Case file DL-2025-02 · Reopened for investors
Who killed Duo?
How a fake funeral became Duolingo’s most valuable campaign, and what investors should ask next.
- 2 weeks dead
- 0 lives left
- 50B XP to revive
- 1.7B impressions
Translate this sentence
¿Quién mató a Duo?
The scene / First impression
The day the owl died
On 11 February 2025, Duolingo posted a black-and-white notice: Duo, its green owl, was dead. The next day it hinted that the killer was a Cybertruck. For two weeks the internet mourned, joked and investigated, until learners earned enough lesson points to bring Duo back on 24 February (Snelling, 2025).
My first reaction was disbelief. Brands protect their mascots, yet Duolingo killed its most valuable asset on purpose. The idea did not even start in marketing. It came from the product team, testing a new app icon to win back inactive users (Snelling, 2025).
For investors, the numbers tell the story. The campaign cost almost nothing, yet it produced 1.7 billion organic impressions, and in the same quarter daily active users rose 49%. Paid subscribers also passed 10 million for the first time, up 40% on a year earlier (Duolingo, Inc., 2025).
This was not a stunt that happened to go viral. It was a retention tool dressed up as a murder mystery.
FIG 1 · CASE TIMELINE
- SEP 2024“Sick Duo” iconAbout 30M impressions in two weeks
- JAN 2025A/B testDead, anxious and chubby icons tested on inactive users
- 11 FEBDeath announcedBlack-and-white notice on social
- 12 FEBThe weaponHints point to a Cybertruck
- FEBBring Back DuoLearners pool 50 billion XP
- 24 FEBDuo returnsExactly two weeks later
- 1 MAYQ1 resultsRecord DAU growth reported
organic impressions in two weeks, against a 70M target (Snelling, 2025)
daily active users year on year, to 46.6 million in Q1 2025 (Duolingo, Inc., 2025)
quarterly revenue growth, to US$230.7 million (Duolingo, Inc., 2025)
The suspects / Segmentation, targeting & positioning
Who the killer was really after
Following the STP process from Module 4, the target was not new learners. It was lost ones.
Situation and segmentation
Duolingo’s mission is to develop the best education in the world and make it universally available. It grows through an “incredible free product” that people should want to return to every day (Duolingo, 2025a). Its biggest problem is not sign-ups but habits.
So the core segmentation variable is behaviour. The dead-owl icon was first tested on inactive users, and the stated aim was “new and resurrected users” (Duolingo, Inc., 2025). This is a consumption-specific variable in Wedel and Kamakura’s (2012) terms, supported by psychographics: people who enjoy dark humour and treat the brand like a friend online.
FIG 2 · SEGMENTS AND TARGET CHECK
| Type | Variable | How Duolingo used it |
|---|---|---|
| Soft | Usage and behaviour | Lapsed and inactive users, reached through the app icon and notifications |
| Soft | Psychographic | Meme-literate fans of dark humour; about 17M TikTok followers |
| Hard | Demographic | Gen Z and young millennials on social media |
| Hard | Geographic | Global, with 13 local accounts telling regional versions |
Lapsed-user segment checked against Kotler’s criteria (Module 4)
Targeting
The strategy is differentiated targeting with a hyper-personalised edge. Lapsed users met the story inside the app, through the icon and notifications, while social audiences followed it on TikTok, X and Instagram. Choosing the icon through A/B testing follows Module 4’s AI segmentation flow: data, model, segment, personalised activation.
The A/B test is the most revealing clue. The dead icon brought users back at about the same rate as the alternatives, and the team chose it because it told a better story (Snelling, 2025). Data showed it would not hurt. Creativity made it work.
Positioning
For people who want to learn a language but struggle to stick with it, the brand Duolingo is the free learning app that feels like a game among all language-learning apps, because its streaks, lessons and a mascot people love bring learners back every day.
Written with the Module 4 template (Kotler & Keller, 2016), this is an emotional selling proposition: learners bond with Duo, not with the syllabus. On the 4 Cs, clarity and consistency are strong, since the “Sick Duo” icon had already drawn about 30 million impressions (Snelling, 2025). Credibility is the risk, because an education brand known for chaos still has to prove that people learn. Competitiveness is strong but capped: Duolingo’s CMO has said the audience for its chaotic content is nearing its ceiling (Pulse, 2026).
FIG 3 · PERCEPTUAL MAP AND 4C SCORECARD
Duolingo owns the playful, free corner of the market almost alone. The open question is how long chaos alone can keep it there.
The crime scene / Social media presence
How a funeral went viral
Dodson’s (2016) four-stage process from Module 5 shows a near-textbook campaign whose weak point is attribution.
FIG 4 · THE FOUR STAGES, APPLIED
Goals
- Win back inactive users
- Keep Duo’s persona alive
- Target: 70M impressions
Channels
- TikTok: the serial
- X: the death notice
- Instagram: local accounts
- App icon and XP page
Implementation
- Six days to prepare
- Plot written daily with fans
- Dua Lipa and other brands joined
Analysis
- Impressions, UGC, mentions
- User lift claimed, not quantified
Goals, channels and implementation
Unlike many brand campaigns, this one had a measurable target: 70 million impressions, against about 100 million for Duolingo’s biggest earlier campaigns (Snelling, 2025).
TikTok led, with the Cybertruck video reaching 25.7 million views and the coffin video 66.3 million (Snelling, 2025). X carried the death notice; Dua Lipa replied there, and her reshare added 22 million views (Weldon, 2025). Inside the app, the icon changed, and a “Bring Back Duo” page asked learners to earn 50 billion XP together, with countries ranked against each other. The marketing team had six days to prepare, then wrote the plot day by day as fans reacted, and brands such as KitKat, Subway and Hilton joined in (Snelling, 2025).
The XP challenge is what sets this apart. It turned followers into a brand community that cooperated and competed, rather than a brand public that simply watched (Module 5).
FIG 5 · WHAT THE EVIDENCE PROVES
Analysis
On Module 5’s measurement ladder, the evidence reaches the top rung, behaviour: users really did come back. The gap is attribution. The shareholder letter claims a “meaningful lift” but gives no figure, and the 49% growth had other causes too (Duolingo, Inc., 2025). Reactions were also mixed; after Duo returned, some people asked what it had all been for (NPR, 2025).
Three things made it work. The story was a serial, so people came back every day for the next clue, which is the same habit Duolingo wants for its lessons. The brand reacted in real time, which Module 5 says gives social media its instant, human feel, along with the risk of slip-ups. And the ending was participatory: fans could not just watch Duo return, they had to earn it.
Choosing a Cybertruck as the weapon was read as a jab at Tesla and Elon Musk (Snelling, 2025). It added fuel, but it pulled an education brand into politics. For me, that was the campaign’s one unnecessary risk.
The alibi / International market presence
The one country where Duo never died
Through Module 11’s decision-making process, the campaign was global by default and local where it mattered.
FIG 6 · INTERNATIONAL MARKETING DECISION-MAKING PROCESS
↺ feedback and reassess continually
Markets, entry and rollout
The rollout used a sprinkler strategy: the death was announced worldwide on the same day, with one icon and one 50-billion-XP goal. More than half of Duolingo’s bookings and revenue come from outside the US (Duolingo, Inc., 2025). As a digital product it enters markets directly through app stores, a high-control, low-commitment mode supported by regional marketing teams.
日本JAPAN · THE ALIBI
In Japan, CEO Luis von Ahn said joking about death is “not as kosher”, so Duo was “just not dead”. The Japanese account used Japanese-language, anime-style posts instead (Business Insider, 2025).
Thirteen local accounts told regional versions of the story (Snelling, 2025), and a McDonald’s partnership in Brazil was paused to make room (Weldon, 2025). This is product extension with communication adaptation, and Japan’s version is local consumer culture positioning. Zou and Cavusgil’s (2002) adaptation drivers, cultural differences and consumer behaviour, explain the exception.
For investors, this matters because so much of Duolingo’s revenue comes from outside the US. One culturally careless joke could cost more in a key market than a viral campaign gains.
FIG 7 · HOW MUCH WAS ADAPTED?
Customs and taboos explain Japan, but they apply elsewhere too. In my experience in Malaysia, many families avoid joking about death and funerals. I would want local teams to assess every shock campaign before launch, not react afterwards.
Hofstede’s dimensions, a cross-national lens covered in Module 11 (Agarwal et al., 2010), point the same way: cultures high in uncertainty avoidance may find shock humour riskier. Others, such as Mexico with Día de Muertos, already joke about death and could have had a richer local version.
The autopsy / Sustainable marketing
What the autopsy found
Following Lim’s (2022) sustainability pyramid from Module 10, I examine the campaign in order: economic, social, then environmental.
FIG 8 · TRIPLE BOTTOM LINE SCORECARD
Economic
Strong- Near-zero campaign cost
- Revenue +38% in Q1 2025
- Risk: AI-first job backlash
Social
Mixed- Buzz converted into lessons
- Free education for all
- Risk: guilt as motivation
Environmental
Absent- Digital campaign, small footprint
- No emissions reporting
- Growing AI energy use
Harm chain: unintended harms found by critical marketing
Kemper and Ballantine’s (2019) levels of sustainability marketing
Economic
The campaign cost almost nothing, while quarterly revenue grew 38% to US$230.7 million (Duolingo, Inc., 2025). In 2025 Duolingo passed 50 million daily users and US$1 billion in bookings (Duolingo, Inc., 2026), although not all of that is the campaign’s doing. The economic risk came weeks later. An “AI-first” memo about relying less on contractors triggered a backlash, and the CEO later admitted it lacked context (Fortune, 2025). Goodwill built on social media can reverse just as fast.
Social
Socially, the campaign turned attention into learning. Reviving Duo required lessons, not shares, which gives it the behavioural goal and motivating exchange of social marketing (Andreasen, 2002) and supports SDGs 4 and 10 through a free product. The harm-chain view of critical marketing (Polonsky et al., 2003, as cited in Gordon et al., 2011) finds the costs: death jokes may hurt people who are grieving, and “study or Duo dies” motivates through guilt rather than progress. Even the CEO now says Duolingo is prioritising “teaching better” (Duolingo, Inc., 2026).
Environmental
The campaign was almost entirely digital, but Duolingo’s 2025 social impact report says nothing about emissions or the SDGs (Duolingo, 2025b). As an “AI-first” company it relies on generative AI, and data-centre electricity demand is expected to more than double by 2030 (International Energy Agency, 2025). On Kemper and Ballantine’s (2019) scale, Duolingo is reformative on education but shows no visible environmental effort. The silence is itself a finding: Principle 7 of the UN Global Compact, covered in Module 10, asks firms to support a precautionary approach to environmental challenges.
Porter and Kramer (2006) argue that social impact must fit strategy, and Duolingo’s does: unhinged marketing made people study. But if the motive is guilt rather than growth, and the company stays silent on climate and jobs, its education mission will face hard questions.
The verdict / Conclusion and recommendations
The verdict
The killer was Duolingo. The motive was retention.
The segmentation was precise, the social execution near-textbook and the international rollout culturally aware. The autopsy, though, found three risks: guilt-driven learning, a blank environmental record and a strategy nearing its ceiling. Each risk points to one of the recommendations below, and each asks Duolingo to turn its talent for attention into proof of learning.
FIG 9 · RECOMMENDED ROADMAP FOR DUOLINGO
Make progress the plot
End the next serial with what learners achieved, not with “study or Duo dies”.
Fixes: guilt-driven motivation (The autopsy)Map the taboos
Have local teams review shock content before launch, and build local versions for Southeast Asia.
Fixes: only Japan adapted (The alibi)Publish the real numbers
Report campaign attribution, AI energy use and the effect of AI on jobs.
Fixes: attribution, environment (The crime scene, The autopsy)A mascot that earns 1.7 billion impressions at almost no cost is a rare marketing asset (The crime scene).
Ask how much growth the campaign really drove, how long chaos can keep working, and how “AI-first” affects goodwill (The suspects, The crime scene, The autopsy).
CASE CLOSED
The owl survived. The question is whether the strategy can.
Sources / APA 7th
References
Campaign evidence
- Business Insider. (2025). Duolingo killed its cartoon owl mascot for another ‘unhinged’ marketing stunt, except in one country. AOL. https://aol.com/duolingo-killed-cartoon-owl-mascot-075514172.html
- Duolingo. (2025a, April 7). Duolingo company strategy. https://blog.duolingo.com/duolingo-company-strategy/
- Duolingo. (2025b). Social impact report 2025. https://blog.duolingo.com/social-impact-report-2025
- Duolingo, Inc. (2025, May 1). Q1 2025 shareholder letter [Exhibit to Form 8-K]. U.S. Securities and Exchange Commission. https://www.sec.gov/Archives/edgar/data/1562088/000156208825000098/q1fy25duolingo3-31x25share.htm
- Duolingo, Inc. (2026, February 26). Duolingo reports fourth quarter and full year 2025 results. https://investors.duolingo.com/news-releases/news-release-details/duolingo-reports-fourth-quarter-and-full-year-2025-results
- Fortune. (2025, August 18). Duolingo CEO admits controversial AI memo did not give enough context, insists company never laid off full-time employees. https://fortune.com/2025/08/18/duolingo-ceo-admits-controversial-ai-memo-did-not-give-enough-context-insists-company-never-laid-off-full-time-employees/
- Gaillot, A.-D., & Tarasova, E. (2025, February 20). Duo is dead: The data on Duolingo’s viral campaign. Meltwater. https://www.meltwater.com/en/blog/duolingo-dead-mascot-campaign
- International Energy Agency. (2025). Energy and AI. https://www.iea.org/reports/energy-and-ai
- NPR. (2025, February 26). Duolingo’s owl mascot is alive after all. What did it gain from faking his death? https://npr.org/2025/02/26/nx-s1-5309785/duolingo-owl-mascot-lives
- Pulse. (2026). Duolingo won the internet with chaos. Now its CMO says it’s time to rein it in. https://www.pulse.bot/cmo-pulse/news/duolingo-won-the-internet-with-chaos-now-its-cmo-says-its-time-to-rein-it-in-6f3f693c-88a1-4366-ac70-d6eac95d557f/
- Snelling, G. (2025, April 10). [Article on how Duolingo built the “Dead Duo” social media campaign]. Fast Company. https://www.fastcompany.com/91313082/duolingo-dead-duo-owl-social-media-campaign
- Weldon, C. (2025, April 22). Duolingo shares PR secrets of viral ‘Death of Duo’ campaign. PR Daily. https://www.prdaily.com/duolingo-shares-pr-secrets-of-viral-death-of-duo-campaign/
Theory (from Modules 4, 5, 10 and 11)
- Agarwal, J., Malhotra, N. K., & Bolton, R. N. (2010). A cross-national and cross-cultural approach to global market segmentation: An application using consumers’ perceived service quality. Journal of International Marketing, 18(3), 18–40.
- Andreasen, A. R. (2002). Marketing social marketing in the social change marketplace. Journal of Public Policy & Marketing, 21(1), 3–13.
- Dodson, I. (2016). The art of digital marketing: The definitive guide to creating strategic, targeted, and measurable online campaigns. John Wiley & Sons.
- Gordon, R., Carrigan, M., & Hastings, G. (2011). A framework for sustainable marketing. Marketing Theory, 11(2), 143–163.
- Kemper, J. A., & Ballantine, P. W. (2019). What do we mean by sustainability marketing? Journal of Marketing Management, 35(3–4), 277–309.
- Kotler, P., & Keller, K. L. (2016). Marketing management (15th global ed.). Pearson Education.
- Lim, W. M. (2022). The sustainability pyramid: A hierarchical approach to greater sustainability and the United Nations Sustainable Development Goals with implications for marketing theory, practice, and public policy. Australasian Marketing Journal, 30(2), 142–150. https://doi.org/10.1177/18393349211069152
- Palmer, R. A., & Millier, P. (2004). Segmentation: Identification, intuition, and implementation. Industrial Marketing Management, 33(8), 779–785.
- Porter, M. E., & Kramer, M. R. (2006). Strategy and society: The link between competitive advantage and corporate social responsibility. Harvard Business Review, 84(12), 78–92.
- Wedel, M., & Kamakura, W. A. (2012). Market segmentation: Conceptual and methodological foundations (2nd ed.). Springer.
- Zou, S., & Cavusgil, S. T. (2002). The GMS: A broad conceptualization of global marketing strategy and its effect on firm performance. Journal of Marketing, 66(4), 40–56.
Generative AI acknowledgement
I used Claude, an AI assistant made by Anthropic and accessed through claude.ai, between 12 and 14 November 2026 to help me plan, research, write and build this digital story. I asked Claude to analyse the assessment brief and marking guide and to compare possible campaigns, and after developing an earlier version on Dove’s “The Code”, I chose Duolingo’s “The Death of Duo” and asked Claude to assess whether it suited the brief. Claude read my Module 4, 5, 10 and 11 lesson notes to identify the frameworks applied here, searched the web for evidence about the campaign and checked the figures it used against the sources in the reference list. It then drafted a planning blueprint, the English text of this page, draft wording for the personal comments, the scripts for my two videos and the APA references, and it wrote the HTML, CSS and JavaScript for the page, including the nine figures, the interactive lesson card, the animations and the video players; it also helped me publish the page on Cloudflare Pages and convert it to PDF. I rewrote most of the page text in my own words, including the personal comments, and checked the final wording and figures against the sources myself. I produced both videos with an AI video generation tool, using the scripts drafted with Claude: the presenter in the first-impression video and the people in the “Funny or too far?” video are AI-generated rather than real people I filmed or interviewed, so the views they express are illustrative scenarios, not primary research data. I take responsibility for the final content, including the choice of campaign, the arguments and the recommendations, and the prompts I used are listed below in the order I used them, with prompts originally written in Chinese given in English translation.
Appendix: GenAI prompts
- “Study this assignment.” (with the assessment brief and marking guide attached)
- “Not yet.” (in reply to whether I had chosen a campaign)
- “I choose option 1.” (Dove, “The Code”)
- “Based on the requirements above, here are the module notes to read.” (with the Module 4, 5, 10 and 11 lesson notes attached)
- “Check whether Dove posted The Code on X (Twitter). If not, delete the X row from the social media platform table; it can be deleted. Screen 6 needs another source on AI energy use; add a suitable one. Give me the video scripts.”
- “Based on the blueprint, design the complete web page. The videos can be temporary, but keep a place for them (YouTube links). The overall design should follow the blueprint’s information, and the UI/UX should match Dove’s design, colours and elements.”
- “Now I want a new topic. The campaign I want to do is Duolingo’s The Death of Duo from 2025. First study whether this campaign is suitable.”
- “Make a Duolingo blueprint; create a new blueprint.”
- “Based on the Duolingo ‘The Death of Duo’ digital story web page blueprint, design a complete web page with the same design approach as the analysis above.”
- “All of these need to be removed, and the 00/01/03/04 numbers can be removed.” (with screenshots of the footer, word count and appendix)
- “Give me the scripts for the two videos.”
- “this can be remove, the home page can make it more attracting with duolingo theme”
- “this is the two link for both video section” (with my two YouTube links)
- “great, add animation effect on this webpage”
- “now i need the zip file to publish it on clouflare, where is the location of the file”
- “checking my Cloudflare cant show this website”, followed by “u direct complete it”, “im in ald” and “use Chrome” while Claude published the site
- “study the link and prepare the PDF file for me”
- “the pdf should be 100% same as website in all design and content”, followed by requests to fix the pink section and pink line in the PDF
- “now check the webpage, click the button can direct play video”
- “this one can be remove in pdf and webpage” (referring to this acknowledgement)
- “add back the generative ai acknowledge back and generate it in full paragraph, can study back the information”
- “i need it in paragraph can combine into one paragraph, all need in english”