"Mid-size companies in tech" is not an ICP. It's a description broad enough to include a 50-person startup and a 5,000-person enterprise, and neither your messaging nor your outreach can serve both well. A real ideal customer profile does one job: it filters out the leads that were never going to convert, before your team wastes time on them.
Start with who actually bought, not who you wish would buy
The most reliable ICP isn't aspirational — it's descriptive. Look at your last 10-15 closed-won deals. What did they have in common? Company size, industry, the specific problem they were trying to solve, who championed the purchase internally. Patterns here are worth more than assumptions about your "dream customer."
Get specific on four dimensions
- Firmographics — industry, company size (by employee count or revenue), geography
- Role — title, seniority, department. "Marketing" is not a role; "Director of Demand Gen" is.
- Trigger — what situation makes someone actively look for a solution like yours right now (recent funding, new hire in the role, a specific pain point)
- Disqualifiers — just as important as who fits: who obviously doesn't, and why. This is what actually does the filtering.
Write the disqualifiers down explicitly
Most ICP documents describe who to target and stop there. The more useful version also states who to exclude — "not a fit if under 20 employees," "not a fit if no marketing headcount," "not a fit outside North America" — because that's what actually keeps your list from bloating back out to "mid-size companies in tech."
An ICP's real job isn't to describe your best customer. It's to give whoever's building your list a clear rule for saying no.
Revisit it every quarter
Your best-fit customer six months into the business often looks different from your best-fit customer at eighteen months, as you learn where you actually win versus where you just occasionally get lucky. Treat the ICP as a living document, not a one-time exercise.
A tight ICP is also the single biggest lever on how good a lead list can be — no amount of data enrichment fixes a target definition that was too broad to begin with. If you want a second pair of eyes on yours, send it over and we'll tell you honestly whether it's specific enough to build a precise list from.