Duolingo (DUOL) is a gamified language learning app with new offerings for learning math, chess, and music. It started in 2011 as an app for people to learn English. DUOL’s English learning and testing offering, accepted by institutions as a certificate of proficiency, makes up most of its revenues.

DUOL has grown at a 40% CAGR for the last three years and is one of the hottest edtech stocks in the world. It trades at 30x P/S and 261 P/E. Those metrics may look outlandish, but they are immaterial for a high-growth company, supposedly at the beginning of its growth stage.

There is not much wrong here or to critique. I nitpicked on the SG&A, but if I said much about it, I would be laughed out of the room.

That said, it is that time of the year, all the 10Ks are out, time for the Orca to go hunting.

For the uninitiated, that is the result of the Beneish M-Score test. This test became famous for exposing Enron as a fraud. Now, just because a company fails the test doesn’t mean it is a fraud. One would need to dig deeper into regulatory filings and decipher what is going on.

In this case, “Declining Depreciation Rate” and “Increased Borrowing” are immaterial and don’t need to be investigated. However, “Expense Capitalization” and “Sales Growth” for sure need to be looked at.

Red Flags Galore

To question the sales growth and the reported numbers, we have to have some “there-there” or hints or even conclusive evidence of shenanigans.

Let’s look at the “User Metrics” reported for Q1 2025. The following is from the Q1 2025 shareholder letter.

Looks awesome; what’s not to like? However, consider the following:

Approximately?

For this blog, I can tell you “EXACTLY” what those numbers are right now or anytime anybody asks within two minutes. And you are telling me a company with $811.2 million in LTM revenue is giving approximate numbers in its regulatory filings? What is the sanctity of anything else that is being reported?

This raises another important issue: differential disclosure. In the shareholder letter, DUOL portrays the numbers as exact but hedges them as approximate in the 10Q. This raises questions about the integrity of the management.

In my younger days, as a small investor, I would have just stopped here and moved on to other things in life. However, I have to complete this article, so let’s plug along.

Essentially, a lot of hedging is going on here. But you may wonder: what about the auditors? Fair point. I will get to them a little later.

TAM

DUOL is still primarily a language learning platform. Half the users are learning English, 20% are learning Spanish, and 10% are learning French. However, DUOL highlights its total addressable users as 1.8 billion (the population of the world is 8.2 billion) who would want to learn English. In isolation, the number would seem awesome; however, let’s put some context.

5.5 to 5.6 billion people have internet connectivity. Of those, how many people would not have learnt English in school? Then, out of those who may not have, how many can afford over $70 per year in DUOL’s subscription fees? You guys are smart, you can figure out where I am going with this.

That said, let me give you a real example. DUOL’s English language learning course is available in India for $70 per year. Those who know India would be rolling on the floor laughing. The majority of the population can read and write English. Those who can’t, can’t afford $70 per year. Also, DUOL is not going to fix the call center guys’ accent, and those without the accent don’t need DUOL. Heck, a Netflix subscription is $2 per month in India. Basically, 1.4 billion people in the world don’t need DUOL.

Now, extrapolate the same logic country by country. It is a pointless exercise, though. How many Italians or Argentines would need to go out of their way to learn English on DUOL? Basically, this company is throwing out numbers that nobody is verifying or questioning.

Other Offerings

DUOL has recently added learning offerings for math, music, and chess. Learning math and music on a gamified app? It is a joke, and it gets better.

In the recent earnings call, DUOL said that the chess offering was built by two people using AI who have no programming experience and can’t play chess. Ok, I appreciate that the management might have been trying to highlight their AI capabilities.

However, this is gaslighting at its finest. I hope the offering was properly tested and QAed. If not, DUOL better be ready to handle a lot of chargebacks or refund requests from customers.

This brings us to the Balance Sheet and Cash from Operations.

Liquidity

DUOL has no debt and one billion in cash on the Balance Sheet. Comforting, right?

While DUOL has a no-refund policy, in the event it gets hit with a spate of refund requests, chargebacks, or disputes, that cash will burn so fast, they will not know what hit them. Can that happen? Possibly. As DUOL relies more and more on AI, remember the chess example above; if customers feel that they are not getting a quality product, DUOL will be in trouble.

Trust me, no merchant wants to get into a dispute with a customer, even when the merchant is at no fault, as the credit card issuer will play hardball and the majority of the time side with the customer. As a merchant, it is best to come to an amicable solution with the customer before the customer disputes the charge.

Hence, the $415 million in deferred revenue can vanish in a heartbeat and could end up causing cash flow problems.

Expense Capitalization

As highlighted in the Beneish M-Score result, “Expense Capitalization” needs to be investigated. So, what is expense capitalization?

Some expenses can be recorded as assets, mostly intangible, on the balance sheet and amortized over time. Essentially, the expense gets spread out over the period the expense is expected to provide benefit. This reduces the expense being recorded in the income statement. Now, while there are guidelines from FASB on what can be capitalized and what can’t, the management has wide leeway.

Given DUOL’s business, research and development, including software development, is a potential area where management can use their judgment to either capitalize or expense.

Now the fun part: Deloitte & Touche, DUOL’s auditor, highlighted “Capitalized Software” as a critical audit matter in the 2024 10K (Refer to PDF Page 82). So, the Beneish M-Score flags it, the auditor highlights it, and I disagree with the auditor’s and management’s interpretation.

In 2024, DUOL capitalized $9.0 million. Hence, instead of an operating income of $63.8 million, DUOL would have reported $54.8 million. Also, as of December 31, 2024, total capitalized software development costs are $35.7 million (includes the $9 million).

Hence, DUOL has been doing it for a while, and it would have made things worse in the context of what we have already discussed, as it would have wrecked its EPS (loss/growth) and potentially caused it to miss the Street’s estimate.

You might be wondering how common this is. Well, it’s pretty common, and people get away with it. This is basically shifting current expenses to a later period. Now, you may counter that the numbers are small in the scheme of things. Hey, all you need is a penny here and a penny there to beat the Street’s estimates.

Verdict

DUOL’s Q1 revenues were bolstered by a viral video they posted on YouTube and social media. I look at spikes caused by viral content as one-offs and a fad rather than a sustainable strategy. It is confirmed by the management that most of their business is returning or reengaging customers or repeats, rather than organic word of mouth and sustained, slow, and steady acquisition.

At the early stage of a company, one can push the boundaries for various reasons, and some missteps are not fatal; however, DUOL doesn’t have that luxury. One misdirected social media post or video, and disaster will strike.

On this blog, I don’t accuse any company of fraud, as the burden of proof for that is much higher than I have presented here. Hence, please exercise your judgment before arriving at any conclusions.

That said, I have serious concerns about DUOL. I believe DUOL will underperform the market over 18 to 24 months, if not become a Peloton.