The Rise of Cohort-Based Courses: A Skeptical Look
In early 2021, a product manager I’ll call Sonia paid two thousand dollars to take a five-week online course on executive communication. She was not buying a certificate. She had a perfectly good MBA. She was buying the Tuesday night Zoom with Wes Kao, the founder herself, and the Slack channel where thirty other operators at companies like Stripe and Airbnb would trade feedback on her memos. Two years later, when she ran a product review at her new company, her VP told her the writing was sharper than anyone else on the team. She credited the course. I asked her if she remembered the content. She said not really, but that she remembered how it felt to be held to the standard of the group.
Cohort-based courses, or CBCs in the shorthand that the genre briefly took on, were supposed to be the post-pandemic answer to the MOOC. The diagnosis was straightforward and mostly correct. Self-paced video courses on Coursera and Udemy had miserable completion rates, somewhere between five and fifteen percent depending on which Katy Jordan meta-analysis you trusted. People bought them with enthusiasm and finished them almost never. The problem was not content quality. The problem was that humans, it turns out, mostly need other humans to do hard things.
Maven, founded in 2021 by Gagan Biyani, was the most visible infrastructure bet on this thesis. Biyani was a Udemy cofounder who had watched the self-paced market mature and flatten; he had seen what happened to completion rates and to instructor compensation, and he had concluded that the economics of online education ran through community, deadlines, and teacher access rather than through video production. Maven raised a 25 million dollar Series A from Andreessen Horowitz and First Round, signed up Wes Kao and Li Jin and Tom Chi and several dozen other operator-instructors, and built a platform that let experts run premium live courses without assembling their own software stack. By 2023 Maven had hosted thousands of courses and moved meaningful revenue to individual teachers, some of whom were earning more from their cohorts than from their day jobs.
The thesis held up, at least on the margins it was meant to. Completion rates on cohort-based courses run roughly ten times higher than on self-paced equivalents; a thirty-percent completion rate is normal, a sixty-percent rate is not rare for small, intense cohorts. The presence of a deadline, of a teacher who might notice you are not doing the work, and of peers whose respect you want to keep, produces a pattern of behavior that is just different from the pattern of behavior a Coursera dashboard produces. Researchers who study what happened to MOOCs after the pandemic have documented the same dynamics from the other side. Scale kills accountability. Accountability requires smallness.
A cohort also changes the curriculum. Because the teacher is in the Zoom, they can rewrite Tuesday’s session in response to Monday’s questions. They can point at a real memo a real student shared in Slack. They can call on the quiet person. These are not revolutionary pedagogical moves, but they are moves that a pre-recorded video cannot make, and their cumulative effect on learning is significant, particularly for material that requires feedback on work you produce rather than memorization of concepts.
Now the uncomfortable part. Many cohort-based courses, including some of the most successful ones, are not really selling education. They are selling access, credential, and network, and they are priced accordingly. A two thousand dollar five-week course does not, on a cost-per-hour basis, compete with a university course or an in-person workshop. It competes with coaching, with consulting, and with the gray market of professional development budgets at tech companies. What you are buying when you pay for a Wes Kao cohort is a conversation with Wes Kao and thirty of your future peers, on a topic that happens to be useful at your job. The skills you learn are real, and also partly incidental to why the transaction made sense.
This is not a scandal. Executive education has always been partly signaling, partly networking, and partly content. The Harvard Business School general management program costs around 90,000 dollars for seven weeks and no one pretends that participants could not, in principle, read the same cases at home for the price of a library card. What makes a cohort-based course interesting is not that it escapes this logic but that it compresses it. You get a version of the experience at a fraction of the cost, for a topic narrow enough to matter to your next quarter, from a teacher who probably still works in the field.
Where the format is weaker is where the signal is too weak to justify the price. A 1,500 dollar course on “better note-taking” or “productivity systems” is a bad trade for most buyers, not because the content is bad but because the network you join is less valuable than the dollars you spent. The people who buy into those cohorts tend to be people who buy into every such cohort, and the resulting room is less useful than the mix of strangers you would meet in a free Slack community around the same topic. Comparing approaches to knowledge work, like the different note-taking systems serious writers actually use, is mostly a matter of reading, experimenting, and adjusting. It is not a matter of paying for a cohort.
The cohort format also filters for learners with specific traits. It rewards people who are already confident enough to speak on video, already connected enough to recognize the instructor’s status, and already well-off enough to expense two thousand dollars without flinching. This is a demographic that skews toward mid-career operators at well-funded tech companies, which is also the demographic that the founders of most successful cohort-based courses implicitly design for. If you are a first-year teacher trying to learn Python, or a nursing student trying to handle your clinicals, the format is not designed with you in mind, and the pricing makes that explicit.
What remains interesting about the model, after the initial hype has settled, is what it suggests about where human teachers retain an edge. The things a live cohort does that a pre-recorded course cannot are mostly things that require judgment, presence, and the capacity to change plans. These are the same dimensions on which researchers distinguish useful tutoring from useless tutoring, and on which thoughtful students distinguish using AI as a study partner from using it to outsource their thinking. The lesson of the CBC era, such as it is, is that software can deliver information at extraordinary scale and still not deliver the experience of being held to a standard by another person.
The genre will not grow infinitely. Many cohorts that launched in 2022 and 2023 have not relaunched. The founders who were attracted to Maven by the promise of escape-velocity teacher income have found that running a cohort is exhausting in a way that recording a video course is not. The students who were attracted by the novelty have become choosier. What will persist is a small, durable layer of high-trust, high-price cohorts run by teachers with real reputations, aimed at professionals whose employers will pay for them, on topics where the feedback loop genuinely matters.
This is a smaller business than the one Maven’s pitch deck imagined in 2021. It is also, quite possibly, the shape of the thing that should exist.
Photo via Unsplash.