Duolingo is hiring a General Manager for India because the market has "much more potential than we've realized so far", and the new performance marketing lead is asked to fix "onboarding and landing experiences for paid users". English for Hindi speakers already has 19.2M learners, Bengali speakers 6.0M, Tamil 1.4M and Telugu 1.1M. The Hindi site is slow for its own visitors, though. Google's field data for hi.duolingo.com shows real Chrome users waiting 5.7s for anything to appear and 5.6s for the main content, with taps taking 555 ms to respond, all failing Google's limits. The page arrives as an empty shell titled "Duolingo", and nothing shows until the app scripts have loaded, in a page that weighs about 4.5 MB. On the mid-range Android phones and patchy mobile data common in India, that wait comes before the "Get started" button.
First action
Build a static, server-rendered version of the hi.duolingo.com homepage and test it against the current one on 10% of traffic.
How to measure it
Hindi-site visitors who start a course
- Now
- –
- Target
- +20%
- By
- 1 quarter
Steps
Baseline
Sign-up rate, bounce and day-7 retention for hi.duolingo.com visitors.
Senior Marketing Analytics Manager 1 week
Define the light page
Headline, course cards and one button; no app bundle until a tap.
General Manager, India 1 week
Build it
Server-rendered template for language-site homepages and course pages.
Web engineering lead 3 weeks
Defer heavy scripts
reCAPTCHA, sign-in and pixels only once sign-up starts.
Web engineering lead 1 week
Localize copy
Hindi, Bengali, Tamil and Telugu versions reviewed by native speakers.
Senior Localization Program Manager 2 weeks
A/B test
10% of hi.duolingo.com traffic, then 50%.
Senior Data Scientist, User Growth 4 weeks
Point paid web traffic
Indian web campaigns land on the fast pages.
Senior Director of Performance Marketing 2 weeks
Roll out
Other language sites by traffic.
Web engineering lead 1 quarter
Why it works
Give the language sites and course pages a lightweight, server-rendered front door: the headline, course list and "Get started" button in the first HTML response, with the full learning app loaded only after a tap. Start with hi.duolingo.com and its 12 sitemap pages, then the Bengali, Tamil and Telugu sites. Defer reCAPTCHA, Google sign-in and ad pixels until a visitor starts sign-up. Run it as an A/B test on web sign-ups and day-7 retention, and point Indian paid web traffic at the fast pages.
Every second of blank screen loses visitors, and most of all on slower phones and networks. The audience is already proven by the course numbers. A faster first screen turns more existing Indian interest into DAUs without extra media spend, and the same pages make paid web traffic cheaper.
Ready to use
ChecklistLight front door checklist
[ ] Title and description in the first HTML [ ] Headline and course cards rendered on the server [ ] "Get started" works before the app bundle loads [ ] reCAPTCHA and Google sign-in load on sign-up only [ ] Ad pixels after consent or first interaction [ ] First content under 1.8s and main content under 2.5s on a mid-range Android [ ] hreflang to the other language versions [ ] A/B test on sign-ups and day-7 retention
Risks and how to handle them
Two templates to maintain
Generate the light page from the same content source as the app.
Sign-up spam without early reCAPTCHA
Load it when the sign-up form opens, not on page load.
Evidence
- HiringGeneral Manager, India (posted October 12, 2026, Bangalore): "India is one of Duolingo's most important growth markets. We've operated there for a while, but believe it has much more potential than we've realized so far." Success "will be measured on Duolingo's core metrics: user growth and monetization"; the GM leads an India-based cross-functional team and is "the voice of India".Source
- HiringSenior Director of Performance Marketing (posted October 16, 2026): paid acquisition "across 100+ markets and our language, chess, and math products" across Google, Apple Search Ads, Meta, TikTok, AppLovin and DSPs; "accountable for proving what our spend actually buys" with "incrementality testing, saturation modeling, and LTV and ROI frameworks that Finance trusts"; partner with Product and Engineering on "onboarding and landing experiences for paid users, and tests that connect acquisition to retention"; "Flex spend with the organic calendar, including product launches, seasonal swings, and viral moments".Source
- Course catalogDuolingo's public course list (checked October 21, 2026): 313 courses taught from 45 languages. The largest is English for Spanish speakers (54.1M learners); English for Hindi speakers has 19.2M, English for Bengali speakers 6.0M, English for Tamil speakers 1.4M and English for Telugu speakers 1.1M; Hindi for English speakers has 13.6M.Source
- Google PageSpeedhi.duolingo.com (the Hindi site) on mobile: real Chrome visitors at the 75th percentile wait 5.7s for first content and 5.6s for the main content, and taps take 555 ms to respond; all three fail Google's limits (1.8s, 2.5s, 200 ms). In the lab: performance 29/100, main content after 16.2s, 4.5 MB downloaded.Source
- Your websiteThe HTML sent for www.duolingo.com/course/es/en/Learn-Spanish is 8.8 KB with the title "Duolingo" and no visible text; the Hindi site's English course page, /math, /super, /pricing and the homepage are the same: the page title and content are filled in by JavaScript. The chess glossary pages are the exception, with real titles in the HTML ("What Is a Bishop Fork in Chess?"). Real visitors to the Learn Spanish page see the main content at 3.7s (p75).Source
- Shareholder letterQ2 2026 (SEC filing): DAUs 58.7M, up 23%; MAUs 140.6M; paid subscribers 12.7M, up 17%; revenue $298.5M, up 18%; advertising revenue $21.1M, up 2%; current user retention at an all-time high of 84%; the one-time Streak Revival in June brought back 15.4 million learners, "nearly 8 million" of whom had no active streak; influencer content supplied about two-thirds of social impressions in China, Indonesia and India; DAUs include users of "the learning section of the website".Source