An ICP That Ends in Named Accounts

Why most ICPs do no work

A typical ideal customer profile reads: "B2B technology companies with 200 to 2,000 employees in Europe and North America." It is true, and it is useless. It matches tens of thousands of companies, so it cannot tell a seller which one to research on Monday.

The test of an ICP is simple. Hand it to two colleagues separately and ask each to name ten companies that fit. If their lists barely overlap, the profile is a description. If they overlap heavily, it is a tool.

Start from customers, not from the market

Do not begin with the market you would like to have. Begin with the customers you already have.

Take your five to ten best customers. "Best" means they bought without a struggle, they use what they bought, they stayed and they would recommend you. Leave out the biggest logo if it was a one-off that came through a founder's friend.

For each one, write down what was true on the day they bought, not what is true today:

• What kind of company they were: sector, size, where they operate.
• What had recently changed there.
• What problem they named, in their own words.
• What they already had in place that made you a fit.
• Who was involved in the decision.

Then read across the rows. The things that repeat are your profile.

The five parts

1. Firmographics. Sector, size band and geography. Be narrower than feels comfortable: "regional logistics operators with two to ten sites" is better than "logistics".

2. Situation. What is usually happening at a company when it buys from you: a new site, a new leader, a system reaching its limit, a regulation with a date. This is the part most profiles leave out, and it is the part that tells you when.

3. Problem. The problem as customers describe it, not as your website does. One sentence.

4. Compatibility. What must already be true for you to deliver: a system they run, a team they have, a way they work.

5. People. The roles that took part in your past deals: who owned the problem, who held the budget, who had to approve the fit.

Add three knock-outs

A knock-out is a fact that removes a company from the list however good it looks otherwise. Write three. Common ones:

• They signed with a competitor in the last year.
• They are below the size at which your price makes sense.
• They run a system you cannot work beside.
• You cannot deliver in their country.

Knock-outs save more time than any scoring. They are also the first thing a team argues about, which is a sign they matter.

From profile to names

Now turn the profile into companies. Aim for 25 to start. A first list longer than that will not be researched.

Work through these sources in order and stop when you have 25:

1. Companies that look like your best customers: same sector, size and situation.
2. Companies your customers compete with, supply or buy from.
3. Companies where someone you know now works.
4. Companies that have shown interest before and went quiet.
5. Companies in the news for the situation in part 2.

For each name, write one line: why this company, and why now. If you cannot write the line, the company is a guess and comes off the list.

Is the profile sharp enough?

Check it against four questions:

1. Did two colleagues, working apart, name mostly the same companies?
2. Does every company on the list have a "why now" line?
3. Would your best three customers have made the list on the day they bought?
4. Is there at least one company you wanted to include that the knock-outs removed?

If the answer to the fourth question is no, the knock-outs are too soft. A profile that excludes nobody you like is not a choice.

What comes next

A list of 25 is a starting point, not a plan. The accounts on it are not equal, and your team cannot give all of them the same effort. The next lesson covers scoring and tiering: how to rank the list and decide who gets the most attention.