Most founders we meet can describe their ICP in two paragraphs. The problem is that the two paragraphs cover roughly four companies, three industries, and a price range that spans 8x.
That isn't an ICP. That's a coverage map of the customers who happened to say yes early.
A fuzzy ICP doesn't fail loudly. It fails quietly, in three places at once:
You don't get out of a fuzzy ICP with another off-site. You get out of it with one piece of evidence: which 12 customers closed in under 30 days, paid full price, and stayed past month four?
Find them. Look at what they have in common. That's the ICP. Everyone else is a distraction you'll eventually have to fire.
You'll have your real ICP by Friday. The hard part is having the conversation with the founder who wrote the original two paragraphs.
Watch for three quiet symptoms: paid channels that consistently underperform, a volatile sales win rate where every deal feels like fresh discovery, and a product roadmap that fights itself because customers want contradictory things.
Pull every closed-won deal from the last 12 months and tag the ones that closed in under 30 days, paid full price and stayed past month four. What those customers have in common is your real ICP.
Not to start. Behavioural evidence from your own closed-won and churn data is more reliable than what people say in interviews. Use interviews to explain the pattern you find, not to look for it.