Research is not expensive. Not doing it is — here is the arithmetic.

“Let's just build it and see” sounds fast. What it actually does is not postpone the bill but enlarge it — because if the idea is wrong, finding out six months from now costs many times more than finding out three days from now.

The bill for six unvalidated months

The arithmetic below assumes a single technical founder. The figures are illustrative — yours will differ — but the ratios do not.

Now the other side of the same ledger: what it costs to research that idea before construction begins. Finding out who your competitors are, what their users complain about, who in the same space is actually generating revenue, and whether this idea has been tried before — an afternoon, or one analysis for about the price of a cup of coffee.

“But I know my market”

You usually do — and that is exactly what makes it risky. The blind spot of a founder who knows their field well is not checking how much of what they know is still true. Whether the complaints on a review site are the same as they were three years ago, whether competitors have since filled that gap, whether a new player in the space has just raised money — these are learned by looking, not by intuition.

• “I have no competitors” usually means “I have not looked yet” — and if there genuinely are none, that is a finding too, and rarely good news.
• “The market is huge” tells you nothing about the share you can take; what matters is who the first hundred people paying you will be.
• “Nobody does it the way I do” — being different is not enough; the difference has to sit somewhere the customer cares about.
• “There's clearly demand” — the existence of demand does not mean that demand converts into payment.

The same arithmetic at institutional scale

For an accelerator, a technology transfer office or a grant programme, the equation inverts: the problem is not spending too much time on one idea but being unable to give enough time to hundreds. The first thirty applications and the last thirty are never read with equal attention — nobody does that deliberately, fatigue does it. The result is a score that varies by reviewer and a decision that is hard to defend.

Here the return on research is not saved time but CONSISTENCY: every application passing through the same criteria, every claim tied to its source, and the reasoning behind a rejection left in writing. The hardest moment for a programme is never explaining an acceptance — it is explaining a rejection.

How much research is enough?

Not unlimited. The goal is not to make the idea certain — that is not available. The goal is to find which assumption, if wrong, would change the idea fundamentally, and to test that one first. You live with the rest of the uncertainty; knowing which uncertainty you are living with beats pretending none of it exists.

That is why Premisey's output is not a “go / no-go” stamp: what is verified, what is an estimate and what is still unknown are shown separately. A number without a source is never labelled verified. You make the decision — but this time you know what it rests on.