How to define a B2B ideal customer profile from your own closed deals

Most ICPs are written in an offsite. Someone puts a slide up, the room agrees on an industry list and a revenue band, and the slide goes into a folder. Six months later the target list in the CRM looks nothing like it, reps prospect whoever looks reachable, and the accounts that actually renew share traits nobody wrote down.

The useful version of your ICP already exists. It is in your closed deals: the accounts that won and stayed, the ones that won and left, and the ones you lost. This guide shows how to read it out of that history, write it as criteria a system can check, and put it where it changes who your team and your agents work on.

ICP, buyer persona, and TAM are three different things

The three terms get used interchangeably, and the confusion is expensive: a persona exercise run in place of an ICP produces a great message aimed at the wrong companies.

The ICP is the narrowest of the three and the one that decides what enters the pipeline. Persona work matters once an account is in; the buying committee across each stage is mapped in buyer intelligence vs. decision intelligence.

Why most ICPs do not change anything

• Aspirational. It describes the customers you want, not the ones that succeed with you.
• Firmographics only. Industry and size are simple to agree on and rarely what separates a win from a loss.
• Untestable. Criteria like “forward-thinking” cannot be checked against an account record, so nobody checks them.
• No disqualifiers. Without a list of who not to sell to, every reachable account looks like a fit.
• Not in the systems. If the ICP is not a field on the account and an input to the score, it does not exist for the people and agents doing the work.

Build it from your deals in five steps

An ICP card you can copy (illustrative)

This example is for a company selling to mid-market revenue teams. The criteria are illustrations of the format, not recommendations; yours come from step 3. Notice that every row names the field it is checked against.

Disqualifiers do as much work as criteria

Teams spend days on what an ideal customer looks like and minutes on what a bad one looks like. The second list is often the more valuable. A clear disqualifier removes accounts before anyone spends research time, sequence slots, or sender reputation on them, and it protects reps from the deals that close and then churn. Find them the same way as the criteria: attributes that show up repeatedly among lost and churned accounts and rarely among best customers.

Disqualifiers also make the low-fit quadrant of a lead score actionable. In the fit and intent model, a low-fit account with high intent gets qualified through a cheaper channel; a disqualified one gets suppressed.

Is your ICP ready to drive targeting?

Run your current ICP through this check before you build a score or a target list on it.

From ICP to execution

Once the ICP is a set of checkable criteria, applying it stops being a quarterly list-building exercise and becomes continuous work, which is the kind of work an agent does well. Scout, the Pipeline Research Analyst, scores accounts against your ICP from intent signals, CRM records, and web telemetry. Account scoring runs on sandboxed deterministic code. No language model performs arithmetic that has a numeric consequence, so a score can be reproduced and checked.

Qualified accounts go to Herald, the Outbound Engagement Rep. Outreach is grounded in the specific signal that triggered it, with the reasoning behind each draft captured alongside it. Your team chooses the mode for each action class, based on its risk tolerance, and can change it at any time. Human-in-the-loop (HITL): The action is drafted and held. It does not execute until a named person on your team approves it. Human-on-the-loop (HOTL): The action executes under a standing policy your team sets. A named person supervises and keeps intervention, override, and revocation authority. Every action, in either mode, is recorded to the audit ledger with its rationale, before-and-after state, and the approver or policy behind it.

Your team still owns the definition: which criteria count, which disqualify, and when the ICP changes. The full four-agent loop is specified on the Revenue Accelerator Stack page, and SaaS-specific applications are on PrescientIQ for SaaS.

How to know your ICP is right

Tag every new opportunity with whether it matched the ICP when it entered the pipeline, then compare the two groups each quarter on three things: win rate, sales cycle length, and retention after the first renewal. If ICP-matched accounts do not do measurably better on at least one of them, the criteria are not separating anything and step 3 needs another pass.

Watch coverage as well. An ICP that is right but only worked on a fraction of the matching accounts leaves pipeline on the table; that gap is what the sales process coverage metric measures.