An ICP should help you decide which accounts deserve pursuit and hold up when you inspect what happens after the sale. Validate it against closed-won and churned-customer data.1 The first version may change. Most founders initially got their ICP wrong.2 A few customers can make a weak idea look viable, leaving the team to support, maintain, and explain it.3 Revise the profile in response to repeated patterns, not one-off excitement.
Write the current profile
Write down the profile you are testing and keep its wording visible. That lets each account confirm or challenge a specific field.
An ICP describes the company type most likely to get lasting value from your product.4 Use company-level fields such as industry, size, business model, technical environment, and buying triggers.5 Keep the person you plan to reach in a separate buyer persona, which describes an individual inside the company.6
Identify which company traits you are testing and which person-level details belong in the buyer persona. The profile is ready to test when it gives you a clear account-level hypothesis that you can check against live results.
Build the evidence set
Use a fixed account sample and test the profile against actual outcomes. Start with closed-won deals. Review the last 30 to 50 closed-won deals and identify common company sizes, industries, geographies, revenue bands, and buyer job titles.7
Bring RevOps, Product, Sales, Marketing, and Success into the review. Each function holds different evidence: data, product patterns, objections, market signal, churn, and expansion.8 Record which profile fields fit each account, what happened in the deal, and what the customer said after purchase.
Look for traits that recur across accounts and treat a trait seen in one account as a hypothesis. The review is ready for testing when it shows patterns you can check in pipeline and customer outcomes.
Check outcomes by ICP tier
Measure win rate, pipeline quality, and churn rate by ICP tier.9
Ask which profile traits appear in wins, where pipeline quality weakens, which groups retain poorly, and whether the account traits explain the outcome or merely reflect the sample. If a criterion appears in wins while the downstream result stays poor, remove it or use it as a qualification rule. Keep a field when it helps explain a consistent outcome pattern. You should be able to distinguish fields that support durable performance from those that only describe surface similarity.
Test the profile with outbound
Use outbound as a live test of the written criteria. It shows how the market responds when you deliberately pursue matching accounts.
Data from outbound sales is a better signal of what is working than leads from investors and friends.10 Use real-world feedback to test, refine, and iterate the profile.11 Keep the account test separate from the person reached, since the same idea can produce different outcomes when pitched to different people.12
Ask what the target account recognized, where the response broke down, and whether the pattern repeats across accounts with the same profile. Then determine whether the response pattern belongs to account fit or to the person reached.
Revise and schedule the review
Make each revision traceable. Record what changed and what result should improve so the next review can judge the change against its intended effect.
After you collect real data about the profile, document it in an ICP Matrix.13 Review the profile quarterly using fresh CRM data, closed-won deal patterns, and updated customer feedback.14
At each review, ask which assumptions survived contact with customers and pipeline, which weakened, and which new pattern deserves a test. Keep the profile flexible enough to change and specific enough to guide account selection.
What not to do
Use these checks before accepting an update:
- Do not update the ICP from assumptions alone. Base updates on real-world data.15
- Do not treat every customer or sale as equally good. That hides the difference between ideal customers and customers who are less than ideal.16
- Do not spend time choosing a problem without also figuring out who to solve it for. Companies that skipped that work often regretted it.17
- Do not conclude an idea is bad after speaking to the wrong people. Talking to the wrong people can make the idea look bad and lead you to give up.18