B2B ideal customer profile: criteria, scoring and a template

What is a B2B ideal customer profile?

A B2B ideal customer profile (ICP) is a written description of the companies most likely to buy from you, succeed with what you sell, and renew. It is stated as criteria that a rep or a system can test any account against in seconds. An ICP describes companies. A buyer persona describes the people inside them.

The written part matters. An ICP that lives in someone's head is an opinion, not a profile. Two reps with the same opinion still pick different accounts, and nobody can check either of them.

The other word that matters is "test". "Mid-market SaaS companies that care about efficiency" is a description. "B2B software, 50 to 500 employees, on NetSuite, with a finance leader hired in the last six months" is a test. Every account passes it or fails it, and the result is the same no matter who runs it.

ICP vs buyer persona

The ICP decides which companies to pursue. Personas decide how to talk to the people inside them. Teams mix the two up all the time, and the result is a persona doc full of company traits or an ICP full of job titles.

You need both because B2B purchases are made by groups. Forrester's State of Business Buying 2024 found that on average 13 people inside an organization are involved in a B2B buying decision, and 89 percent of purchases involve two or more departments. The ICP gets you to the right account. Inside it, profile the committee, not one champion. A deal with a single contact dies when that contact goes quiet.

B2B ICP criteria: the four groups

B2B ICP criteria fall into four groups. Firmographics are the structural facts: industry, employee count, revenue band, region, business model. Technographics are the tools in place, which signal both fit and integration effort. Situation triggers are the events that create a need now, such as a funding round, a new sales leader, or a hiring push. Disqualifiers are the facts that rule an account out regardless of the rest. Most published ICPs skip the fourth group, and it is the one that does the sorting.

Firmographics

These are the facts a data provider can give you for almost any company: industry, employee count, revenue, headquarters region, ownership, and whether it sells to businesses or consumers. They are cheap to check and they cut a market down fast.

Write each one with a threshold. "50 to 500 employees" can be tested. "Growing companies" cannot. Where a range matters more than a minimum, give both ends. Very large accounts often fail for reasons a minimum will not catch, such as procurement cycles or security reviews you cannot support.

Technographics

Technographics are the tools a company already runs: its CRM, ERP, data warehouse, or the competitor product you would replace. They tell you two things at once. Fit, because some stacks only work with what you sell. Effort, because a stack you already integrate with makes onboarding short, and one you do not makes it a project.

Sources are install-detection tools, job posts that name the stack, and what your own customers run. If your best accounts all share one platform, that is a criterion, and it usually deserves a heavy weight.

Situation triggers

A trigger is a change that creates a need now: a funding round, a new leader in the department you sell to, a hiring push for the role your product supports, an acquisition, a new market entry, a contract renewal coming up on the tool you replace.

Triggers say the most about timing, and they expire the fastest. A new CFO matters in month two and much less in month fourteen. Put a time window on every trigger ("in the last six months") and re-check them more often than the rest of the profile.

Disqualifiers

Disqualifiers are the facts that rule an account out, whatever its other scores. They come from the accounts that churned, the deals that stalled in security review, and the customers support still remembers.

Typical disqualifiers are a platform you cannot integrate with, a buying process you cannot support (public tenders, for example), a size where the product breaks, a region you cannot serve, or a regulated use case you do not cover. Two or three sharp ones are worth more than a long list.

This is the group that makes an ICP useful. A profile that only says who to pursue still lets every account through, because almost every company fits a few positive criteria. Here is a quick test for your current ICP: name five real companies that look like a fit on paper and that you would turn down. If nobody on the team can, the profile does not sort yet. Writing the disqualifiers down is also what lets a rep say no to a bad account in ten seconds instead of after a discovery call.

How to build an ideal customer profile in seven steps

Build it from the accounts you already won and lost. Most of the work is pulling clean data. The analysis itself is quick.

1. Pull your last 12 to 24 months of closed-won, closed-lost, and churned accounts.
2. List what the best 20 percent of customers by revenue and retention share, and what the churned ones share.
3. Write each shared trait as a testable criterion with a threshold ("50 to 500 employees", not "mid-market").
4. Write the disqualifiers from the churned and lost list.
5. Weight the criteria to 100 points, heaviest on the traits that separate winners from churn.
6. Score 50 known accounts by hand. If the scores do not match what the team knows, adjust the weights.
7. Set tier cut-offs and name an owner who re-runs the analysis each quarter.

The 20 percent and the 50 accounts are working choices, not rules. A company with 40 customers looks at all of them. Teams skip the hand-scoring step most often, and it is the cheapest place to catch a bad weight. If a churned account scores 85, a weight is wrong. Better to find that before the model routes a quarter of rep time.

How to score accounts against your ICP

ICP scoring turns the criteria into a number so accounts can be ranked. Give each criterion a weight so the weights total 100, score every account against them, and treat any disqualifier as an automatic zero. Then set tier cut-offs, for example 80 and above as tier 1 and 60 to 79 as tier 2. The weights come from your closed-won analysis. The criteria your best customers share get the most points.

Weights

Start simple. Each criterion is met or not met and earns its full weight or nothing. Partial credit (half the points for a near miss on headcount) is worth adding later, once the simple version has been checked against real outcomes.

Put the weight where the evidence is. If every one of your best customers runs the same ERP and most churned ones did not, that criterion might carry 20 points. If region made no difference to win rate, it carries 5 or it comes out.

Tiers and what each one gets

A score only matters if it changes what the team does. Tie each tier to a level of effort:

The cut-offs are an example. Set yours so that tier 1 is small enough for your reps to research properly.

A fit score says nothing about timing. A tier 1 account that is not in market will still ignore you. Most of your market is not buying at any given moment, which is why fit is one axis and timing is the other. Use the ICP score to decide who is worth the effort, and use triggers or intent to decide when.

ICP template

Copy the template below, or download it as a CSV. The rows are filled in for the illustrative example in the next section. Replace the thresholds and weights with your own.

Tier key: 80 to 100 is tier 1, 60 to 79 is tier 2, 40 to 59 is tier 3, and anything below 40 or with a disqualifier is out.

B2B ideal customer profile example

The example is illustrative: a fictional company, Ledgerline, that sells accounts payable automation to mid-sized companies. Its closed-won review showed that its best customers ran one of three ERPs it integrates with, had a new finance leader, and were hiring AP clerks when they bought. Churned customers were mostly larger companies on SAP, where the integration never worked. That is why ERP carries the most points and why SAP is a disqualifier rather than a low score.

Three accounts run through the template:

Account C is the one that teaches. It matches the industry and has the strongest trigger on the list, and a team without disqualifiers would chase it. Ledgerline's own churn says the deal would close and then fail, so the ICP turns it down before anyone spends a call on it.

Account B shows what the tiers are for. It fits well on structure and stack but has no trigger. It goes into a sequence now and moves up when its AP tool renewal comes within six months.

Keeping the ICP current

Re-run the closed-won analysis every quarter and name one owner. Markets move, the product changes what it can support, and the triggers that predicted wins last year stop working.

Three things go stale on their own. The data goes first, because people change jobs and companies change size, so contact and account data drifts even when the ICP does not. Triggers expire, so re-check their time windows. Weights drift too. Compare the scores of last quarter's wins and losses, and move points toward the criteria that separated them.

Be honest about what the model is. A scoring model built on a few dozen closed-won accounts is a hypothesis to test, not a finding. Treat the first version as a guess with a method behind it, and expect the hand-scoring step and the first quarter of outcomes to change it.

The ICP also needs a system that can apply it. A profile that reps have to score by hand in a spreadsheet gets skipped the first busy week. Writing the criteria down is the first step toward running them in the stack instead of in someone's memory. graph8 Data scores every account for ICP fit and shows the buying committee by role on the company record, so the list a rep works from is already sorted.