Thinking Like A Venture Capitalist #2

Alright, we are back. Last time I wrote about the gap in my knowledge, which was essentially that I had all this experience building things and selling things but absolutely zero framework for understanding how a company is actually valued from the outside looking in. I introduced the lens I was trying to develop, and I used CookUnity as my first case study to test it. The big takeaway was that CookUnity is not a food company. It is a creator marketplace that removes the massive barrier to entry for independent chefs, the same infrastructure strategy you see at YouTube and Shopify. I left off saying I wanted to look at something closer to the software world next.

That takeaway from CookUnity honestly stuck with me for weeks. The idea that a company can be worth so much more than what it literally does on the surface was a genuinely new mental model for me. It made me want to find another company I already interact with every single day and see if I could spot the same thing underneath the hood.

I did not have to look very hard. The company I chose this time is Duolingo. I have had it downloaded on my phone for probably two years at this point. I have used it for a grand total of about four days. I do not think I have ever actually learned anything from it. But the app has sent me more notifications than any other app on my phone and I have never fully deleted it. I thought that was funny until I realized it was actually the whole point.

The Company I Was Completely Wrong About

Before I did any research, I assumed Duolingo was just a nonprofit passion project that somebody made to teach Spanish to people who could not afford textbooks. I genuinely thought the business model was mostly donation-based or grant-funded, like a Wikipedia for language learning. It looked too simple and too free to be a real commercial enterprise worth billions of dollars. I was thinking about it like a builder, not like an investor.

Then I looked it up. Duolingo went public in 2021 at a valuation of around 5 billion dollars. Their 2024 annual revenue crossed 700 million dollars. The majority of their users pay absolutely nothing. The free product is not the gimmick. The free product is the entire growth engine. When I read that, I had the exact same moment I had with CookUnity, where my original assumption about what a company actually is got completely flipped.

The Streak is Not a Feature

The most important thing I learned about Duolingo is that the streak mechanic is not designed to help you learn a language. It is designed to make you afraid to stop opening the app. There is a real field of study around this called variable reward systems, and casinos have been using the same psychological principles for decades. When you build a 30-day streak on Duolingo, you have created a sunk cost in your own head. Losing that streak feels like losing something real, even though it is a number on a screen. Duolingo engineered that feeling deliberately, and it is the single biggest reason their daily active user numbers look the way they do.

The subscription tier called Super Duolingo does not give you better language instruction. It removes friction. It takes away the ads. It removes the hearts system that penalizes you for wrong answers. It sells you a way to protect your streak without consequences. People do not pay for it because they are more serious about French. They pay for it because they are emotionally attached to a counter on their phone. That is one of the most clever pieces of product design I have ever come across, and it works because the free version is genuinely fun enough that people get attached before they ever see a paywall.

The Platform Angle Again

The thing that surprised me most, and directly connected back to what I learned from CookUnity, was how Duolingo actually builds its course content. I assumed a company that teaches 40 different languages must have massive internal teams of linguists writing every single exercise. That is not how it works at all. Duolingo has a contributor model where language experts, academics and native speakers outside the company design the core curriculum. Duolingo provides the gamification framework, the infrastructure and the hundreds of millions of users. The contributors build on top of it.

This is the exact same dynamic I described with CookUnity. The platform provides the rails. The creators and experts build the actual content. The difference in Duolingo’s case is that the content contributors are not compensated per lesson the way a CookUnity chef gets paid per meal sold. That makes the economics of acquiring course content dramatically cheaper for Duolingo than you would ever guess just from using the app. When I put that together, I started understanding why this model is so attractive to investors. The distribution costs stay high but the content acquisition costs stay relatively low, and that gap is where a lot of the margin comes from.

The Data Nobody Talks About

There is also an asset inside Duolingo that I almost completely missed because it is invisible to a normal user. Every single mistake I make on a sentence answer is a data point. Duolingo collects data on exactly which grammatical concept confused me, at what point in my learning progression, and how it compares to millions of other users at the same stage. They are running the largest real-world study on human language acquisition that has ever existed, and most people using the app have no idea that is happening.

From a VC standpoint, that dataset is a competitive moat that has nothing to do with green owls. If Duolingo wanted to license curriculum data to school systems, to AI companies training language models, or to governments building literacy programs, they would have an incredibly powerful asset to sell. The data is not just a byproduct of the product. The data might actually be more valuable than the product itself in the long run. I do not think I would have noticed that a few months ago. But after starting to look at companies through the investor lens, I could not unsee it.

Moving Forward

Looking at both CookUnity and Duolingo back to back was genuinely useful because they ended up teaching me the same core lesson from two completely different angles. CookUnity showed me the power of removing a barrier to entry for creators so that the platform benefits from their output. Duolingo showed me the power of engineering habit so deeply that users feel a psychological cost when they even think about leaving. Both of those things answer the same fundamental investor question, which is why will users not leave. I am starting to think that question is the most important one you can ask about any consumer business.

I am also noticing that the products I thought I understood the most are the ones that surprised me the most when I actually looked at them analytically. I had opinions about both of these companies before I ever did any research. Both of those opinions were basically completely wrong. I think that is going to be the pattern here. The whole point of this series is to unlearn the surface-level takes I have carried around and replace them with something more honest.

I am two articles in and maybe a third of the way to having a real framework. September is getting closer and I still have a lot of ground to cover before I walk onto the NYU campus. Next time I want to move away from consumer apps entirely and look at something in the B2B software world, because every company I have analyzed so far has individual users as the customer, and I suspect the dynamics look very different when the customer is another business. We will see.

Keep building, but start asking why your users would never uninstall.