Measuring Incubator Impact Without Fooling Yourself
The attribution problem nobody solves: did the programme cause the outcome, or select companies that would have got there anyway?
The claim everyone makes and nobody can support
Read any incubator's annual report and you will find some version of: our companies raised ₹X crore and created Y jobs.
Both numbers may be entirely accurate. Neither tells you whether the incubator did anything, because both are compatible with a programme that admitted companies which were going to raise and hire anyway.
This is the attribution problem, and no incubator solves it properly. What separates honest reporting from the rest is not solving it - it is knowing it exists and narrowing the gap where you can.
The selection effect, from the inside
An incubator that is good at selection will report strong outcomes even if the programme itself adds nothing.
That is not cynicism, it is arithmetic. If you admit the eight strongest applicants out of ninety, those eight would have outperformed the other eighty-two regardless. Your outcome numbers reflect your judgement, which is real and valuable - but it is a different thing from your programme working.
The uncomfortable part is that it is invisible from inside. Strong outcomes feel like validation whichever mechanism produced them.
Four things that narrow the gap
None is proof. All are much better than the standard claim.
1. Record a baseline at entry
The minimum viable step, and the one most centres skip because it feels administrative at the moment a cohort is starting.
Stage, revenue or usage, headcount, funding raised, and what the founder says the constraint is - captured on day one, in twenty minutes per company.
Without it, every later claim is about absolute state. With it, you can talk about movement, and movement is much closer to contribution.
2. Measure movement, not state
"Eight companies at first revenue at graduation" tells you very little. Some may have arrived there.
"Six of eight moved up at least one stage; two did not" is a statement about the period they spent with you. Still not causal, but it is about change, and change is what a programme could plausibly cause.
3. Compare against your near-misses
The strongest practical method available to most incubators, and almost nobody uses it.
You rank applicants. You admit the top eight. Companies ranked ninth to sixteenth were judged similar in quality and did not get the programme. They are the closest thing to a control group you will ever have.
Follow up with them - politely, once a year, with a short form. Many will answer, particularly if you were decent to them at rejection. If admitted companies consistently outperform near-misses, that is meaningful evidence. If they do not, that is more useful still, and it is a finding about your programme rather than about your applicants.
4. Ask founders what actually helped
Qualitative, unfashionable, and informative. At graduation and again a year later, ask an open question: what, if anything, from the programme changed what you did?
The answers cluster in ways that are frequently unflattering to the programme's centrepiece. Many incubators discover that the thing founders name is the desk, or one specific introduction, rather than the curriculum that took most of the staff effort.
That is worth knowing. It is also only obtainable by asking.
Label claims accurately
The difference between honest and dishonest reporting is usually one word.
| Instead of | Say |
|---|---|
| We helped companies raise ₹X | Companies in our portfolio raised ₹X |
| We created Y jobs | Companies we supported employ Y people |
| Our programme has an 80% success rate | 80% of companies that entered are still trading at 24 months |
| We accelerated their growth | Six of eight advanced at least one stage during the programme |
The right-hand column is not weaker. It is checkable, and a reader who can check one claim believes the next one.
Report against the entering cohort
Restating it because it is the most common way incubator numbers mislead unintentionally.
Survival measured against graduates excludes everyone who dropped out, which is exactly the population that did worst. Twelve of fifteen graduates still trading is "80%"; against the twenty who entered it is 60%.
Pick the entering cohort, say so, and keep it consistent. A denominator that changes between years makes every trend meaningless, including the good ones.
What to publish
A defensible annual impact section:
Activity - applications, admissions, companies supported by engagement level.
Movement - stage progression against baseline, for the entering cohort.
Outcomes - survival at 12 and 24 months, companies that raised anything with median and total, jobs with the definition attached.
Comparison - where you have it, admitted versus near-miss.
What founders said helped - two or three lines, honestly.
That is a stronger document than a large unattributed total, and it is the kind of thing a funder's evaluator has rarely seen from an incubator.
The honest position
You will not prove causation. Nobody in this sector does.
What you can do is record a baseline, measure movement, follow your near-misses, ask founders directly, and label your claims accurately. An incubator doing those five things knows more about its own effect than almost any of its peers - and it can say so without overclaiming, which is the part funders remember.
Frequently asked questions
- Can an incubator prove it caused an outcome?
- Not rigorously, without a control group nobody has. What you can do is narrow the gap: record a baseline at entry, measure movement rather than absolute state, and compare admitted companies with the near-misses you rejected. None of that is proof, and all of it is far better than the usual claim.
- What is the selection effect?
- The tendency for good outcomes to reflect who you admitted rather than what you did. An incubator that selects companies already likely to succeed reports strong numbers without having changed anything, and cannot tell the difference from the inside.
- How do you measure a programme's contribution?
- Movement against a recorded baseline is the best practical proxy - stage at entry versus stage at exit. Comparing your cohort with your waitlist is the strongest thing most incubators can do, because those companies were judged similar and did not get the programme.
- Should incubators report impact they cannot attribute?
- Report it, label it accurately. "Companies in our programme raised X" is a true statement about a portfolio. "We helped companies raise X" is a causal claim you probably cannot support. The first is honest and still useful.
