How to build an ICP that actually converts

Ask ten B2B teams for their ICP and nine will hand you a slide with three industries, a headcount range and a geography. That is a segment, not a profile. It describes who you happen to have sold to, which is not the same as who is likely to buy next.

A useful ICP starts from closed-won data, not aspiration. Export your last fifty wins and your last fifty losses, then look for the attributes that appear in one group and not the other. Headcount rarely survives this test. Things like whether the company has a dedicated RevOps hire, whether they run a specific CRM, or whether they posted a role you sell into within the last ninety days usually do.

Those last examples matter because they are signals rather than static traits. A company with 200 employees today had 200 employees last quarter and will next quarter. A company that just hired its first VP of Sales has a budget cycle opening right now, and that is the difference between a cold email and a timely one.

In SigmaGTM you express this as a saved search combining firmographics with live signals — tech stack detected on the site, hiring activity scraped from careers pages, funding events, and page changes we track through Apify. Anything matching enters a sequence automatically instead of waiting for someone to run a report.

Finally, keep the profile narrow enough to be wrong. If your ICP matches 40,000 companies, it is not selecting anything. Aim for a list you could genuinely work in a quarter — for most teams that is between 400 and 2,000 accounts — and expand only once conversion holds.