Many early-stage founders treat the ICP as a decision: a workshop, a persona slide, a paragraph in the deck. Then they go to market and wonder why the "ideal" customer keeps not buying. The mistake is upstream. An ICP is not something you choose. It is something the market reveals, if you force it to. The reason many founders never force it is simple: they refine on compliments. "Interesting." "Send me something." "Let's stay in touch."
These feel like signal. They are noise. Interest is free. The only evidence that validates a customer profile is behaviour that costs the buyer something: a named deadline, a budget owner in the room, money paid. The moment you rank evidence that way, ICP discovery stops being a guessing game and becomes a controlled experiment. Segments stop being debated in meetings and start being forced up an evidence ladder, or out.
The structure is not complicated:
1. ICP hypothesis - structured segment including demographics and triggers
2. Sales engine design - prospecting, landing surface per segment, outreach / inbound
3. Run the engine - fixed window, no extensions
4. Log the evidence - conversations that deliver proof, not interest
5. Face the gate - kill, narrow, pivot, or hold
6. Repeat, one segment at a time
The difficulty was never the structure. It is step 5. Every gate ends in one of exactly four decisions. No fifth option exists - and many teams would rather collect another month of encouraging conversations than let a segment fail cleanly.
I run this with founders on two separate clocks. Infrastructure time - steps 1 and 2 - is real calendar time, and pretending otherwise kills the experiment before it starts. Measurement time is different: the qualifying calls get a fixed four-eight week window, wide enough for prospects to actually arrive, and when it closes, a decision gets made whether the team feels ready or not.
That second clock is the method. It never stretches. Run segments staggered and a verdict lands every few weeks. A full validation cycle takes roughly a quarter. Not fast. But at the end you are no longer hoping a segment works - you have watched one pay, repeatedly, for the same reason. Founders who "haven't found product-market fit yet" are usually waiting for clarity. The ones a cycle away from it are manufacturing it.