Most ideal customer profiles are written from a gut feel. Yours is already sitting in your CRM.
An ideal customer profile, or ICP, says which kind of account to spend your time on. The evidence for it is the deals you closed, including the ones you lost.
Start with the export
Export every deal that reached an outcome in the last 12 to 24 months: won, lost and no decision. Keep the lost ones. A pattern that only looks at wins cannot tell you what to avoid. Do not clean away the awkward customers either. The one who paid late and asked for everything is data.
Use your own deals only, and follow your employer’s rules on what can leave the CRM.
Define “best customer” with a rule
Pick a rule you can see, not a feeling. Mine has five tests. A best-fit deal was:
- Won.
- At or above a minimum deal size.
- At or under a maximum sales cycle.
- Easy to serve: a 4 or 5 on a 1 to 5 scale, which is my default. That is a judgment. Write it once and do not edit it later.
- Still a customer.
I start the size and cycle limits at the median of the won deals. The median is the middle value, so one huge deal does not drag it. The rule is strict on purpose. A best customer is worth winning, quick to win and easy to keep.
Then count
Group the deals by industry, size band, buyer role, how they found you and the trigger that made them buy. For each group, work out the win rate: won divided by won plus lost. Leave no-decision deals out of the rate. Then read each group like this:
- Lean in: win rate at or above your overall rate, and at least one best-fit deal. Build the profile around it.
- Watch: win rate at or above overall, but no best-fit deal yet.
- Mixed: win rate below overall, but at least one best-fit deal.
- Avoid: win rate below overall and no best-fit deal.
- Too few deals: fewer than three closed deals. Do not read anything into it.
Three deals is my default minimum, not an industry standard.
An example
Redwater Integration is a fictional drone maker I use for examples. In my made-up data it has 30 deals: 17 won, 10 lost and 3 with no decision. Its overall win rate is 17 of 27, about 63 percent.
Electric utilities show 6 wins out of 7 closed deals, with two best-fit deals. That is Lean in. County roads departments show 2 wins out of 4 and no best-fit deal. That is Avoid. None of these numbers is real. They are there to show the method.
The catch
Your deals only show what happened to the accounts you reached. A group can look weak because you rarely sold to it. A group with three deals can swing on one. Treat what you find as a hypothesis, then test it: see Test your ICP on 20 accounts before you build on it.
I built a free Claude skill that interviews you about your best customers and drafts the profile for you to approve: get the free ICP builder skill (opens in a new tab). The full method, with the spreadsheet that does the counting, is in the ICP discovery workbook.
I’m Seth Levanen. I find the AI shortcuts that keep the quality and drop the busywork. Email me if you want to talk about it.