 or Q3 call last year, given the growth in MAU, so that denominator of that number is just growing really, really nicely, and the retention trends, we would expect it to actually instead of adding 1 or 2 points every quarter year-over-year, be closer to 1% this year, just because, again, that denominator is growing really nicely. So it will ultimately turn into subscribers over time. But in the short term, it will be about 1 point. And then, Luis, you want to talk about Max and tutors?

我認為展望未來，正如我們在去年第二季度或第三季度的電話會議上提到的那樣，考慮到 MAU 的增長，所以這個數字的分母真的非常非常好地增長，而保留趨勢，我們預計它實際上，與其每季度同比增加 1 或 2 個百分點，不如今年接近 1%，只是因為分母再次增長得非常好。因此，隨著時間的推移，它最終會變成訂閱者。但短期內，會在1點左右。然後，Luis，你想談談 Max 和導師嗎？

Yes, on human tutoring. I mean the first question was about human tutoring -- or the second question was about human tutoring. This is over the years, we've waffled with human tutoring. At first, we launched Duolingo with no humans involved. Then we tried it. We tried putting it in. We tried -- and over the last year, we've made the decision that I think is the right decision that we are not going to get into human tutoring at all. We've tried it. We are out. There's a number of reasons for that. There's nothing -- it's not like we have anything against human tutors. It's just on our end, we are a technology company, and that's what we excel at.

是的，關於人工輔導。我的意思是第一個問題是關於人工輔導的——或者第二個問題是關於人工輔導的。多年來，我們一直在為人工輔導而胡思亂想。起初，我們在沒有人參與的情況下推出了 Duolingo。然後我們試了一下。我們試著把它放進去。我們試過了——在過去的一年裡，我們做出了一個我認為是正確的決定，即我們根本不打算進行人工輔導。我們試過了。我們出去了。這有很多原因。沒有什麼——我們並不反對人類導師。就在我們這邊，我們是一家科技公司，這就是我們擅長的。

And in addition to that, if you look at it, human tutors are great, but they're not getting any better. They're just there. Whereas technology, in particular AI, and in particular a lot of the stuff coming from generative AI, is just getting better and better every year. So we're going to bet on that. Much in the same way, like 10 years ago, we were told we could not teach a language with synthesized computerized voices. We bet on that, even though 10 years ago, synthesized computerized voices still sounded kind of robotic. Today, they sound excellent. We feel the same with AI that in 5 years, I don't know, 5, 10 years, we will be very happy that we'll have that on technology rather than humans.

除此之外，如果你看一下，人類導師很棒，但他們並沒有變得更好。他們就在那裡。而技術，尤其是人工智能，尤其是來自生成人工智能的許多東西，每年都在變得越來越好。所以我們要打賭。以同樣的方式，就像 10 年前，我們被告知我們不能用合成的計算機語音來教授語言。我們敢打賭，即使在 10 年前，合成的計算機語音聽起來仍然有點像機器人。今天，它們聽起來很棒。我們對 AI 有同樣的感覺，我不知道，5 年，10 年，我們會很高興我們將在技術而不是人類上擁有它。

And next question comes from Nat Schindler from Bank of America.

下一個問題來自美國銀行的 Nat Schindler。

Yes. I actually want to just go a little deeper on some of the questions you had before, really about how the users have changed. You saw a dramatic acceleration in both users and paid subscribers over this year versus the prior year. Has there been a real change in the demographics of those? Is there anything substantially different about users this year than there were last year?

是的。實際上，我想更深入地談談您之前提出的一些問題，實際上是關於用戶如何變化的問題。與上一年相比，今年的用戶和付費訂閱者都出現了顯著增長。這些人群的人口統計數據是否發生了真正的變化？今年的用戶與去年相比有什麼實質性的不同嗎？

It's a great question. If there has been, it's small. There's -- generally, our user base is very wide. We have users from every single country in the world. Every single country in the world is represented. We have users from every single socioeconomic status from all ages as long as they know how to read. So it's kind of from 6 years to about 100. Just -- we just really have users from all over the place, and it's a very wide user base. So it hasn't changed too much.

這是一個很好的問題。如果有的話，它很小。通常，我們的用戶群非常廣泛。我們的用戶來自世界上每一個國家。世界上每個國家都有代表。我們擁有來自各個社會經濟地位的各個年齡段的用戶，只要他們會閱讀即可。所以它是從 6 年到大約 100 年。只是 - 我們真的有來自各地的用戶，而且它是一個非常廣泛的用戶群。所以它並沒有改變太多。

The one thing we've noticed is it has gotten slightly, ever so slightly younger, and that probably has to do with our -- a lot of our prominence on social media like TikTok, but that's it. So we haven't really noticed too much of a change.

我們注意到的一件事是它變得稍微年輕了一點，這可能與我們——我們在 TikTok 等社交媒體上的知名度有關，但僅此而已。所以我們並沒有真正注意到太多的變化。

And as those changes towards going slightly younger, has that happened in both free users and paid?

隨著這些變化變得更年輕化，免費用戶和付費用戶是否都發生了變化？

The honest answer is I don't know. I know that it's -- free users is getting slightly younger. I don't know for paid. Maybe...

誠實的回答是我不知道。我知道這是——免費用戶正在變得更年輕。我不知道付費。或許...

Yes. I'd have to go back and double check, Nat, on the paid. I think what we've seen on paid is that the conversion trends haven't really changed. Some have accelerated. For example, just overall conversion of kind of cohorts across the board has increased over the course of last year. So folks who are new to the platform in the first couple of months, they converted more rapidly than in 2021.

是的。我得回去仔細檢查一下，Nat，支付的費用。我認為我們在付費方面看到的是轉換趨勢並沒有真正改變。有些已經加速了。例如，在去年的過程中，整體同類群組的整體轉化率有所增加。因此，在頭幾個月剛接觸該平台的人們，他們的轉化速度比 2021 年更快。

Folks that have been on the platform for a long period of time, they converted a little bit faster than they had in the past. We've always seen that people who use the app more convert more. That stayed the same or got a little bit better. So yes, it was kind of a broad base. Just like our users grew around the world in a pretty broad-based way, the same trend on the free to paid conversion.

已經在該平台上呆了很長時間的人，他們的轉化速度比過去快了一點。我們一直看到，使用應用程序越多的人轉化率越高。那保持不變或變得更好一點。所以是的，這是一個廣泛的基礎。就像我們的用戶在全球範圍內以廣泛的方式增長一樣，免費到付費的轉換趨勢也是如此。

Thank you very much. And once again, very impressive.

非常感謝。再一次，非常令人印象深刻。

Thanks, Nat. And next question comes from Mark Mahaney of Evercore.

謝謝，奈特。下一個問題來自 Evercore 的 Mark Mahaney。

Two questions. You had this line in the release about how the number of returning users is roughly the same as the number of brand-new users. I would just have assumed that, that has always been the case or it's been the case for a while. But is that something new that's happened? Because I get your point about how people will use it, and then they'll either come back to it when they realize importance in learning a new language or there'll be another opportunity. Maybe they're traveling to a different country. So is that a new trend or not? What's the so what out of that? And just real quickly, Duo Math, though, that math product, just where is that? Are you still tweaking it? Is it far from being where you want it to be? Just talk about kind of adoption, how much interest you've really seen in that.

兩個問題。你在發布中有這樣一句話，關於返回用戶的數量如何與全新用戶的數量大致相同。我只是假設，情況一直如此，或者一段時間以來一直如此。但這是發生了什麼新事嗎？因為我明白你關於人們將如何使用它的觀點，然後當他們意識到學習一門新語言的重要性時，他們會重新使用它，或者會有另一個機會。也許他們要去另一個國家旅行。那麼這是否是一種新趨勢？那是什麼？很快，Duo Math，那個數學產品，它在哪裡？你還在調整它嗎？它遠離你想要的地方嗎？只是談談採用的種類，你真正看到了多少興趣。

For your first question, this is not a new thing. It's been a while since returning users are the same as new users. I'm going to assume at the very early of Duolingo, that was not the case. But it's been a while that has been like that. And you're right. It's because it is rare for us to see a user that uses Duolingo goes away and never comes back. That is rare. Usually, people may even go away for 2 years and then come back. It's just because there's such a little friction. So we see that. And the reason we put that there is mainly to emphasize that it is not right to think of our users as coming, having churn and been gone forever. That is rare. It's more that they come. They stay for a few months or something, then they may take a break for a couple of months and they come back or something. So something like that.

對於你的第一個問題，這不是什麼新鮮事。回頭用戶與新用戶一樣已經有一段時間了。我假設在 Duolingo 的早期，情況並非如此。但這種情況已經有一段時間了。你是對的。這是因為我們很少看到使用 Duolingo 的用戶離開後再也沒有回來。那是很少見的。通常，人們甚至可能離開 2 年然後再回來。只是因為有這麼一點點摩擦。所以我們看到了。我們之所以這樣說，主要是為了強調認為我們的用戶來了，流失了，然後永遠離開了，這是不對的。那是很少見的。更何況他們來了。他們會待幾個月之類的，然後他們可能會休息幾個月，然後他們回來之類的。所以類似的東西。

In terms of math, we're very happy with the progress. It's growing rapidly. We're happy with that. Just to remind you of the plan, we are -- right now, the goal is to grow our math product organically for it to grow by itself. Once we get to a pretty decent number of users, call it 1 million daily active users, this is kind of -- it doesn't have to be exactly a million. Once we get there, we're going to say, okay, great. We have good product market. It is time to start monetizing. We're not there yet, but it is growing pretty nicely, and so we're happy with the results.

在數學方面，我們對取得的進展感到非常滿意。它正在迅速增長。我們對此很滿意。只是為了提醒您這個計劃，我們現在的目標是有機地發展我們的數學產品，讓它自己成長。一旦我們獲得相當數量的用戶，稱之為 100 萬日活躍用戶，這有點——它不一定是一百萬。一旦我們到達那裡，我們會說，好的，太好了。我們有很好的產品市場。是時候開始貨幣化了。我們還沒有到那一步，但它的增長非常好，所以我們對結果感到滿意。

In terms of tweaking guests, we're working very hard on tweaking the app. We're adding a lot of the features that we know work well on the main Duolingo app to the math product. So improvements to the Street. We're thinking of adding things like leaderboards, better notifications. So a lot of the stuff that we've learned over the last 10 years on Duolingo and that we have the code for, we're starting to add to the math app to get it -- to have higher retention. We're also adding more content to it.

在調整訪客方面，我們正在非常努力地調整應用程序。我們正在將許多我們知道在 Duolingo 主應用程序上運行良好的功能添加到數學產品中。所以對街道進行了改進。我們正在考慮添加排行榜、更好的通知等內容。因此，我們在過去 10 年中在 Duolingo 上學到的很多東西以及我們擁有代碼的東西，我們開始添加到數學應用程序中以獲取它——以獲得更高的保留率。我們還向其中添加了更多內容。

One of the things that happens with the math app right now is people actually run out of content because if you are very -- it turns out we have some very heavy users that use the whole thing, and then they kind of run out of content. So we're adding more content to it. So you'll see us over the next year just make the product better and better. And because of that, it will continue growing.

數學應用程序現在發生的一件事是人們實際上已經用完了內容，因為如果你非常——事實證明我們有一些非常重度的用戶使用整個東西，然後他們就會用完內容.所以我們正在向它添加更多內容。所以你會在明年看到我們讓產品變得越來越好。正因為如此，它將繼續增長。

Next question comes from Eric Sheridan of Goldman Sachs.

下一個問題來自高盛的 Eric Sheridan。

Great. Maybe first question would be following up on something that was implied in Mark's first question. Travel has obviously been a big tailwind, and the shift to services over goods has been a big tailwind in the macroeconomic environment. Can you talk a little bit about what you've seen with travel potentially as a catalyst for usage, both top of the funnel and the bottom of the funnel as that's continued to build momentum and how maybe as a result of cross-border travel continuing to return, that could be a nice tailwind for the business going forward over 2023 and beyond.

偉大的。也許第一個問題是跟進 Mark 第一個問題中暗示的內容。旅行顯然是一個巨大的順風，在宏觀經濟環境中，向服務而非商品的轉變是一個巨大的順風。你能談談你所看到的旅行可能成為使用催化劑的情況嗎，無論是漏斗頂部還是漏斗底部，因為它繼續建立勢頭，以及跨境旅行持續的結果如何歸根結底，這對於 2023 年及以後的業務發展來說可能是一個很好的推動因素。

And then the second dynamic would be, you talked a lot about virality in TikTok. Can you talk a little bit about where you are excited to experiment and think about marketing or building virality on platforms on the content side or the distribution platform side, maybe even away from TikTok and how we should be thinking about that building in the years ahead?

然後第二個動態是，你談到了 TikTok 中的病毒式傳播。你能談談你對在內容端或分發平台端進行營銷或建立病毒式傳播感到興奮的地方嗎，甚至可能遠離 TikTok，以及我們在未來幾年應該如何考慮這種建設？

Great question, Eric. Let's see. In terms of travel, we don't see in our metrics that more and more people are using Duolingo for travel. We did see a very temporary dip in the height of COVID a couple of years ago. That is like a fewer -- slightly fewer. Since then, it's been about the same.

好問題，埃里克。讓我們來看看。在旅行方面，我們沒有在我們的指標中看到越來越多的人在使用 Duolingo 旅行。幾年前，我們確實看到 COVID 的高度暫時下降。這就像少了一點——稍微少了一點。從那時起，情況就差不多了。

And we just -- travel is interesting. For a lot of people like us that are, to be honest, from the worldwide perspective, pretty wealthy. I think travel is like a big motivation for using Duolingo. But for a lot of our users, they're using it to learn English or as a hobby or something. So we don't see these things that's like, "Oh, my god, travel has opened up, and therefore, Duolingo is exploding." That just -- we just haven't seen much of that. We do ask when people sign up, why they're using Duolingo. And some fraction tell us they're using it for travel. I don't know the number off the top of my head, I'm going to assume it's something like 15% to tell us that they're using it for travel. But it's just not something that we look too much to.

我們只是——旅行很有趣。對於很多像我們這樣的人來說，老實說，從全世界的角度來看，他們都非常富有。我認為旅行是使用 Duolingo 的一大動力。但對於我們的很多用戶來說，他們用它來學習英語或作為一種愛好或其他東西。所以我們不會看到這樣的事情，“哦，天哪，旅行已經開放，因此，Duolingo 正在爆炸式增長。”那隻是 - 我們還沒有看到太多。我們確實會在人們註冊時詢問他們為什麼使用 Duolingo。一些人告訴我們他們正在用它來旅行。我不知道我腦海中浮現出的數字，我假設大約有 15% 告訴我們他們將其用於旅行。但這不是我們看得太多的東西。

In terms of your second question about virality. So, so far, we have experimented with a few social media platforms to get our word out, by the way. And this is mainly organic, as in we don't pay for this stuff. We just make really good content. TikTok has been really good for us. Not just TikTok in the U.S., we've actually localized our TikTok to Portuguese and Spanish and other languages. So one of the things that we're excited about is TikTok accounts in the other languages. They're actually growing really fast. And it's actually amazing how -- by the way, this is just amazing thing about Brazil. Anything you do with social in Brazil does well. And so as soon as we put TikTok in Portuguese, it just kind of went nuts. And if you go see those videos, they're amazing. We just had Duo dressed up as Wednesday from the Netflix show, Wednesday. And we got, I don't know how many, 10 million views or something. So we're excited about platforms like TikTok and actually also Twitter. We're doing pretty well in Twitter.

關於你關於病毒式傳播的第二個問題。所以，到目前為止，我們已經嘗試了一些社交媒體平台來宣傳我們的想法。這主要是有機的，因為我們不為這些東西付費。我們只是製作非常好的內容。 TikTok 對我們來說真的很好。不僅僅是美國的 TikTok，我們實際上已經將 TikTok 本地化為葡萄牙語、西班牙語和其他語言。因此，我們感到興奮的一件事是其他語言的 TikTok 帳戶。他們實際上增長得非常快。它實際上是驚人的——順便說一下，這對巴西來說真是太神奇了。在巴西，你在社交方面做的任何事情都做得很好。因此，一旦我們將 TikTok 置於葡萄牙語中，它就有點瘋狂了。如果你去看那些視頻，它們會很棒。我們剛剛讓 Duo 打扮成星期三 Netflix 節目中的星期三。我們得到了，我不知道有多少，1000 萬次瀏覽量之類的。所以我們對像 TikTok 這樣的平台感到興奮，實際上還有 Twitter。我們在 Twitter 上做得很好。

Now when you ask about virality in general, another thing that we're doing is we're just adding a lot more things inside the product. This is not social media or anything, inside the product to get more people to share things. And we now have a team behind it. It's a really good team, too, and they are able to get a lot more people to share things.

現在，當您一般詢問病毒式傳播時，我們正在做的另一件事是我們只是在產品中添加了更多的東西。這不是社交媒體或任何東西，在產品內部讓更多人分享東西。我們現在有一個團隊支持它。這也是一個非常好的團隊，他們能夠讓更多的人分享東西。

One of the main things that people share, there's now a way to share sentences much more easily. One of the main things people share are weird sentences that they see on Duolingo. So some of the courses, for one reason or another, have a very weird sentences. I don't know if I can think of one of them. Like the German course has something like -- the sentence has something like, "right next to the bed, lay the body of the husband," or something like really weird stuff that people share. And that just gets a lot of people. So we're working on getting people to share like that, and it's working out pretty well.

人們分享的主要內容之一是，現在有一種方法可以更輕鬆地分享句子。人們分享的主要內容之一是他們在 Duolingo 上看到的奇怪句子。所以有些課程，出於某種原因，有一個非常奇怪的句子。我不知道我是否能想到其中之一。就像德語課程有類似的東西——句子有類似的東西，“就在床邊，躺著丈夫的屍體”，或者類似人們分享的非常奇怪的東西。這只會吸引很多人。所以我們正在努力讓人們像這樣分享，而且效果很好。

I want to say Matt probably wrote that question about the -- that sentence about the husband. But, okay.

我想說 Matt 可能寫了關於 - 關於丈夫的那句話的問題。但是，好吧。

Next question comes from Mario Lu of Barclays.

下一個問題來自巴克萊銀行的 Mario Lu。

Great. So the first one is on app store fees. Just curious to your thought on the recent trend of apps bypassing the app store. For example, games like Fortnite and apps like Spotify, Netflix have done this for a while. But more recently, more traditional mobile games are implementing like a direct-to-consumer payment method. Just curious your thoughts on any initiatives that you guys have on your end.

偉大的。所以第一個是應用程序商店費用。只是想知道您對最近繞過應用程序商店的應用程序趨勢的看法。例如，Fortnite 等遊戲和 Spotify、Netflix 等應用程序已經這樣做了一段時間。但最近，越來越多的傳統手機遊戲開始採用直接面向消費者的支付方式。只是好奇你們對你們這邊的任何倡議的想法。

We're not currently working on that. For us, we love the app stores. Obviously, it's hard to say that we wouldn't love it if they decrease their fees. Of course, we would. But we do really well with the App Stores. They're an amazing distribution mechanism. And the thing when you go off app stores is that it's harder to distribute your product. And also, it has quite a bit of friction. One of the amazing thing about the app stores is that people already have the credit card in there, and they just have to tap once and it gets to there.

我們目前沒有在做這方面的工作。對我們來說，我們喜歡應用商店。顯然，很難說如果他們降低費用我們會不喜歡。當然，我們會的。但我們在 App Store 方面做得非常好。它們是一種了不起的分配機制。當你離開應用程序商店時，事情就是更難分發你的產品。而且，它有相當多的摩擦。應用程序商店的一個令人驚奇的事情是人們已經在那裡擁有信用卡，他們只需點擊一次它就會到達那裡。

So you may be able to lower your fees, but you add so much friction to your users. But it's not clear to us that we would make that much more money. It's something that we're always thinking about and may eventually do it, but it's not something that we're currently working on too much.

所以你可以降低你的費用，但是你給你的用戶增加了很多摩擦。但我們並不清楚我們是否會賺那麼多錢。這是我們一直在考慮並最終可能會做的事情，但這並不是我們目前正在做太多的事情。

And remember, Mario, we did get a quite significant bump from Google lowering their app store fees last year. So...

請記住，馬里奧，去年谷歌降低了他們的應用商店費用，我們確實得到了相當大的提升。所以...

Got it. Helpful. And then second one just on kind of changes to the Duolingo app itself. I believe past few quarters, you talked about the change to the single learning track being at parity. Any updates to that? And just in general, how do you balance between kind of optimizing for learning and versus having a more casual app that kind of attracts the masses? So I think just user feedbacks, could be like a lab minority. But the removal of like things like oral prompts or type answers, I believe got some negative feedback. I just wanted to hear your thoughts there.

知道了。有幫助。然後是第二個關於 Duolingo 應用程序本身的變化。我相信在過去的幾個季度中，您談到了對單一學習軌道的改變。有什麼更新嗎？總的來說，您如何在優化學習與擁有更休閒的吸引大眾的應用程序之間取得平衡？所以我認為只是用戶反饋，可能就像實驗室的少數群體。但是刪除口頭提示或類型答案之類的東西，我相信會得到一些負面反饋。我只是想听聽你的想法。

Mario, you read Reddit.

馬里奧，你讀過 Reddit。

I do.

我願意。

Yes. So it's a great question about the changes to the app. So we made some pretty significant changes. We're always making the app better. I mean this is what we work on. This is why we spent so much on R&D. We're literally always making the app better. Over the last year, we made this large change to our home screen, going from -- giving a lot of choice about what to learn next to this linear path.

是的。所以這是一個關於應用程序更改的好問題。所以我們做了一些非常重要的改變。我們一直在使應用程序變得更好。我的意思是這就是我們的工作。這就是我們在研發上投入如此多資金的原因。我們實際上一直在使應用程序變得更好。在過去的一年裡，我們對我們的主屏幕進行了巨大的改變，從——在這條線性路徑旁邊提供了很多關於學習內容的選擇。

The main reason we did that linear path, by the way, was to teach better. We noticed that we were -- when we were giving people so much choice, what was going on was a significant fraction of them, we're just coming in to extend the streak with really easy lessons that we're not actually getting them to learn. Whereas now with the linear path, we've encouraged a significantly larger fraction of users to whenever they come to increase their streak, they actually also learn something. So the main reason we did that was actually to teach better and also to simplify our product, which has really helped us. But that was actually to teach better.

順便說一下，我們採用線性路徑的主要原因是為了更好地教學。我們注意到我們 - 當我們為人們提供如此多的選擇時，發生的事情是他們中的很大一部分，我們只是通過非常簡單的課程來延長連勝，而我們實際上並沒有讓他們接受學習。而現在有了線性路徑，我們鼓勵了很大一部分用戶在他們來增加他們的連勝時，他們實際上也學到了一些東西。所以我們這樣做的主要原因實際上是為了更好地教學並簡化我們的產品，這確實幫助了我們。但那實際上是為了教得更好。

The way we think of it is when we're making changes, we look at -- we have 3 things that we're improving. We're improving how well we teach. We're improving how engaging the app is, so basically, how much time people spend on it, how -- retention, et cetera, and we're improving how well we monetize. And we have large teams behind each one of these kind of group of metrics. And what happens is that each team, for example, there's the monetization team, what they're trying to do is get more people to subscribe. Whenever they run a test to try to get more people to subscribe, what we enforce is that they cannot screw up the other metrics. So yes, you can get more people to subscribe, but you cannot decrease our number of DAUs. You cannot decrease the time people spend learning, or you cannot decrease the learning outcome.

我們的想法是，當我們進行更改時，我們會查看 - 我們正在改進 3 件事。我們正在提高我們的教學水平。我們正在改進應用程序的吸引力，所以基本上，人們在上面花費了多少時間，如何 - 保留等等，我們正在改進我們的貨幣化程度。我們在這些指標組中的每一個背後都有龐大的團隊。發生的情況是每個團隊，例如貨幣化團隊，他們試圖做的是讓更多人訂閱。每當他們進行測試以嘗試讓更多人訂閱時，我們強制執行的是他們不能搞砸其他指標。所以是的，你可以讓更多人訂閱，但你不能減少我們的 DAU 數量。你不能減少人們花在學習上的時間，也不能減少學習成果。

Similarly, when we get people to -- the group of people that is dedicated to teach better, we tell them, okay, you can make changes that teach better, but you cannot screw up the monetization or you cannot screw up the user engagement. That's how we balance it. And we found that, that works pretty well. And most every change in the app, you can actually trace it down. That change was made by the group or by the team of people that is trying to make us grow faster or make us make more money, et cetera. And so that's how we do it. Hopefully, that gives you some color.

同樣，當我們讓人們——一群致力於更好地教學的人時，我們告訴他們，好吧，你可以做出改變來更好地教學，但你不能搞砸貨幣化，也不能搞砸用戶參與。這就是我們如何平衡它。我們發現，效果很好。應用程序中的大多數更改，您實際上都可以追踪到。這種改變是由試圖讓我們成長得更快或讓我們賺更多錢等的團體或團隊做出的。這就是我們的做法。希望這能給你一些顏色。

Our next question comes from Arvind Ramnani from Piper Sandler.

我們的下一個問題來自 Piper Sandler 的 Arvind Ramnani。

I guess my first question is, among the management team, who has the longest streak in terms of usage? But let me now ask that.

我想我的第一個問題是，在管理團隊中，誰的連續使用時間最長？但是現在讓我問一下。

But I don't know the answer to the question, but I'm willing to bet it's me, but I don't know the answer to the question.

但我不知道問題的答案，但我願意打賭是我，但我不知道問題的答案。

We can be pretty sure about that.

我們可以很確定這一點。

Yes, yes. So a really good metric with user engagement and kind of the revenue growth and bookings. But I know there's a whole lot of other metrics you track that -- beyond what you'll share with us. So can you -- conceptually, if you're not able to share hard numbers, are you able to conceptually share a bit more underneath on some of the metrics that's giving you confidence in kind of the shape of the growth for this year?

是的是的。因此，這是一個非常好的衡量用戶參與度以及收入增長和預訂量的指標。但我知道您跟踪的還有很多其他指標——超出了您將與我們分享的指標。那麼你能——從概念上講，如果你不能分享硬數據，你是否能夠在概念上分享一些讓你對今年的增長形式充滿信心的指標？

Yes. I mean I don't know if we can share precise numbers of -- there's certain things that we share and stuff that we don't share, but we do track a lot of metrics. In general, as I was saying before, there are 3 groups of metrics. There is what we call the growth metrics, which is just usually our entire user base. Are they using the product more? Do we have more DAUs, et cetera? There is the monetization metrics, which is how much money we make, basically. And then there's the metrics about how well we teach. In each one of them, we have a ton of metrics.

是的。我的意思是我不知道我們是否可以分享精確的數字——有些東西我們分享，有些東西我們不分享，但我們確實跟踪了很多指標。一般來說，正如我之前所說，有 3 組指標。我們稱之為增長指標，通常就是我們的整個用戶群。他們更多地使用該產品嗎？我們有更多的 DAU 等等嗎？有貨幣化指標，基本上就是我們賺了多少錢。然後是衡量我們教學水平的指標。在每一個中，我們都有大量的指標。

In terms of growth, some of the things that at least I look at, I look at time spent learning. I look at DAU to MAU ratio. I look at resurrection rates, so kind of number of people that are being resurrected. I look at the retention of resurrected users. I look at the retention of current users. I look at the retention of new users. So that's the type of stuff.

在成長方面，至少我看的一些事情，我看的是花在學習上的時間。我查看 DAU 與 MAU 的比率。我看復活率，所以有多少人被復活了。我看重複活用戶的留存率。我看的是當前用戶的保留率。我著眼於新用戶的保留。這就是事情的類型。

In monetization, the types of stuff we look at are things like conversion from free to paid, LTV, retention of paying users. Daily bookings is something we look at a lot, number of users or a fraction of users that have an IAP or do a purchase on an IAP.

在貨幣化方面，我們關注的東西類型包括從免費到付費的轉化、LTV、付費用戶的保留。每日預訂量是我們關注的很多方面，擁有 IAP 或在 IAP 上進行購買的用戶數量或用戶的一小部分。

In learning, we look at a number of things, but one of the main things we look at is difficulty of the content that we're able to give out without scaring users all away. So there's a lot of metrics that we track, and I don't even know all of them off the top of my head.

在學習中，我們會考慮很多事情，但我們關注的主要事情之一是我們能夠在不嚇跑用戶的情況下提供的內容的難度。所以我們跟踪了很多指標，我什至不知道所有這些指標。

Yes. That's helpful. And then kind of bookings growth, kind of sitting externally in some sense, it's a bit of a black box. But are you able to share like when you project for bookings growth, what are the inputs that go into kind of coming up to the type of projections for bookings growth? I mean we don't even know the specific numbers. But just conceptually, what goes into predicting user growth -- booking growth, sorry?

是的。這很有幫助。然後是預訂量的增長，在某種意義上有點像一個黑盒子。但是，當您預測預訂量增長時，您能否分享一下，進入預訂量增長預測類型的輸入是什麼？我的意思是我們甚至不知道具體數字。但從概念上講，預測用戶增長是什麼——預訂增長，抱歉？

We just take a ruler and go up like that. It works pretty well. I'm kidding. Matt has a much better answer than that.

我們只是拿著一把尺子，然後像那樣往上走。它工作得很好。我在開玩笑。馬特有比這更好的答案。

Well, maybe not. I mean the main thing to think about with bookings growth is that there's just the 2 dimensions. There's new bookings growth, new subscribers and there's renewals, and so -- on the subscription side. And then you can add in adds, obviously, and IAP.

好吧，也許不是。我的意思是預訂量增長要考慮的主要問題是只有兩個維度。在訂閱方面，有新的預訂增長、新訂戶和續訂等。然後你可以添加添加，很明顯，還有 IAP。

On the new users, Luis basically talked about how we have a really complicated model, a markup model, if you want to read our blog post on our growth model, how we predict users. An input into that is new users, and that's a combination of organic growth and marketing spend. So we predict new users. That flows into some element of new bookings. And then we're just basically doing cohort math for our cohorts of new users over time to see what we think they will convert into. And then we're adding into that for the subscription business, the renewal cohorts, which we have really good data on at this point. We feel pretty confident about. So you add that together, you get your subscription bookings. Add in adds and IAP, which again are a bit less detailed in the cohorts, make a little less sense for that. And then you got bookings. It's really that simple, Arvind.

關於新用戶，路易斯基本上談到了我們如何擁有一個非常複雜的模型，一個標記模型，如果你想閱讀我們關於增長模型的博客文章，我們如何預測用戶。對此的投入是新用戶，這是有機增長和營銷支出的結合。所以我們預測新用戶。這流入了新預訂的某些元素。然後我們基本上只是在為我們的新用戶隊列隨著時間的推移做隊列數學，看看我們認為他們會轉化成什麼。然後我們將訂閱業務添加到續訂隊列中，我們目前擁有非常好的數據。我們感到相當有信心。所以你把它們加在一起，你就得到了你的訂閱預訂。加上 adds 和 IAP，它們在隊列中的細節又有點少，因此意義不大。然後你得到了預訂。真的就這麼簡單，Arvind。

Great. Well, that's all we have for questions. So I'm going to turn it back over to Luis.

偉大的。好吧，這就是我們所有的問題。所以我要把它轉回給路易斯。

Thank you for all the excellent questions. And please, please, I beg you, do your Duolingo lesson.

感謝您提出的所有優秀問題。拜託，求求你，上 Duolingo 課吧。

Goodbye.

再見。