Cohort or self-paced: choose by the constraint that binds.

You are not choosing a format. You are choosing what you will pay with.

Cohort or self-paced is usually argued as a quality question — which one teaches better, which one people finish. It is really a resource question, and the two models bill you in different currencies.

A cohort is paid for in calendar. You commit hours you cannot move, for the length of every run, and in exchange you get a deadline you did not have to manufacture, a room where people can see each other, and a price that can include access to you. A self-paced course is paid for up front, in build time and in design effort, and in exchange it costs you almost nothing per additional student — and supplies none of the pressure the calendar was supplying.

That is the whole trade, and almost every other difference follows from it. Which is why the comparison usually goes wrong at the first step: people start from the format they would prefer to teach and reason backwards, rather than starting from the hours, the audience and the price they actually have.

A word about completion rates

You will find confident figures claiming one model is completed far more often than the other. Treat them the way you would treat any average taken across price points, subjects, audiences and free-versus-paid enrolment: as a description of populations rather than of formats. A cohort priced at real money, sold to people who chose a start date, is being compared with free video courses nobody paid for. The mechanism is the part worth reasoning about — one model hands the student a deadline set by someone else, and the other asks the student to bring their own. Everything downstream of that is design work you control, and completion systems is where it gets done.

Five differences that are real. The rest are preferences.

Read the table as a bill rather than as a scoreboard. Each row costs you something in both columns; the question is which cost you can absorb every month rather than which column has more ticks in it.

Two rows do most of the deciding. Calendar cost is the one people underestimate, because the first run is interesting and the fourth is a Tuesday evening you would rather have back. Launch cadence is the one people forget entirely, and it is the reason a business built on cohorts feels like a series of sprints while a business built on self-paced courses feels like a slow leak that nobody notices until the quarter ends.

A cohort is the right answer when the teaching is not finished.

The most under-stated advantage of running live is that you do not have to know the curriculum yet. You find out what the curriculum is by watching where a room of real people gets stuck, and you get paid while finding out.

That is also why a cohort suits a subject that moves. If a third of what you teach would be wrong in eighteen months, recording it thoroughly is an expensive way to build something you will have to rebuild. Teaching it live keeps the material current at no extra cost, because updating it is just saying different words next Tuesday.

The third case is work that needs looking at. Some outcomes cannot be assessed by a quiz — a plan, a design, a piece of writing, a business decision — and the value of the programme is your judgement applied to that person’s specific attempt. A video cannot do that. Neither can an automated assessment, though a rubric plus peer review gets closer than most people expect.

And who a cohort quietly fails

The shift worker. The parent with an unpredictable evening. The person eleven timezones away who would have to be awake at three in the morning every week for eight weeks. The person who needs this now and cannot wait until March. None of them will tell you they could not attend; they will simply not buy, and you will read that as weak demand for the subject rather than as a scheduling exclusion.

It also fails you, specifically, if the honest answer to “can I hold these hours for eight straight weeks” is no. A cohort where the host keeps rescheduling is worse than no cohort at all, because the deadline was the product and you have now taught the room that it is negotiable.

Self-paced is the right answer when the teaching has settled.

Once you know the shape of the curriculum — which lesson has to come before which, where people get confused, what the first week should feel like — recording it stops being a gamble and starts being leverage.

Its real advantage is not scale, which most people never need. It is availability. Someone decides they want this on a Sunday night in a timezone you have never sold into, and they can start immediately. A cohort makes that person wait until the next intake, and a surprising share of buying intent does not survive the wait.

The second advantage is that it does not compete with your other work. Self-paced revenue arrives without a calendar negotiation, which matters enormously if the course is not the whole business — if you also consult, run webinars that sell it, or have a job.

What you have to build to replace the calendar

A self-paced course inherits none of the pressure a cohort supplies for free, so the pressure has to be designed in. That means a first-week onboarding sequence that gets someone to a small win before their enthusiasm decays, lesson lengths that fit the time people actually have, a visible sense of position, and a defined response when somebody stops. It is more design work than a cohort, not less — it is simply front- loaded, and it happens once instead of every run.

And who self-paced quietly fails

The person who has already bought three courses and finished none of them, and who knows it. The person whose work needs review rather than instruction. The person who would have thrived in a room and will instead buy a fourth course they do not finish. If those people are your market, the self-paced version is cheaper to run and worse at the job.

Most working courses are hybrids. Say which one you are running.

A self-paced core with some live contact is the arrangement most established programmes settle into. The four designs below differ in one thing only — how much calendar they cost you — and that is the number to choose on.

1. 01Self-paced core with a standing open hour. The material is finished and evergreen; one recurring session is the only thing on your calendar, and it costs the same hour whether four people enrol this month or four hundred. The cheapest hybrid to run and the easiest to sustain, which is why it is the one most often abandoned too early — attendance in the first weeks is always thin. Office-hours structure covers the scheduling models, the question queue and the contingency for the week nobody asks anything.
2. 02Self-paced core with dated checkpoints. Students still move at their own speed, but everyone starts on an intake date and three scheduled sessions sit at the points where people historically stall. It restores the deadline without committing you to a weekly hour for eight weeks. The design work is choosing where the checkpoints go, which is a question your feedback log answers better than your intuition does, and group call structure covers how to run the sessions themselves.
3. 03Cohort first, self-paced afterwards. Run it live to a small group, record everything, and let the edited recordings become the self-paced product. This is the sequence most first courses should use: the live run prices the unfinished version correctly, the questions in the room tell you what the curriculum is actually missing, and you get paid while building it. The trap is assuming the recordings are the course — recording course videos covers what has to be cut before a live session becomes something a stranger can follow.
4. 04One curriculum, two prices. Everyone gets the same material; the higher tier adds live access, review of their work, or both. It sells to two budgets without building two products, and it lets you cap the live tier at a number you can actually serve. Decide the cap before you open it rather than after, and see pricing an online course for how far apart the tiers should sit — and subscription versus one-time if the live tier is really a membership wearing a course’s clothes.

Three questions, in this order.

Work them in sequence. The first one eliminates a model outright for most people, and it is the one most often answered optimistically.

1. 01What can you defend on a calendar for a whole run? Not what you could do in an exciting week — what you can hold when something goes wrong at work and somebody in the house is ill. If the honest answer is “nothing weekly, reliably”, you are not choosing a cohort, and the rest of the decision is about which hybrid you can sustain.
2. 02How many people can you actually reach? A cohort needs enough people to fill one room on one date. A self-paced course needs enough continuous reach to keep selling something that is always available. Small reach plus a real price fits a cohort. Broad reach plus a lower price fits self-paced. Small reach plus a low price fits neither yet, and the next move is demand, not delivery.
3. 03What price does the outcome carry today? Divide the price by the hours a run would cost you. If the result is a number you would refuse for consulting work, the cohort is a job you are underpaying yourself for — either raise the price, cut the live contact to checkpoints, or sell the material and stop trading hours.

Notice what is not in the list: which format you enjoy, which one the internet is currently in favour of, and which one your competitor uses. Those are worth something at the margin and nothing at all if the first three answers point elsewhere.

The decision is reversible. The people who already bought are not.

Switching model is normal and usually a sign the business is working. What makes it go badly is almost never the mechanics — it is failing to say anything to the people who bought the previous version.

Cohort → self-paced

1. 01Cut everything that was addressed to one person. A live run is full of answers to a specific student’s specific situation, and those passages are dead weight to a stranger who does not know who Priya is or what she was building. Either edit them out or reframe them as worked examples with the context restored.
2. 02Replace the deadline you just removed. The cohort supplied a date, a room and a mild social obligation. Self-paced supplies none of them, so the sequencing, the first-week experience and the nudges have to do that job deliberately rather than by accident.
3. 03Move the answers out of the calls and into the course. The questions asked in the live run are the curriculum gaps you already found. Fixing the lessons is worth more than publishing a recording of the call where you answered them, because nobody watches nine hours of somebody else’s office hours.
4. 04Decide what the live buyers keep. People who paid a cohort price partly bought access to you. If the self-paced version is much cheaper, say what the difference bought and honour it — continued access, a session, the next run at no charge. Silence here is what turns a price change into a complaint.

Self-paced → cohort

1. 01Check the calendar honestly, twice. The first run is exciting and the fourth is a Tuesday. Look at what the commitment costs in a normal month rather than in a launch month, and assume at least one week where you would rather be doing anything else.
2. 02Cap the room before you sell it. A cohort is priced on attention, so the number you can serve is a design decision, not a result. Set it, say it publicly, and let it create the scarcity rather than inventing scarcity afterwards.
3. 03Give existing self-paced buyers a real answer. They bought the material; you are now selling the material plus you, at a higher price. The clean move is a defined upgrade path priced at the difference, offered to them first. Grandfathering costs less than the goodwill it protects.
4. 04Change one variable per run. Model, price and promise moving together produce a result you cannot read. If the cohort underperforms you will not know whether the format, the number or the pitch was wrong, and the next decision inherits the confusion.

One rule covers both directions: change one variable per run. If the model, the price and the promise all move at once, the result is unreadable — you will not know which change produced it, and the next decision will inherit the confusion. Move the model, run it, read what happened, then move the price.