AI Product Marketing
ICP card

What is an ICP (ideal customer profile)?

In the method we use, an ICP, or ideal customer profile, is a card for one buyer segment that writers and sellers work from. It says which companies buy, how they behave, who decides, what makes it urgent and who it is not for. Every row on the card is sourced or marked ASSUMPTION.

Last updated by Omair Dawood.

On this page

What an ICP card contains

The method is our own: the icp-persona skill behind our ICP card tool. These are the parts of a card, in the order the method's output format sets them.

  1. A header and one line. The header dates the card and names its sources and the outputs meant to cite it. Then one sentence of 20 words or fewer: who buys, what makes it urgent, and who it is not for.
  2. Firmographics. Industry, size, stage, geography, tech context and budget owner, kept short because it does the least work.
  3. Behavior. What they already do, buy and complain about in public, each with why it predicts fit. The quality bar asks for at least as many behavior rows as firmographic rows.
  4. The buying committee. One row each for the economic buyer, champion, user and blocker. Each row holds their real question in their own words, what they fear, and the proof that answers it.
  5. Trigger events. What happens just before the problem turns urgent, and how an outsider would see it happen.
  6. Where they look. Named communities, newsletters, people and search phrases. The method's example: "LinkedIn" is a category, not an answer.
  7. Disqualifiers. At least three kinds of company the card is not for, each with the tell that spots it early.
  8. Their words, and the card's own test. Phrases they say and vendor words they never say, then a citation ledger and a list of open assumptions.

The method refuses to invent a buyer. A row is evidenced or carries the literal label ASSUMPTION, and there is no third state. A card with no evidence at all may exist, but it says so on its face. And in the method's words, a card that nothing cites is decoration.

The eight steps that build one

These follow the method's own order. Step 7 is the one the method says people skip.

  1. List the evidence first. Start from the positioning's "who cares a lot" line, then sales notes, lost deals, reviews and public threads, each dated. Name one buyer seat.
  2. Write the firmographics, briefly. Every row is evidenced or labelled ASSUMPTION.
  3. Write the behavior. List observable habits that predict fit, such as how they solve the problem without you today.
  4. Map the committee in their words. "How do I know this will not become my problem at 2am?" is a question. "Wants reliability" is a summary, and a writer cannot use it.
  5. Name the trigger events. A funding round, a person leaving, a missed target or a renewal date, and where each one shows from outside.
  6. Name where they look. Specific places, people and search phrases.
  7. Write at least three disqualifiers. Give each the reason it fails and the tell. If one feels uncomfortable to write down, the method treats that as a sign it belongs on the card.
  8. Label every line, then set the test. Sweep for unlabelled rows, open the citation ledger, and log where card and evidence disagree. Write the one line last.

Worked sample: Basecamp's small-business page, read as a card

In our view, a real ICP card stays inside the company, so a public "who we are for" page is the closest public copy. Basecamp's small-business page says, under the heading "We’re for the Fortune 5,000,000":

"Most big software companies fight over the Fortune 500. The whales, the thousand-seat contracts, the enterprise deals."

"They can have them."

"Our favorite customers are the Fortune 5,000,000. The small and medium-sized businesses of the world, the individual freelancers, the creative shops that do the best work, not the most work."

Read on . Source: Basecamp, "Come small, come all"

A paragraph further down reads:

"Have 3 employees? We’d be honored to have you as a customer. Just you on your own? We’re here for you."

Read on . Source: Basecamp, "Come small, come all"

What the copy does well. It writes a disqualifier out loud, which is the step the method says most cards skip. In our view, "the thousand-seat contracts" even works as a tell, because a seller could spot it early. The copy also gives the size band in plain numbers. "Have 3 employees?" and "Just you on your own?" let a reader place their own company at once.

What is missing, against the card. The quoted copy names three segments at once: small and medium-sized businesses, freelancers and creative shops. The method writes one card per segment, so this is three cards. The copy describes size and type, which is the firmographic part, and says nothing about behavior. It names no buying committee and no trigger event. In our view, a page for buyers can skip all of that, but the card writers work from cannot.

The card, filled from the page. Each open slot stays a question, because the public page does not answer it.

Card partFrom the pageStill open
Segmentsmall and medium-sized businesses, freelancers, creative shops[the one segment this card covers: needs a choice]
Firmographics"3 employees", "Just you on your own"[the upper size limit and the budget owner: needs evidence]
Behaviornothing on the quoted copy[what they already use and complain about: needs evidence]
Disqualifiers"the thousand-seat contracts, the enterprise deals"[two more, each with a tell]
Trigger eventsnothing[what happens just before they look: needs evidence]

The one line, from the page's own words, with its gap left open:

In one line: Small businesses and freelancers buy [after what: needs evidence]; thousand-seat enterprise deals never will.

Common mistakes

  • Guesses that look like facts. The method labels every unevidenced row ASSUMPTION. Its reason: an invented buyer presented as fact spoils every asset built on it.
  • A card of job titles and headcounts. The method's point is that nobody can build a value pillar out of a headcount range. Add behavior rows until they match the firmographic rows.
  • Summarized wants. Write the committee member's question as they would ask it, with a question mark at the end.
  • No "not for" section. Write at least three disqualifiers, each with a tell a seller can spot early.
  • Channel names instead of places. Replace "LinkedIn" or "podcasts" with the named community, show or person.
  • Two segments on one card. Split it. In the method's words, a card for two segments is two cards pretending to be one.

Questions people ask

How do you build an ideal customer profile?

Start from evidence, not a template. List what you know about real buyers, dated. Then follow the eight steps above in order. Label every unevidenced row ASSUMPTION, and write the one-line summary last.

What does "buyer persona" mean?

In our method, the persona is the people half of the card. The ICP rows say which companies buy and how to spot them. The buying committee rows say which people inside those companies decide, and what each one is really asking. Our card holds both on one page.

How many buyer personas should I have?

The method sets no number. It sets one card per segment, with one row per committee role inside it. In our view, start with the segment your real deals come from, then add a card only when a second segment shows up in your evidence. Two assets written for different buyers is the method's signal that a new card is due.

Sources

Build yours with the ICP card tool

The tool asks what you sell and for one buyer segment, with your product website if you have one. Then it asks what you know about real buyers, and what only you know about them. You get who buys by company and by behavior, the committee with their real questions, triggers, where they look and who it is not for. Every row is sourced or marked ASSUMPTION. With no buyer evidence, most rows say ASSUMPTION and the card says so. It takes about 2 minutes of typing. The tool's next step is a positioning statement.

Build an ICP card for one buyer segment