Commoditize Your Complements

This is a cornerstone of our strategy and the strategies of many other companies.

One of my favorite pieces on business strategy is this famous blog post from Joel Spolsky. It’s main point is simple: Commoditize your complements. Allow me to pull out my economics bachelor’s degree and explain.

A commodity is a good that is undifferentiated. Think of crude oil. Regardless of whether that oil comes from Venezuela or Canada or Qatar, after some processing, it all ends up the same stuff we pump into our cars. The opposite is something like wine, which can be highly differentiated. In any wine shop, the bottles at the top of the shelf will cost multiples more than the bottles at the bottom. A wine producer from Bordeaux can charge a higher price because once people see “Bordeaux” on the label, they assume it is better and are willing to pay more for it.

But with commodities like crude oil, producers can only compete using one aspect: Price. If the Canadians can produce gas more cheaply than the Qataris, then they can sell it at a lower price, and sell a lot more of it. Producers of commodities are in a downward race to the lowest prices (which is why they form cartels like OPEC to cooperate and keep prices higher).

A complement is something that is purchased alongside something else — like hot dogs and hot dog buns. A more formal definition is that two goods are complements if an increase in the price of one leads to a decrease in the demand for the other and vice-versa. For example, tennis balls and tennis rackets. If tennis balls get more expensive, people will be less interested in tennis, so they will purchase fewer rackets.

Thus, you want the complements of whatever you are selling to be as cheap as possible. That way there is maximum demand for your product.

What is the lowest possible price? Free. This is why the big tech companies invest so much in open-source projects and providing other free things to the world. They are commoditizing their complements. Meta, for instance, is investing billions into open-source LLMs. Why? Meta makes money when people post content to Facebook or Instagram that other people want to watch. Anything that helps people create content, like LLMs, is a complement to Meta’s core business. In contrast, OpenAI makes money by charging people to use LLMs. If the models were free commodities, OpenAI wouldn’t make any money, which is how we end up with the ironic situation that OpenAI is the leader in closed-source models.

At Arno, we monetize primarily through our AI-driven feedback and scoring that shows students exactly how to improve their English. Our main complement is content — videos and articles — that teach people strategies for the Duolingo English Test and general English concepts like vocabulary and grammar. Content like this is a complement to our core business because people can consume a lot of it, but they need Arno’s scoring and feedback to know if they’re actually improving.

There is a ton of free content out there related to English as a second language. Even for DET prep specifically, there’s more content than any one person could reasonably consume. However, many of our competitors sell access to “premium” content. We tried this for a while and found it confusing: Which content do we make free and which content do we charge for? We never had a clear answer to this.

Increasingly this “premium” content model seemed like a weak way to monetize. We came to see that our competitors did it because they didn’t have a choice. They couldn’t build software like we could. Most importantly, they didn’t have a rockstar technical co-founder like Makoto who could build and maintain a globally-reliable, feature-rich, and intuitively-designed platform 😁

Once we realized that our core value prop and differentiator was software-driven features, this has freed us up to more aggressively commoditize our complements. Now our goal is to make the best DET preparation content free. This is going to seriously undercut our competitors, make a lot of value accessible to people around the world, and increase demand for Arno.

That’s the neat, easy-to-remember story. But I would be remiss if I didn’t get into some of the nuances that complicate matters. Content like YouTube videos is a complement to our core software-driven features, but it is also a substitute for those same features. It depends on the user. A lot of students can simply watch YouTube videos to learn what they need in order to get their target scores on the test. They don’t need to pay for Arno.

However, because there is so much content out there, people struggle to find what will be most helpful to them. The more content that people need to search through, the more value there is in personalization. This is the next phase of Arno, the first pieces of which we’ll be rolling out over the next 2 months 😁