Nine questions before you build anything

Conclusion: whether a need deserves work is not decided by how plausible it sounds, but by whether you can answer these nine questions line by line. Any unanswerable one means you are not at the coding stage yet.

The original

Is the need purely imagined / is the need big enough / are my information sources correct / is my judgement real / can it be simplified so investors and ordinary people understand it / how big is the market / what do the cost and difficulty of the financial model look like / are my guesses about buyer behaviour verified / does the buyer segment need another cut — from my own notes, 2025 (redacted), filed as “CEO self-audit checklist”

Why these nine work

They are not nine parallel considerations. They are four filters from “me” to “market”:

1. Self-doubt layer (1, 4): did I invent this need? Is my judgement contaminated by my own taste?
2. Evidence layer (3, 8): who are my sources? Is buyer behaviour inferred or verified?
3. Scale layer (2, 6, 9): big enough? how to segment? what is left after segmenting?
4. Feasibility layer (5, 7): can it be explained simply? do the cost and difficulty add up?

Each layer filters a different failure mode: layer one filters the self-important, layer two the hand-wavers, layer three the linear thinkers, layer four the people who cannot do arithmetic.

Auditing this site with it

Applying the nine to the site I launched today, questions 3 and 8 are the uncomfortable ones:

• Who are my information sources? Readers. But I have no reader data yet, so the honest answer is “not verified”.
• Is buyer behaviour inferred or verified? Inferred. Which is why I wrote a 30-day experiment instead of a conclusion.

The rest hold up reasonably. The need (“let someone check whether a person has been doing a thing for a long time”) is not large, but it has the shape of a real need: as long as people must judge whether someone is worth working with, the question persists.

How to use it

Do not use it as a scorecard (nine “roughly fine” equals nine failures). Use it as a veto: anything you cannot answer means stop and go get evidence first.

That is why it lives on this site: it is the cheapest rule I have ever written.