Ideal customer profile: build it from the customers you already won

Build an ideal customer profile from the deals you won and kept, interview those wins to learn why, and test the profile against losses and churn.

Instant Expert EditorialPublished Updated 5 min read

Build your ideal customer profile (ICP) from the accounts you have already won and kept, then interview them to learn why they bought. An ICP describes the kind of company that gets the most value from your product and is likeliest to buy and stay: its industry, size and current tools, and above all the situation that makes your product matter to it. A profile written in a whiteboard session usually describes the market you hope for.

An ICP picks companies; a persona picks people

An ICP describes an account. A buyer persona describes a person inside that account who shapes the purchase. HubSpot's ICP guide, produced with Clearbit, bases an ICP on company-level firmographic and technographic data points such as employee count, technologies used and industry, and treats personas as a separate tool alongside it.

Ideal customer profileBuyer persona
DescribesA type of companyA type of person in the buying decision
Typical fieldsIndustry, size, location, tools used, trigger situationRole, trigger, expected result, concerns, evaluation steps
Main useDeciding which accounts to pursue and which to skipDeciding what to say, and to whom
Main evidenceWin, loss and retention records, plus interviewsInterviews with recent buyers

Use the ICP to choose accounts, then personas to plan the conversation inside each one. Still choosing between very different markets? Start with how to choose your first customer segment and market segmentation research.

The example: job costing for construction subcontractors

Here is a made-up example: you sell job-costing software to construction subcontractors. The company and every number below are invented.

Over the last year you had 60 qualified opportunities and won 18, a 30% win rate (18 / 60). Your current ICP says "construction subcontractors, 10 to 500 employees, US." That covers almost all 60 opportunities, so it helps nobody decide which to chase. (A profile that fits everyone is a polite way of having no profile.)

1. Split your wins, losses and churn by facts you can check

Export every qualified opportunity with its outcome and facts you can verify for each account: trade, employee count, number of active jobs, accounting system and whether they are still a customer. In the example, split by accounting system:

  • Subcontractors already on one particular mid-market accounting package: 20 opportunities, 12 won (12 / 20 = 60%).
  • Everyone else: 40 opportunities, 6 won (6 / 40 = 15%).

Then check retention among the 18 wins. Suppose 11 of the 12 in the first group are still customers after a year (92%), against 2 of the 6 in the second (33%). That is a stronger signal than win rate alone: a group you win easily but lose within months is a poor fit.

With counts this small, treat the pattern as a lead, not a proven rule. One referral partner or one strong salesperson could explain the difference, and interviews tell you whether it has a cause.

2. Interview recent wins to learn why they bought

April Dunford, author of Obviously Awesome, argues for win analysis over loss analysis. In a conversation with Dreamdata, Dunford says companies lose deals for all kinds of reasons, while wins show why a customer chose you over the alternatives. Dunford's questions cover what triggered the need to do something different, how the buyer made a shortlist, where they did their research, who was on the shortlist, why they picked you and why not the others. Best fit, in Dunford's framing, is the set of characteristics that make an account care a great deal about the value only you deliver.

Ask recent wins:

Before you looked at us, how were you tracking costs against each job? What happened that made that stop working?

Suppose the subcontractors who stayed shared one situation: a controller reconciled field hours with the accounting system by hand every week and had missed a cost overrun on at least one large job. The accounting package mattered because your integration removed that manual step. Now the ICP has a reason behind it, and you can look for that situation in companies on other accounting systems.

Interview a few losses and churned customers too, using how to interview customers who chose another product and how to interview churned customers.

3. Fit the ICP on one page, including who to skip

FieldHypothetical exampleEvidence
Industry and typeSpecialty trade subcontractors (electrical, mechanical)Win records
SizeRoughly 25 to 150 field staff, 10 or more active jobsWin and retention records
Current toolsMid-market accounting package; field hours in spreadsheetsWin records, interviews
Trigger situationA cost overrun discovered after a job closed7 of 9 win interviews
Who feels it, who signsController feels it; owner signsInterviews
Poor-fit signalsFewer than 10 active jobs; owner does the books personallyChurn records, 3 churn interviews

Keep the poor-fit row. It saves as much sales time as the positive criteria.

4. Test it on the next deals before you retarget

Check the profile against the next month or two of new opportunities. Do matching accounts move faster and close more often? If the difference disappears, go back to the interviews. Revisit it whenever you change the product, pricing or market, since each can change who gets the most value.

Your next step

Export last year's qualified opportunities, split them by two or three verifiable facts, and compare win rate and retention for each group. Then book interviews with five recent wins in the strongest group.

To hear from companies that match your profile but aren't customers yet, Instant Expert can find people who fit a description such as "controllers at electrical subcontractors with 25 to 150 field staff." It sends your invitations, and you pay for each call that gets booked. The directory page for finance professionals in construction shows the kinds of people you could reach.