An ideal customer profile describes the type of company that gets the most value from your product and is therefore most worth pursuing. It is a company-level description, which distinguishes it from a persona — a description of an individual.
Most ICPs are written in a workshop, circulated once and ignored. The ones that get used are built from data and expressed in terms a rep can check in thirty seconds.
Build it from your own customers
The aspirational approach — describing the customers you wish you had — produces a document nobody trusts. Start with who you actually serve well.
- List your customers with contract value, tenure, retention, expansion, support load and whether they would act as a reference.
- Score them. A simple composite of value, tenure and expansion identifies the top quartile.
- Look for shared attributes in that quartile: size, sector, maturity, technology in use, team structure, how they found you.
- Look at the bottom quartile too. The common attributes of customers who churn fast and consume support are as useful as the positive ones, and frequently more actionable.
- Check the difference is real. With fewer than 50 customers, patterns may be coincidence. Treat early versions as hypotheses.
What to include
Firmographics. Employee count, revenue range, geography, industry. Keep ranges rather than points, and be honest about where the edges are.
Technographics. What they already use. Often the strongest predictor in software, because a company already running a particular system has the problem your product solves.
Situational triggers. What is happening that creates the need: rapid hiring, a funding round, a new regulation, an executive hire in a relevant function, an expiring contract. These are the most useful and the most commonly omitted.
Organisational signals. Does a dedicated function exist for the problem you solve? Is there someone whose job it is to care?
Disqualifiers. Explicit criteria for who is not a fit. This section is what makes an ICP operationally useful and is the one most often left out because it feels negative.
What to leave out
Three things that bloat ICPs without helping.
Aspirational segments you have never sold to. Include them in a strategy document, not in the profile used to route leads.
Individual personas. Related and separate. The ICP says which companies; personas say which people inside them. Combining the two produces a document that serves neither purpose.
Unverifiable attributes. “Values innovation” cannot be checked against a company record, so it cannot be used to score a lead.
A workable format
One page. Longer is not read.
A short paragraph describing the fit in plain language. A table of attributes with target ranges. A list of three to five trigger events. A list of disqualifiers. Two or three named example customers who exemplify it, which is the part reps actually use because it is concrete.
If the document cannot be used by a rep to make a go or no-go decision in under a minute, it is too long.
Making it operational
An ICP that lives in a document changes nothing. Four ways to put it into the workflow.
Lead scoring. Attributes become points, feeding routing and prioritisation. See lead qualification frameworks.
Target account lists. Build the list of companies that match and work it deliberately rather than waiting for inbound.
Disqualification permission. Reps must be allowed, and expected, to decline poor-fit opportunities. If quota pressure makes that impossible, the ICP will be ignored.
Content and messaging. Written for the ICP specifically rather than for a general audience.
The uncomfortable part
A genuine ICP excludes business you could win. That is the point and it is politically difficult, because it means turning away revenue in a quarter when the number is short.
The argument that works is economic rather than strategic. Poor-fit customers churn faster, consume disproportionate support, generate worse references and rarely expand. Their lifetime value is frequently negative once support cost is counted, and they occupy capacity that would serve good-fit customers.
Quantifying this from your own data — retention and support cost by segment — turns a strategy argument into an arithmetic one, which is considerably easier to win.
Keeping it current
An ICP written three years ago describes a product that has changed and a market that has moved. Review it every six to twelve months with the same data exercise.
Watch for drift in particular: if sales is consistently closing outside the profile and those customers are retaining well, the profile is wrong rather than the sales team. The opposite pattern — closing outside the profile and churning — is the one that justifies tightening enforcement.
Using the ICP to score leads
Turning the profile into a score is what makes it operational, and the arithmetic should stay simple enough that a rep can sanity-check it.
Assign points to each attribute by how strongly it predicted retention in your own data. Employee count in range might be worth 20, the right industry 15, a relevant technology in their stack 25, a recent trigger event 20, and so on. Disqualifiers subtract heavily or cap the total.
Then set two thresholds rather than one: a score above which leads go to sales immediately, and a score below which they are not worked at all. The band in between goes to nurture.
Validate the scoring the same way you built the profile — check whether high-scoring leads from the last year actually converted and retained better. A score nobody has validated is a ranking of assumptions.
What an ICP does for the rest of the company
The profile is usually treated as a sales and marketing document, and its most valuable uses are often elsewhere.
Product. Knowing which customers get the most value tells you whose feature requests to weight. A roadmap driven by the loudest customer rather than the best-fit one drifts away from the market you serve well.
Customer success. Accounts outside the profile need different handling and often more of it. Knowing which are which at onboarding allows the effort to be allocated honestly.
Pricing and packaging. If the profile describes companies of a particular size, the packaging should fit them. Many pricing problems are really ICP problems.
Hiring. Reps who have sold into your profile before ramp faster than reps with a generically impressive record.
Testing the profile
A profile is a hypothesis until it predicts something. Two tests confirm it.
The retrospective test: score last year’s closed deals against the profile and check whether the high scorers won more often, retained longer and expanded more. If the profile does not separate them, it is describing something other than fit.
The prospective test: for one quarter, record the profile score of every opportunity at creation, then compare outcomes. This is slower and more convincing, because it cannot be influenced by hindsight.
Either test may show the profile is wrong, which is useful. A profile nobody has tested tends to be defended rather than improved.
From profile to target list
The profile describes a type of company; the next step is naming the actual companies that match it. That list is what turns the ICP from a document into a plan.
Building and working it deliberately is the core of account-based marketing, which is worth considering whenever the total number of realistic customers is small enough to count.
Frequently asked questions
How is an ICP different from a persona?
An ICP describes the company; a persona describes a person within it. A complete picture needs both, kept as separate documents.
Can a company have more than one ICP?
Yes, where the product genuinely serves distinct segments differently. Two is manageable; four usually means the profiles are not specific enough to be useful.
What if we are too early to have customer data?
Start with a hypothesis based on the problem you solve and who has it most acutely, label it as such, and replace it with data as soon as you have twenty or thirty customers.
Who owns the ICP?
It should be agreed jointly by sales, marketing and customer success, with one named owner. An ICP owned by marketing alone tends not to be used by sales.
