Who you sell to, written down.

An ideal customer profile describes the company most likely to buy from you and the person inside it who decides. This is how a profile should be built, and how we build yours.

What it is

An ideal customer profile, or ICP, has two halves. The first describes a company: its industry, its size, its type and where it is. These are called firmographics, and they say which companies can buy. The second describes a person inside that company: the roles that decide on a purchase and the roles that use the product. This half is the buyer persona.

Both halves are needed. A profile with only the company half produces a list of companies and nobody to write to. A profile with only the person half produces job titles at companies that will never buy.

A profile measures fit, and fit is static. It says who could buy, and it says nothing about when. Timing is a separate measurement, covered by buying signals. The two are kept apart on purpose, so each can be checked on its own.

How to build one

The sound method starts from customers, not from the product. The clearest picture of who buys is the set of companies that already bought. You take your best customers, record the same attributes for each one, and look for what they share: the industries that repeat, the size band most of them fall in, the countries, the titles of the people who signed. The profile is the pattern in that data.

The unsound method starts from the homepage. You read what the product does and reason about who should want it. The answer sounds right and describes who the company hopes to sell to.

An early company has few customers, and sometimes none it can name. The method still holds, with one extension. When your own customers are too few to show a pattern, the next best evidence is the customers of products like yours: who reviews them, who appears in their case studies. They bought the same kind of thing for the same kind of reason.

Two rules complete the method. Every field should rest on evidence that can be named, so the profile can be checked and corrected. And the values should be standard ones, so they can be used directly as search filters.

Reading the site

We follow that method, starting from the one thing every company has: its website. An agent reads the landing page and at most two more pages, chosen from product, pricing, customers and about. It does not read blog posts or documentation.

From those pages it records what you sell and what the product does, and whether you sell to businesses, to consumers or to both. Then it collects the evidence the site carries about buyers: every customer you name in a logo, a case study or a testimonial, the job title printed under each quote, and every competitor you mention. One web search adds the alternatives that buyers compare you with.

The evidence

The evidence is then gathered in a single pass. Each customer named on your site is looked up in company data: its industry, its headcount, its type and its country. The answers are counted, so the profile can state how many of your named customers sit in each industry and in each size band.

For your closest competitors we read two things. Their public review pages, which break reviewers down by company size, industry and role. And their own customer pages, which name who bought.

The decision

Each field of the profile is decided from the strongest evidence that exists for it, in a fixed order. Your own named customers come first, because they already bought from you. The reviewers and customers of your competitors come second, because they bought the same kind of product. The wording of your pricing and plans comes last, and it is used only when there is nothing else.

The company half uses standard values. Industries are picked from a fixed list of industry names. Company size is a set of headcount bands, from 1 to 10 employees up to more than 10,000. Company type is one of a fixed set, such as startup, private company, public company or agency. Locations are the markets your site says you sell in.

The person half comes from three sources: the job titles under your testimonials and case studies, the roles your own pages address, and the roles of the people who review competing products. The titles are then grouped by their part in a purchase. Senior titles decide to buy. Hands-on titles use the product every day. For a product sold to businesses the profile keeps both groups. For a consumer product it keeps the users.

The profile also names the segments that should not be targeted.

The basis

Each main field carries its basis: one line that names the evidence behind it, with the counts. When a field rests on nothing but the wording of your site, the line says exactly that. The research notes and the full evidence are stored with the profile.

What you tell us about your business outranks what the site suggests. A stated fact beats an inferred one.

Two more lists

The same research produces two lists that the next step searches with.

The first is your competitors: the companies your customers could buy from instead. Each is marked as a direct competitor, a substitute, a partner or integration, or a company that serves the same customers, and each is tied to its own website.

The second is your topics: ten to twenty short phrases that your buyers write and react to on LinkedIn and X when the subject is the problem you solve. They are taken from the language of testimonials and reviews, in the buyers' own words. Brand names and single generic words are left out. These phrases become the search terms for finding people who post about the problem.