An ideal customer profile is a written description of the companies your offer can serve well: industry, size, location and the observable traits that make a first conversation worth having. It describes an organization, not an individual, and it only earns its keep when it also tells you which companies to leave out.
What it is
An ICP is a filter you can run, not a paragraph of ambition. A usable one names four things: the industry or category, the geography you can actually serve, a size band you can support, and at least one trait you can see from the outside before anyone answers the phone. That last item is what separates an ICP from wishful thinking. "Companies that value good design" cannot be searched. "Dental clinics with two or more locations in Toronto whose site has no online booking" can.
How it works
You build it backward from evidence. List the accounts that closed fastest and stayed longest, then ask what they had in common before they were customers, not after. Turn each shared trait into a field you can query: category, city, headcount band, whether a website exists, how many public reviews the location has. Run that query, look at the first twenty companies it returns, and correct the profile where the results disappoint you. An ICP that has never been tested against a live search is still a hypothesis.
Why it matters for winning clients
Most wasted outbound is not a copywriting failure, it is a targeting failure that copy cannot rescue. A clear ICP raises reply rates by removing companies that were never going to buy, and it makes every downstream decision cheaper: which channel to use, which case study to reference, what to charge. For a small agency it also protects capacity. Saying no to a poor fit in the research stage costs minutes; discovering it in the third meeting costs a week.
Example in LeadCanvas
A web development studio in Manchester writes its ICP as: independent trades businesses, United Kingdom, one to ten staff, active on Google Maps with recent reviews, no website shown. It runs that as a search in LeadCanvas, reviews the returned listings and finds the criterion is tighter than expected. Our own July 2026 study of 231,349 deduplicated local business listings found 6.8% of United Kingdom businesses show no website, against 37.0% in Brazil, so the studio widens the geography rather than loosening the qualifying trait.
Common mistakes
Writing the ICP after the list instead of before it, which turns the profile into a description of whoever happened to answer. Mixing the company profile with the job title of the person you want to reach: those are two different filters and they belong in two different fields. Setting criteria nobody can verify from public data, so the search cannot be repeated. And never revisiting it, which is how a profile written for last year's offer quietly steers this year's pipeline.
Related terms
Lead scoring ranks the companies that already match this profile. Firmographic data supplies the structural fields the profile is written in. Total addressable market answers how many companies fit it at all. The links below open the checklist that walks through writing one and the search tool that tests it.