Ideal customer profile (ICP): what it is and how to build one

ICP stands for ideal customer profile. In B2B it describes an account, meaning a company, and the profile is only useful if two people reading it would pick the same companies.

In sales, the ICP decides where reps spend their time. It sets which accounts get assigned, which inbound leads get a fast reply, and which deals get qualified out early. A rep with a sharp ICP spends their week on accounts that can close.

In marketing, the same profile chooses ad audiences, the account lists behind ABM, and the topics content is written about. When marketing and sales keep separate versions, marketing fills the funnel with leads sales will never call. Keep one document and let both teams edit it.

Some companies say "ideal client profile" instead, especially services firms. It means the same thing.

Pick the accounts with the ICP first, then choose the people with the persona. Doing it the other way round gets you a list of VPs of Sales at companies that will never buy.

Most ICPs are written in a workshop from memory. The better source is your CRM, because it already records who bought and who walked away.

No closed-won data yet? Write the ICP as a hypothesis from the problem you solve, test it with a small outbound list per segment, and rebuild it from replies and meetings after the first 90 days.

Copy this table into a doc or spreadsheet and fill in the middle column. The right-hand column is the test: if you cannot name where the data comes from, the field is not finished.

A made-up company that sells contract management software to in-house legal teams. Its closed-won deals clustered at mid-size software and services companies that had just hired their first or second lawyer, and its losses were mostly very large companies already locked into an enterprise suite.

Notice the last row. The ICP alone gives a list too large to work by hand. The trigger cuts it to the accounts worth contacting this month, which is the job of signals.

A written ICP does nothing until it ranks real companies. The steps below take the profile from a document to a list a rep can work.

The GTM playbook covers what happens next: sequences, routing, and how to measure each segment.

Every outbound system I build starts with the client's ICP written as code: the filters, the exclusion list, and a fit score stored next to each account. A few rules come from doing this many times.

Disqualify before you score. Industry codes often cannot tell a customer from a competitor, since both file under the same category. I match on what a company says it does in its own description, and keep that check separate from the score.

Show the gaps. When an input cannot be read, say the stage is unknown or the homepage will not load, that part of the score gets half credit and is stored on its own. A missing field stays visible instead of hiding inside a total.

Measure team shape directly. "Lean sales team" sounds like a judgment call, but headcount by department turns it into a number: sales and marketing headcount as a share of the company. I compare that number within a segment only, because services firms and software companies count their people differently.

Pair fit with a signal. High fit decides who is on the list. A signal decides who hears from us this week. For Antimetal, a fit score ranked every account by size, engineering footprint, and how much reliability matters to them, and signals decided when reps reached out.

If you want to see where your own targeting stands, the free GTM scorecard grades it along with the rest of your go-to-market setup, and the GTM glossary defines the terms used here.