How Long Should an Incubation Programme Be?

Length is a design decision, not a convention

Most programmes pick a length by copying one they admire. Twelve weeks because that is what accelerators do, or open-ended because the building is there and nobody set an end date.

Both can work. Both fail in specific, predictable ways when applied to the wrong companies.

The question worth answering first is not "how long" but: what has to be true at the end, and how long does that actually take?

Twelve weeks: intensity, and a hard stop

Works when companies arrive with a product, some evidence it works, and a specific goal - raising, a first enterprise customer, a launch.

Twelve weeks is enough to sharpen, focus and prepare. It is not enough to build. A pre-product company in a twelve-week programme spends the first month working out what to build and the last month preparing a demo day pitch about something that barely exists.

Requires: a real end event that participants can act towards, high-intensity contact, and companies selected for readiness rather than promise.

Fails when: applied to early-stage companies because the format is fashionable, or when the end event has no audience who can act.

Six to twelve months: the common incubation shape

The default for centre-based incubation in India, and it fits what most incubators actually have: space, a mentor network, some grant funding, and companies that are early.

Long enough to build something and get initial evidence. Short enough to create pressure. It also fits a grant cycle and an academic year, which matters more practically than it should.

Requires: genuine review points. Six months with no checkpoint until month six is open-ended incubation with an end date attached. A quarterly review that can change something - funding release, continued space, exit - is what makes the length work.

Fails when: the middle months have no structure. Cohorts reliably disengage around month three or four if nothing is due.

Open-ended residency: the landlord risk

Companies stay as long as they are useful to the centre and the centre is useful to them.

This genuinely works for deep-tech and research-linked ventures where three years is a realistic timeline and the value is lab access rather than programme content.

Requires: hard review points and a real exit mechanism. Without them, the centre becomes a subsidised office with an occasional event, space never turns over, and the newest companies cannot get in because the oldest never left.

Fails when: review is nominal. If nobody has ever been asked to leave, review is not happening.

Structure the middle, not just the ends

Whatever the length, the failure is nearly always the same: a strong start, a strong finish, and a middle where nothing is due.

The fix is a small number of points where something must be ready and something can change:

Point	What is due	What can change
Entry	Baseline recorded, agreement signed	Nothing - this is the start line
One third	First milestone, first update	Tranche release, mentor rematch
Halfway	Substantive review	Continued funding, continued space, exit
Two thirds	Second milestone	Final tranche, demo day inclusion
End	Final report, outcome recorded	Graduate, extend, or move to alumni terms

The halfway review is the one that matters most and the one most often skipped. It is the only point at which a programme can honestly say to a company that this is not working, while there is still time for that to be useful to them.

Separate programme length from space tenure

A useful and under-used distinction.

The programme is structured support, and it should have a defined end. The space allocation is a commercial arrangement, and it can continue on different terms.

Separating them means graduation stops being an eviction. A company finishes the programme and moves to alumni space terms - higher rate, less support, still connected. The desk becomes reclaimable through a renewal conversation rather than a difficult one, and the relationship continues.

Centres that bundle the two have the hardest graduation conversations, because leaving the programme and losing the office are the same event.

Run one length well before running two

Multiple programme formats - an accelerator, an incubator, a pre-incubation track - look impressive and are usually premature.

Each format needs its own selection, its own rubric, its own schedule and its own reporting. Running two badly is worse than running one well, and the staffing article makes the point plainly: a second format needs a second programme manager, or it takes capacity from the first.

If you do run more than one, keep the machinery shared - one question bank, one set of stage definitions, one reporting cadence - so the data stays comparable across formats. Otherwise you end up with two programmes whose outcomes cannot be compared, which makes it impossible to tell which one is working.

The test

Ask two questions about your current length:

1. What has to be true at the end, specifically, for a company to have benefited?
2. How long does that actually take for the companies you admit?

If the answers do not match your programme length, the length is wrong. That is a more useful diagnosis than a disengaged cohort, which is only the symptom.