Duolingo: Disrupting the $7.3T Education Market.
Q3 2024 ER
Edited by Brian Birnbaum and an update of my original Duolingo deep dive.
Seemingly a harmless language app, Duolingo is building an infrastructure that promises to disrupt the global education market, delivering radically improved student outcomes at a fraction of the current cost over time.
Process power is the force that enables the execution of an organisational algorithm. By the latter I refer to a series of steps that are performed daily, with the ultimate goal of enhancing shareholder value over time. In the network-defined economy, moats tend to be invisible because information is abundant. Indeed, the only reason that one should participate in another network is that others choose to do so beforehand. And they often choose to do so driven by largely invisible forces, that are the result of world-class process power sustained over time. Process power is thus a novel form of moat that in many cases the market hasn’t learned to identify yet, before it equates to extraordinary financial performance.
Duolingo’s organisational algorithm consists of enhancing student motivation daily. Higher levels of motivation lead to better learning outcomes, which drives higher engagement across the board and thus equates to a rising earning power. This is the sole driver of Duolingo’s financial performance, which is elegantly summarised by the graph below. Duolingo’s free cash flow per share is up considerably since 2021, driven by the organisation’s ability to delight students across the globe despite abundant and free educational content on the Internet. In theory students should have no reason to pay Duolingo, but the company’s unreasonable focus on enhancing student motivation offsets the availability of free content.
One of the main lessons that I have learned in the past five years is that betting on organisations with extraordinary process power tends to equate to success. At times the market spins narratives that bring stock prices down, but ultimately process power continues equating to improved earning power over time and thus higher levels of free cash flow per share, which brings stock prices back up. In Q3 2024 Duolingo’s DAUs (daily active users) were up 50% YoY and once again, Duolingo’s co-founder and CEO Luis von Ahn wasn’t able to point to any specific driver. The repeated absence of any concrete driver necessarily signals that performance is driven by a collection of small drivers, which is in the abstract a definition of process power.
Luis’s words in the Q3 2024 earnings call are highly insightful:
Our 50%+ year over year DAU growth is especially impressive considering that we’re building on 60%+ growth in the third quarter of last year.
We have been able to sustain this growth rate because we constantly improve features that were already considered home runs, like launching Friends Streak as an extension of one of our most powerful retention mechanics.
What we’ve learned from over a decade of building habit-forming products is that these minor improvements compound over time.
My major concern about Duolingo when I first studied the company was that over 50% of revenue was coming from English speaking countries, when Duolingo’s fundamental value proposition at present is teach folks English. In many countries around the world, learning English immediately equates to a higher salary. In turn, Duolingo’s user activity in English-speaking countries is fundamentally recreational, putting the corresponding percentage of overall revenue at risk during a real economic downturn. As of Q3 2024, Duolingo is now making great progress on this front, with over 2M DAUs learning intermediate English on the platform. According to management, this growth is “pretty fast.”
When I wrote my deep dive, I couldn’t have foreseen that Duolingo was going to be well on its way to fixing the above issue by Q3 2024. But I did see the company’s extraordinary process power and thus, did assign a high probability of Duolingo fixing pretty much any issue that it finds along its to path towards maximising student motivation and shareholder returns over time. In the concluding section of my original Duolingo deep dive, I wrote:
Although the company is richly valued at a price to sales ratio of 20, I am tempted to make an investment. Companies with this level of process power tend to do well over time and especially so in a winner-takes-all economy, defined by networks.
I can clearly see what Duolingo is going to evolve into, and I wish that I had spotted this company earlier.
I have often explained that culture is the underlying enabler of process power. Culture is what makes people go to work everyday obsessing about a few particular things, which then equates to the unreasonable focus on the end customer and the unusually fast pace of iteration. However, an often under appreciated technical driver of process power is a decreasing cost of change, as I explain in my original Tesla deep dive. For the first time ever, in the Q3 2024 earnings call I got an insight into how Duolingo decreases the cost of change over time, by increasing the pace of experimental frequency.
CEO Luis von Ahn said the following about this during the call:
The good news is we're running -- the number of tests that we're running per quarter is increasing, so we are running more tests per quarter.
The success rate for the test is about the same. And it's actually a funny thing, it's about almost exactly 50%.
So every test that we run has about a 50% chance of succeeding. And by succeeding, I mean that it did what we wanted it to do, like it increased whatever metric we wanted it to do to increase.
And CFO Matthew Skaruppa added:
And Andrew, the other thing I would add is just that the teams do run experiments like from time to time to try to enable it themselves, to speed up their own experimental frequency, to get data faster, to analyse the experiments more rapidly. And so we're always running trying to run those. We're going to run some this quarter, for example. So we think things can speed up.
Per the mental framework that I teach in my Tech Stock Goldmine course, the Duolingo thesis could be reduced to a single KPI (key performance indicator): the speed of experimental frequency. So long as it continues to go up over time, the company’s free cash flow per share levels will likely track along. Indeed, Duolingo’s business is not only about increasing engagement, but also getting better at monetising said engagement over time. However, both skills derive from a single sub-skill, which is optimising the user experience. And ultimately, speed of experimental frequency is the best predictor of how user experience will evolve.
Additionally, Duolingo now seems to be positioned to meaningfully increase its operating leverage going forward. Generative AI is exponentially decreasing the cost of high performing content creation. And by high performing, I mean content that drives incremental user engagement for Duolingo. In Q3 2024 Duolingo announced the launch of Lily, an AI agent that Duolingo users will be able to converse with and practice any language. This feature will be able available in Duolingo’s highest priced tier, called Duolingo Max. Lily promises to radically accelerate the learning curve of intermediate and advanced English learners, but at a marginal cost.
This type of feature therefore positions Duolingo to meaningfully increase its earning power. And per Duolingo’s process power, I believe this will likely not be the last feature with such potential. As the cost of training LLMs (large language models) and making inferences with them continues to decline, Duolingo’s proprietary data moat is set to shine. I’ve heard some folks on X say that ChatGPT will disrupt Duolingo (which is a narrative that I’d love to see amplified), but at present no one has more granular data on what increases or decreases student motivation, across a growing range of subjects. This data will most likely eventually create an AI super-teacher, that will disrupt the entire educational system.
In my original deep dive I said that I wished I found this company earlier, which remains true. However, valued at just $15.48B I believe its high price to sales ratio of 24.4 will seem reasonable in hindsight. As often happens in technology, Duolingo seems like a harmless toy to the broader education industry. However, the infrastructure that they are building is orders of magnitude more powerful than it seems. And I believe this infrastructure is about to rear its head, in the coming year or two. It will be very hard for traditional education companies to compete with Duolingo. Indeed, Duolingo is currently pointing at a $7.3T opportunity - which is the size of the global education market.
Until next time!
⚡ If you enjoyed the post, please feel free to share with friends, drop a like and leave me a comment.
You can also reach me at:
Twitter: @alc2022
LinkedIn: antoniolinaresc