The ICP on your wall is a hypothesis

Most startups have an ideal customer profile pinned up somewhere. It looks like a fact. It is a guess, written before anyone bought anything.

That's fine. Every company has to start with a guess. The trouble starts when the guess stops being tested.

Meet Mesa Foods. Mesa is the fictional company in our Startup workbook: two founders making high-protein tortillas. Their ICP was clear. Athletes, gym-goers, post-workout recovery. The launch campaign said so.

Then they looked at their first 40 customers. Twenty-eight were pregnant or postpartum women. Six were meal-preppers. Four were parents packing school lunches. Two were athletes. The customer they designed for was 5% of who showed up.

The customers explained why, in their own words. One message said: “I eat one of these at 3am when I wake up to feed the baby. I don’t have to cook anything and I’m full enough to go back to sleep.” Others said “it’s like real food that takes ten seconds” and “I use it when I can’t deal with cooking.” Nobody said “high protein.” Nobody said “athletes.” They said fast, real, filling.

It passed two of the four checks so far, and that is useful. The 3am story turned up five times: two Instagram messages, a farmers market conversation, a Google review and a post in a breastfeeding Facebook group. That looks like five sources. But five places are not five people, and Mesa hadn’t yet checked whether one customer was behind more than one of them. So “repeated” was not yet: count the customers, not the channels. It was consistent: the middle of the night, no preparation. It cost them something: three customers reordered within two weeks without being asked. Whether it lasts is not tested yet.

So they wrote both beliefs down, with what would prove each wrong. Athletes: wrong if they don't reorder within 30 days. Neither of the two did, but two early customers are far too few to judge, and a launch campaign aimed elsewhere might have found more athletes. So athlete demand isn't ruled out. It stays not tested yet, with the next test named: reach athletes directly and see whether they reorder. New mothers: wrong if they choose protein bars over Mesa at the same price. That one is not tested yet, so it stays a working belief with the next test named.

The hard part wasn't the data. It was resisting two easy conclusions: dropping athletes on the strength of two customers, or keeping them on the wall with “early days, they’ll come once we market harder.” The first reads too much into too little. The second protects an ICP from evidence.

Try it on yours. Count your first customers by who they actually are, not who you hoped for. Write next to your ICP what would prove it wrong. If you can't, it isn't a profile yet. It's a wish.