What is an ICP?

The definition

An ICP describes a company, not a person. Size, sector, geography, structure and situation. A buyer persona describes the individual inside it, and the two get confused constantly.

The ICP decides who you contact. The persona decides what you say. Mixing them produces a document that is too vague to build a list from and too generic to write from.

A usable ICP is short. One paragraph of firmographics, one paragraph of situational criteria, and a named list of exclusions. If it runs to three pages it is a positioning document, not a targeting one.

ICP template

A usable ICP fits on one page: a paragraph of firmographics, a paragraph of situational criteria and the exclusions.

What belongs in one

Firmographics you can actually filter a list by: employee count, revenue band, country, region, industry classification. These are the criteria that exist as fields in a company register or a data provider, which is what makes them usable.

Then the situational criteria that separate a good fit from a merely eligible one. Whether they run their own fleet. Whether they hold a particular certification. Whether they have opened a second site, entered a new market or started hiring a function they did not have before.

If a criterion cannot be found in data or inferred from a public signal, it cannot be used to build a list and belongs in the persona instead. "Values quality" is not an ICP criterion. "Holds ISO 9001" is.

ICP vs. buyer persona

You need both documents. They answer different questions, and a team that merges them ends up with a list it cannot build and a message it cannot write.

What an ICP does to the numbers

The point of an ICP is subtraction, and the scale of it surprises people. Take the Polish company register as a worked example, using our own snapshot.

Four criteria take a market of three quarters of a million companies down to three and a half thousand. That is a list a small team can genuinely work, and it is why an hour spent on the ICP is worth a week spent on sending.

It also exposes the cost of each filter. Requiring a filed revenue figure alone discards more than half the sector, and the companies it discards skew smaller and younger rather than being a random sample. Every criterion introduces a bias, and you should know which ones you are accepting.

An ICP example

The Polish logistics filter above, written as an ICP you could hand to whoever builds the list:

• Firmographics. Active Polish company, transport and logistics sector, filed revenue above 2m PLN.
• Situation. Runs its own fleet and posted a driver vacancy in the last 90 days.
• Exclusions. Current customers, open deals, and companies tied to a competitor by contract.
• Buyer. Owner or operations director. The persona document covers what they care about.

The first line comes straight from the register and leaves 3,666 companies. The second line needs a signal source, job adverts in this case, and cuts the list again. Note which source feeds each line, or nobody on your team can rebuild the list next quarter.

To build the profile from your own customer base, follow how to define your ICP. To turn it into a LinkedIn search, use the Sales Navigator guide.

The mistake that makes most ICPs useless

Writing it from ambition rather than evidence. Teams describe the customer they want rather than the one that has actually closed, renewed and referred.

The fix is unglamorous: list your best twenty customers, find what they share, and write that down. Look at deal size, cycle length, churn and referral behaviour, not just industry.

If your best customers have nothing in common, that is itself the finding. It usually means you are selling a horizontal product on relationships rather than on fit, and outbound will be expensive until something changes.

How to test it cheaply

Build a list of two hundred companies that match, send to them, and read the replies. Two hundred is enough to distinguish a targeting problem from a message problem and small enough to run in a fortnight.

Uniform irrelevance means the profile is wrong. Polite brush-offs from the right kind of company mean the profile is right and the message is not. Interested replies from companies outside the profile mean the profile is too narrow.

That distinction is worth more than any amount of internal debate, and it is the only cheap way to find out that your ICP describes a market that does not exist.

How often to revisit it

Once a quarter is enough for most companies, and only if something changed: a new product, a new market, a churn pattern, a competitor moving.

The failure mode is the opposite of neglect. Teams rewrite the ICP every time a campaign underperforms, which destroys the ability to tell whether the profile or the execution was at fault. Change one variable at a time.

The ICP defines who goes on the list; building the contact list is the next step. Once the profile is defined, see TAM analysis: mapping your real market.