The test this page has to pass
Hand your ICP to someone who does not work with you and ask them to find five companies that match. If they can do it in twenty minutes without asking a question, your ICP is real. If they come back with a list you'd reject half of, it isn't.
That's the whole standard. Most ICPs fail it because they describe a mood — ambitious, fast-growing, customer-obsessed — rather than a set of facts you could filter a list by.
Nine prompts. One page. No adjectives.
- 1Who has the problem? Name the role and the seniority, not the department.
- 2Who feels the pain and who signs the cheque? If they are different people, write both.
- 3What has to be true about the company for the problem to exist at all? Headcount, stage, structure, volume, geography, regulation.
- 4What is the trigger? The event that turns this from a background annoyance into something someone acts on this quarter.
- 5What are they doing instead today? Name the tool, the spreadsheet, or the person.
- 6What does it cost them to keep doing that? In hours, money, or risk — a number, not an impression.
- 7Where do they already go for answers? Communities, conferences, publications, individual people.
- 8What disqualifies someone? The conditions under which you would tell a prospect they aren't a fit.
- 9Name five real companies that match every line above.
Why the trigger line matters most
Most founders can describe who has the problem. Very few can describe when the problem becomes urgent, and that gap is the difference between a pipeline and a mailing list.
Triggers are usually events: a funding round, a new head of something, a failed audit, a doubled headcount, a churned enterprise account, a regulatory deadline, a migration. They are observable from outside, which is what makes them targetable.
If you cannot name a trigger, you are selling an improvement rather than a resolution, and improvements get deferred indefinitely.
Anti-patterns to strike out
- —'Fast-growing teams' — describes a mood, filters nothing.
- —'SMBs and enterprise' — two different products with different buyers.
- —'Anyone who does X' — if the market is everyone, the message is nobody's.
- —Persona names and stock photos — decoration standing in for a definition.
- —Demographics with no behaviour attached — age and job title without a trigger.
- —An ICP with no disqualifiers — a filter that excludes nothing isn't a filter.
Where the evidence comes from
Every line on the page should be traceable to something someone said or did. Write the source in the margin: an interview, a support ticket, a lost deal, a usage log.
Lines with no source are hypotheses. Mark them as such and put them at the top of your next round of conversations. A one-pager where half the lines are honestly marked as assumptions is far more useful than one where they're all quietly presented as fact.
How often to redo it
Every time you close or lose three deals, and every time you finish a round of interviews. The ICP is not a founding document; it's a working hypothesis that should get narrower as evidence accumulates.
Narrower is the direction of travel. If your ICP is getting broader over time, that is usually a sign of pipeline anxiety rather than learning.