Same name,
different companies.
A school with a building sells access to a place as much as to teaching. Students arrive, see each other, develop the mild social obligation that gets them through the months where learning stops feeling like progress, and hand over money partly for the room. A school without a building sells the teaching alone, to people who chose it from a list of alternatives and can leave it just as easily. Both can be excellent and profitable. Almost nothing transfers between them except the syllabus.
That is why the question cannot be answered by preference, and why "we will do both" is not the safe middle it appears to be. The useful way to decide is to look at four structural asymmetries and ask which set of problems you would rather have, because you will certainly have one set.
What actually differs.
The first two rows are usually the whole of the conversation, and they cut in opposite directions. Premises commit the largest cost in the business before you know whether anyone will come, which is a genuinely bad way to test an idea. Online removes that commitment entirely, which is why it is the right way to find out whether there is demand for what you teach — but the saving is not free money, because the marketing spend that replaces rent does not stop when the classes fill, and a school that budgeted rent-free without budgeting acquisition has simply moved its fixed cost to a line it did not plan for.
The third row is the one founders underestimate. A premises school competes with three or four schools in its town, and its competitive advantage can be as ordinary as being nearer the station. An online school competes with everyone teaching that language, including people charging very little and people who look enormously well-funded. Being merely good is a viable position locally; online it is not a position at all, and you will need a reason to be chosen that fits in a sentence.
Leaving online
costs the student nothing.
Language learning has a shape: an encouraging start, a long middle where progress becomes invisible, and — for those who get through it — competence. Most students who quit, quit in the middle, and they quit not because the teaching failed but because the returns stopped feeling immediate. What carries people through that stretch is rarely motivation. It is friction and company: the class they have already walked to, the people who will notice, the teacher who will ask where they were.
A building supplies that friction for free. Online, you have to build it deliberately, and a school that does not will lose students in the middle months and call it churn. The things that work are unglamorous: small stable groups rather than shuffling rosters, a teacher who contacts an absent student the same day, visible progress that is not just a score, and a reason to turn up that involves other people — which is what breakout rooms are for, and why a live class that is mostly a lecture converts to an empty class. The absence follow-up article covers the same-day habit, and it matters roughly twice as much online.
The corollary is a pricing one. Because leaving is frictionless, online schools that sell month to month invite a monthly decision to continue. Selling a course with a defined end — a term, a level, a set number of weeks — replaces twelve small decisions with one, and gives the student a finish line to aim at rather than an open-ended subscription to reconsider every payday.
Hybrid is a choice.
Usually it is a drift.
Almost every premises school now offers something online, and almost none of them decided to. A student asked to join remotely while travelling, it was easy to say yes, and within a year the school is running classes where some people are in the room and some are on a screen — which is the hardest format to teach well and the one nobody chose. The teacher is now managing two audiences with one attention, the people on the screen can hear less than they admit, and the school is paying for premises while delivering part of its value without them.
Hybrid is defensible when it is designed. Three versions work. Separate cohorts: the same course runs as an in-room class and as an online class, each taught as itself, never mixed. Premises core with online reach: the school is physical, and online exists only to serve students who could never attend — a different catchment, priced and marketed separately. Blended by design: some scheduled elements are online because that is genuinely better for them, and everyone is remote for those.
What fails is the fourth version, the one that arrives by accident: one class, some people present, decided lesson by lesson. If you are going to do it, write down which version you are doing, tell teachers, and price it.
Ask what would
have to be true.
Rather than weighing preferences, turn each option into a condition and test it cheaply. Premises make sense if there is local demand you can name, if the students you want are children or people who value a place, and if you can fund the lease through the months before the timetable fills. Online makes sense if the students you want are adults or specialists, if you have a reason to be chosen that survives comparison with strangers, and if you can spend on being found every month rather than once.
Then test before committing. Online is the cheaper experiment by a wide margin, which is the practical reason to start there even if you intend to end up with a building: run a term, see what enrols, find out what people actually want to learn and at what hours, and take that evidence to the lease negotiation instead of a hope. A school website on a free address, a timetable, live classes and a way to take payment are enough to find out, and on our free tier one educator with ten students costs nothing — which is a real first cohort, not a demo.