Edtech and SaaS: Will Anyone Be Able to Build Like Duolingo Again?
In response to Matt Tower's recent post and question on EdSheet
Day 17 of writing daily.
Duolingo has an amazing track record, founders who I deeply admire and don’t look like your typical founders1 so everything from the core problem they set out to solve (goal of helping millions around the world learn English) to their approach to product, growth and track record is incredible. They’re one of the few edtech companies that has continued to grow after the pandemic and ride the AI wave.
Last week, Matt Tower posed an interesting question over at The EdSheet2 given Duolingo’s latest results:
Why aren’t there more companies copying Duolingo’s product strategy?3
And with AI and product development, esp. for an example like Magic School, Matt calls out that the “novel company dimension that stands out to [him] is the velocity of product releases. These companies are prioritizing shots on goal rather than perfection with each launch, knowing that the best use cases for AI are emergent and the cost of building new software tools is dropping dramatically.”
I’d love to talk a litte more about Duolingo today, but also in a way that references the broader context of edtech. And this is hot take that I’ve thought in other contexts before, but never quite coalesced this way.
tl;dr:
Language turned out to be an amazing subject choice to build Duolingo (especially when learning foreign languages became the primary use case).
Learning in other academic or skill areas requires much larger investments in content and integrations (which takes away from product, AI/ML, data, etc).
It took a lot of resources and time for Duolingo to get here, which investors no longer have the appetite or patience to invest in anything that’s not AI.
㊗️ Language Learning as a Content Playground
Language turned out to be the perfect subject for this approach because
broad market appeal and demand4
the stakes are low
the baseline expectations of quality are low (the expectation of learning a foreign language in the US in school or with alternative options)
relatively straightforward content requirements: vocabulary and rules around pronunciation, vocabulary, grammar, syntax, writing & speaking
This mix of factors made it easier for Duolingo to scale content creation and apply AI/ML effectively. It’s harder to find other subjects5 that hit this sweet spot of broad demand and simplicity in potential scalable content creation.
📚 Investment in Content vs. Product
For more complex or higher-stakes content areas, the required investment in content development is much higher. This often diverts resources from pure product innovation to content creation and maintenance, making it harder to achieve the same level of product-led growth that Duolingo has.
Academically: things that are academic with higher stakes (math, ELA, science) run into higher stakes and the requirements of the system which drive higher investments in content
does this help a student/my kid pass (grade to grade)
does this raise their scores (for high stakes standardized test scores like SAT, ACT, NWEA MAP aligned to Common Core, etc.)
Skills-based: for things that are more skills/workforce based, many companies have historically (drastically) underestimated the amount you have to invest to keep content up to date, e.g., with coding/programming and languages, frameworks, and more changing regularly it's a lot more applied learning
additionally, the the question of do these programs get you hired (e.g., bootcamps in coding/programming) which while this has been notoriously hard to measure/validate, is super important to end users driving the threshold of content quality required to be much higher
B2C vs. B2B: I'm only talking B2C above if you actually went B2B (K12) your product investments had to go into integrations, data dashboarding, and other investments needed to go into GTM/sales, customer service on a different level
💰 Overall Investment Required To Scale
The flipside of B2C: it takes a huge investment and runway to grow.
To quote Matt again: “Duolingo followed a pattern of a long R&D period that ended with a product with minimal marginal costs (no instructors or implementation costs).” Funding this R&D these days typically seems to require something that engages the school system B2B vs. B2C. Sure some of this R&D could take place at universities but many across education and workforce seem to eschew the B2C business model because of concerns of equity and accessibility.
Duolingo has raised $180+ million (over 8 public rounds from Series A in 2012 to Series H in 2020 before raising more money via IPO in 20216)
They built and grew for four years before beginning to test monetization: from ads and in app purchases in 2016 to subscriptions in 2017.7
The key point here is not that it wasn’t worth it for Duolingo, but is anyone else going to be able to raise for this?
Yes, there have been a few other successes during this time (e.g., Quizlet, Newsela, and a few other PLG companies that have grown to significant traction across K12 edtech), but I’m not just indexing on the past, but look forward.
It doesn’t seem like it’ll be a pattern that can be replicated.
Yes, if you’re building in AI, including AI and edtech. Elsewhere? Call me skeptical.
More on Duolingo’s cofounder Luis von Ahn, including being born and raised in Guatemala, and that his mother (one of twelve siblings) was one of the first women there to earn a medical degree! And Severin Hacker who grew up in a small town outside Zurich, Switzerland.
Matt was previously writing ETCH, and recently announced an acquisition by and joining the team at Whiteboard Advisors!
In addition to Lenny’s great breakdown, this is one of my favorite all time pieces breaking down Duolingo here, with a great title - the anti bazaar (via Nib Nalin).
Albeit different than what Luis predicted: he expected global demand for learning English would be the driver, but ultimately it was demand in the US for learning foreign languages (and in some cases, English itself, at least according to some data/stats from 2017 about Duolingo use broken down by state).
Yes, Duolingo has branched into math and music, but even there the focus seems to be on operations/fluency and music symbols which link very well to the foundation in language that Duolingo has built.
Using the the 424B4 after the S-1 (the formal term for an IPO prospectus), you can see how much was sold by shareholders vs. primary to raise funds for Duolingo. Duolingo raised ~$355 million (before fees) from its IPO.
This was after SaaS/subscription revenue models were taking shape and taking off across B2C and B2B, and freemium to paid conversion has become the way that Duolingo makes money. An idea early on was perhaps becoming a translation service that can monetize B2B, according to that same Techcrunch article talking about Duolingo launching subscriptions.