Calculating your real ICP (not the one you think you have)

Calculating your real ICP means starting from the customers you've closed who are still with you. You look for what they share that the ones who left or never signed don't, then turn that pattern into verifiable criteria everyone on the team applies the same way. You don't start from the market you'd like to have. You start from your CRM.

Why your ICP on paper isn't who actually buys

The ICP you have usually describes who you want to sell to. Who actually buys only shows up when you look at the data on who signed and stayed. If you ask five people on your team what your ICP is, you'll probably get five slightly different answers. In many B2B teams the ICP is only partly defined and isn't written down: there's a broadly shared idea, but each new person interprets it their own way and there's no document to check it against.

The consequences show up day to day:

• The target account list is rebuilt from scratch every quarter, because nobody remembers what criteria the last one was built on.
• Your best customers by margin and by retention share traits that appear nowhere in the ICP.
• When the quarter looks weak, you take on customers who don't fit, and those are exactly the projects that go worst.
• The target profile is still the one from the first year, even though the product and the price are now different.

Start with your best current customers, not your ideal market

The starting point is the list of customers you closed who are still with you today, pulled from the CRM and completed with what the CRM doesn't have. The aspirational buyer persona can wait.

What data to gather for each customer

For each won account, gather in a single table:

1. Company data: sector, size, country, business model and the tools they use, if relevant to what you sell.
2. Deal data: creation and close dates, amount, discount given, source and who took part in the decision.
3. Post-sale data: whether they're still a customer, whether they've renewed or expanded, how much delivery effort they've required and, if you have it, the real margin.

Do the same with lost opportunities and with the ones that stalled. Without that contrast you can't tell what separates those who buy from those who don't.

Fill in whatever the CRM is missing from your billing tool and from the people who deliver the service. It's a one-off manual job, and it's what makes the result worth having.

What to discard before looking for patterns

• Accounts that came in through a personal relationship or as an exception, and that you wouldn't go after again.
• Customers who bought something you no longer sell, or at a price that no longer exists.
• Deals closed with an exceptional discount to save a quarter.
• Duplicates: the same company recorded several times in the CRM, with records kept by different people.

Separate "who buys" from "who stays"

A customer can fit perfectly at the moment of sale and no longer fit months later, so the real ICP has to hold at both moments. To see it, make three cuts on the same table:

1. Won versus lost: which traits appear more in what you closed than in what you lost or what stalled.
2. Customers who stay versus customers who leave: what those who renew or expand share, and what the ones who left had.
3. Profitable versus costly: what distinguishes customers with good margin and little delivery effort from those who drain the team.

The real ICP lies at the intersection of the three. The first cut only tells you who finds it easy to buy from you.

The same exercise produces the anti-ICP: if the customers who turn out worst share a trait you can see before they sign, that trait is an exclusion criterion. Write it down as one.

Firmographic criteria and situational criteria

A useful ICP combines firmographic criteria, which say which companies fit, with situational criteria, which say which ones are ready to buy now. Many teams stop at the first kind: size and sector, without looking at what's happening to the company.

• Firmographic: sector, size, geography, business model, sales team structure. They build the universe of accounts that fit.
• Situational: what's happening in the company that creates the need. For example, it has just hired its first dedicated salesperson, it's changing CRM or it's opening a new market. They decide the timing of the contact.

Firmographic criteria define the list; situational ones decide who on that list gets called this week. Without the second kind, you reach out when there's a gap in the diary, not when the account shows a buying signal.

The After case shows the combination: the firmographic criterion was companies of 500 to 1,000 employees in Spain, in finance, software, consulting, legal and pharma; the situational one, young workforces with an intense workload. Those criteria produced 925 companies in Clay.

Quantify the fit, don't just describe it

An ICP criterion is only useful if someone else can check it against a data point without asking you. "Technology companies that value innovation" isn't an actionable ICP; a specific employee range, a specific minimum deal size and at least one person dedicated to selling are.

For each criterion, write the field, the value or range that meets it and the source of the data. If a criterion has no source, it's an opinion. A simple test: give the same list of accounts to two people on the team and ask them to mark which ones fit. If they don't agree, the ICP isn't written precisely enough yet.

From ICP to a scoring model the team uses

The documented ICP becomes useful when it's translated into a score that orders the accounts inside the CRM:

1. Few criteria: the ones that separated the customers who stay from those who leave, not every one you can think of.
2. Two separate scores: one for fit, which changes little, and one for signals, which changes every week. Mixing them hides why an account moves up.
3. Tiers with consequences: for example, A accounts with account-by-account research, B accounts in a sequence by segment and C accounts that are only worked if a signal appears. A tier that doesn't change what the team does is no use.
4. The score explained: each account shows why it has the score it has. A score nobody understands gets ignored.
5. Inside the CRM: the criteria are fields and the score is calculated where the team works, not in a separate spreadsheet.

Review it when your product or your pricing changes

An ICP defined before a change of product, price or market probably no longer describes who you sell to today, so each of those changes forces a review. On top of that, set a periodic review in advance; every six months is a reasonable starting point.

In each review look for four things: deals won outside the ICP, deals lost inside it, customers inside the ICP who have left and criteria the team has stopped applying. If the contacts were enriched once and nobody has touched them since, your scores are built on stale data. Keep each version with its date, so that "what criteria was this list built on" always has an answer.

Common mistakes when calculating the ICP

• Starting from the market you want and then looking for data that confirms it.
• Looking only at who buys, without looking at who stays and who is profitable.
• Not writing the anti-ICP, and letting every weak quarter decide the exceptions.
• Keeping it in a presentation instead of in the CRM's fields.
• Piling up criteria until no account meets them and the team stops applying them.

Where to start

This exercise is the starting point for any pipeline generation effort: without verifiable criteria, the target account list is rebuilt from scratch every quarter. For a first reading of where your team stands, the free sales self-assessment (in Spanish) includes the ICP among the statements you tick.

It's also the first thing we review in the RevOps & GTM Diagnostic, on the real data in your CRM. If you already know this is the missing piece, ICP definition, TAM sourcing and account scoring are in the RevOps & GTM Projects catalogue. If you don't trust your data, read the signs that your CRM needs an audit.

Frequently asked questions

What is the difference between an ICP and a buyer persona?

The ICP describes the type of company that's a good fit for you. The buyer persona describes the people inside it who decide on, evaluate or use what you sell. You need both, in that order: a good message for the right role is no use if the company doesn't fit.

How many customers do I need to calculate the ICP?

Enough that the pattern doesn't depend on a single case. If you still have few, write the ICP as a documented hypothesis, with its criteria and its sources, and test it against every win and every loss. A written hypothesis can be corrected; a hunch can't.