Why Ed-tech Startups Don't Scale

November 2022

I have yet to find a successful ed-tech model that works at scale.

Issues abound, but it often comes down to how ed-tech grows and how it monetizes its audience.

Let’s start with growth.

Ed-tech products are hard to grow

Ed-tech products are hard to grow, even when they’re free.

The education market is geographically siloed. As an early-stage startup that just found product-market fit in one region, you are unlikely to be able to scale it by replicating your model as venture-backed businesses usually do, because different regions have wildly different markets.

You’ll quickly find yourself in front of a conundrum: either you dilute your PMF by geographically expanding the product, or you forgo growth and hunker down in your initial market. The former will likely result in a lot of churn as you acquire users who are not the right fit for your product, while the latter means that your growth potential is massively capped.

This was the issue when I built Uniwhere, which ultimately made us sell our company. We found a perfect product-market fit in Italy that did not easily transfer to adjacent markets in Europe and North America. And we tried — hard.

See, when you build an education product, you’re competing against two massive institutions: the formal education system (schools, colleges, state universities, etc.), and the laws and cultural expectations around that system. The former is highly resistant to change, but can be fought with a sizable war chest. The latter cannot be changed in a startup timescale – it takes too long.

In the US, it’s culturally expected that college students take repeated internships every year. In Europe, it’s not. This means that you can build a successful internship product in the States that will fail miserably in most other countries (see, for instance, Handshake). In the US there are many ways to finance one’s college degree: from grants to loans to parents’ funds. In Brazil, although loans are starting to appear, the vast majority of college students finance their degrees out of pocket. In Italy, the price is low enough that students usually pay for it while working part-time as waiters. In Germany, it’s free.

Yes, product localization will always be difficult in any product category. But growth friction across countries must always be offset with expected returns. They’re inversely correlated.

Some startup industries are just better and more investable than others — products like Netflix, Airbnb, or Doordash face consumer habits that cross borders and cultures, which is why they can scale effortlessly. The product experience you will get from Disney Plus in Ecuador will be precisely the same as the experience you’ll get in Mongolia. Education is not like that. Inter-market differences are profound, almost unshakable, and extremely granular.

Some industries are hard to scale across borders, but it’s worth it, because the expected returns are massive. The issue with education is that expected returns are just too small.

Why, though? Why do ed-tech companies always end up with abysmal market caps? It definitely has to do with the difficulty in scaling the company, but it’s not the only thing.

Monetization at scale is an unsolved problem

Monetizing an ed-tech product seems immediate on the surface, but gets stupidly hard when you get deep into the woods.

For most ed-tech products, the financial incentives are misaligned: parents pay, students consume. You need a product that entices parents, but optimizes for student retention. Sometimes it’s even worse: institutions pay for the product, educators administer the product, students consume the product, and parents passively evaluate the ex-post performance of the product. Many stakeholders, with potentially competing needs, and you must satisfy them all. It’s an intrinsically hard design problem.

What ends up happening is a race to the bottom on either price or quality, which means that most K-12 ed-tech products either suck or don’t sell.

To solve this issue, you can target users that can pay — as in, adults. You then get caught in another problem: the time horizon is misaligned. You need high effort to learn now, but benefits come much later. I’ve already written in the past about how learning, to be effective, must be hard. Customers must actively engage with education products in an effortful way, which limits the pool of prospects in a severely distracted world.

Startups found novel ways to get around these intrinsic limitations of how we structurally (and biologically) learn. All of them have shortcomings. None checks all the boxes that make ed-tech companies scale massive businesses.

The first way is to bundle users together in small groups, the so-called cohort-based courses. CBCs do manage to provide a rather successful (and often enjoyable) learning experience. The issue is that, by very design, they’re extremely hard to scale while maintaining high product quality.

The second way is to gamify the learning experience. A classic example is Duolingo. The issue with that approach is that it’s mostly successful in acquiring new users, but churn is off the chart essentially because you’re not actually learning anything, and people eventually figure it out.

Other companies solved the time-horizon issue by postponing the cash-out event — like Bloom Tech (fka Lambda School). They popularized Income Sharing Agreements — a financial scheme where the education provider gets paid with a fraction of the future income of the student. It was a disruptive, incredible feat. But they still suffered from the same scaling issue as any other cohort-based course. The more you try to scale them, the more their effectiveness dilutes.

Many companies (my old one included!) think that they can solve the monetization issue by selling recruiting ads. If you can’t sell content — the thinking goes — you can sell profiles to paying companies that can’t find fitting profiles for their open roles. Recruiting and education are two sides of the same coin. After all, the more you learn, the more your profile will become sellable on the hiring market.

Unfortunately, it doesn’t work.

There are many reasons, very similar to the ones plaguing education models. One of them, for instance, is that the customers that need the product the most — small companies and local businesses — are those with the least amount of money to pay for it. As with education, a recruiting business will get worse and worse the more you scale it. And there’s an extreme variance between small and big companies recruiting needs, specular (and maybe partially attributable) to the regional fragmentation of the education market.

One potential angle of attack is to make companies pay for learning content, for employee reskilling and retention. I won’t say it can’t work, but I’m bearish. Large companies can leverage a higher base comp and better employer branding to attract and retain employees — which is much more effective than pure reskilling. Typically, when large companies invest in reskilling content, they’d rather make their own materials and tools when they can afford to do so; if we consider only those who can’t, the market shrinks considerably.

The only, effective way to easily monetize ed-tech products at scale today is to avoid the “education” part in the first place, and sell entertainment instead. Masterclass is the perfect example. Substacks, Morning Brew, and even magazines like The Economist fall into this category. Edutainment is pleasant and immediately satisfying. It works, as a business. But it doesn’t teach anything.

It’s not ed-tech.


In the past, some ed-tech companies did manage to grow to become robust global businesses. I’m thinking of products like Udemy, Lynda, Coursera, maybe Udacity. They leveraged product primitives that quickly saturated, and today are no longer exploitable. More interestingly, no one that I can think of ended up actually becoming a generational company.

Take Udemy, for example: it’s a peer-to-peer marketplace. From a product design perspective, it’s no different from eBay, or Uber. It’s the startup class of the early 2000s. And, most critically, only a $2 billion market cap. Hardly generational.

Interestingly, Coursera and Duolingo too have a $2 billion market cap, as if there’s an informal consensus on the ceiling the market expects from these kinds of businesses.

The thing is – it’s very hard to build something new, with a high product-market fit, easily monetizable, that scales so much that it becomes a generational company.

You’ll either find companies with high PMF that never scale beyond their initial successes (eg. Uniwhere), or companies that can scale but can’t achieve high PMF because regional players will always build better products — and better products on a smaller scale will inevitably monetize better.

In a nutshell, you have to deal with a balkanization of ed-tech products. It’s too capital-intensive to compete when the upside is so low and the likelihood of global success so minuscule.

I don't want to say that building a venture-scale, generational company in the education space can’t be done. Maybe someone will figure it out.

It's just so hard that no one has yet.


Want more?

I post irregular, bite-sized updates on Twitter. If you liked this post, feel free to tag along or reach out.
You can also find more essays here.


I’m publishing a private newsletter on what I’m thinking and what I’m building. Join 1,200+ readers here:

Thanks to Mathurah Ravigulan, Aman Mathur, Giulio Michelon, Nick Drage, John Lanza, Christ Angelis, Daniel Sisson, Micheal Shafer, and Jeremy Nguyen for reviewing earlier drafts of this post.
© Gianluca Segato 2022