Skip to content
ArticleUpdated 4 min read

How to build an ideal customer profile from your last ten deals

Build your ideal customer profile from evidence, not aspiration: list your last ten customers who paid and stayed, record the same eight facts about each, and keep only the traits most of them share. Add a list of who is not a fit, write the result as plain sentences, then test it on fifty real prospects before trusting it.

Start with evidence, not ambition

Most ideal customer profiles are wish lists. "Mid-market SaaS companies with a growth mindset" describes who the founder would like to sell to, not who buys. The fix is to start from the customers you already have, specifically the ones who paid and stayed.

If you have fewer than ten, use what you have plus anyone who said yes to a pilot or a paid trial. If you have none, use the people who took a call and described the problem. Label it a hypothesis.

Step 1: pick the ten

From your customer list, take the ten best by this rule: paid, still paying (or renewed), and you would happily have ten more like them. Leave out the big logo that churned, the friend who bought as a favour and the account that takes all of your support time.

Step 2: record the same eight facts about each

Use the same columns for every customer. Fill in what you know; look up the rest.

FactWhy it mattersExample
Industry or tradeWho has the problemPhysiotherapy clinics
Company sizeBudget and complexity3 to 15 staff
LocationLanguage, law, time zoneUK and Ireland
Who found you or signedThe person you will be writing toClinic owner or practice manager
What they used beforeThe thing you replacePaper diary, or a general booking tool
The triggerWhy they looked when they didOpened a second site, hired a receptionist
The words they usedHow they describe the problem"Patients keep not turning up"
How fast they boughtSales cycle lengthTwo weeks

The "words they used" row is the one most profiles skip and the one that is most useful later. It becomes your search phrases and your opening lines.

Step 3: keep only what most share

Go column by column. A trait shared by seven or more of the ten goes in the profile. Four to six: note it as "often". Three or fewer: leave it out.

Say your ten look like this (illustrative): nine are clinics with 3 to 15 staff, eight are in the UK, seven were signed by the owner, six moved off a general booking tool, and the trigger varies. Your profile is clinics of 3 to 15 staff in the UK, sold to the owner, often replacing a general booking tool. The trigger is not yet a pattern; keep collecting.

Step 4: write the not-a-fit list

This is as valuable as the profile itself. Write down who looks like a fit but is not:

  • Too small to pay (sole practitioners, in this example).
  • Too big to buy without procurement.
  • Already locked into a competitor's multi-year contract.
  • Your own customers and open deals. See stop pitching your own customers.
  • Anyone in a sector you cannot serve legally or practically.

Step 5: write it as sentences

Columns are for analysis. The profile you use should be readable in thirty seconds. Something like:

We sell appointment scheduling to physiotherapy clinics in the UK with 3 to 15 staff. The buyer is usually the owner. They come to us when no-shows start costing real money, often after opening a second room or site. They describe it as "patients not turning up" or "the diary is a mess". We are not a fit for sole practitioners or hospital groups.

That paragraph tells a colleague, a contractor or a scoring tool exactly who to look for and how they talk.

Step 6: test it on fifty real people

A profile is a prediction. Test it.

  1. Find fifty people the profile says are a fit.
  2. Contact them with the same message.
  3. Record who replies, and who replies positively.
  4. Compare against fifty people the profile scores lower, if you can afford to.

If the fit group does not reply noticeably more than the others, a trait in your profile is not doing what you think. The lead scores calibration guide shows how to read these numbers without fooling yourself, including why fifty is the smallest sample worth reading.

Common mistakes

  • Describing the company you want to become. Profile today's buyers; update as you move upmarket.
  • Too many traits. Eight facts is plenty. Twenty means nothing matches.
  • No language. Without their words, you cannot search for them or write to them.
  • Never revisiting it. Every ten customers, rerun the worksheet.

Turning the profile into a lead list

The profile tells you who; you still need to find them. Firmographic traits (industry, size, location, title) suit a directory search. The words and triggers suit signals: posts, reviews and job adverts where people describe the problem. The glossary entry on ideal customer profile covers the terms; firmographics covers the company-level traits.

In Sluice, the paragraph from step 5 is what you type as a customer profile. Every person found is scored 0 to 100 against it on fit, worth approaching, and whether their words show the problem now; 70 is the line for worth contacting. Insights then shows reply rate by predicted score, which is step 6 done continuously. Agency plans allow one profile per client. See pricing.

Do this today

Open a spreadsheet, list your ten best customers, and fill in the eight columns. It takes about an hour, and it will change who you contact next week.

Questions people ask

What is the difference between an ICP and a buyer persona?
An ideal customer profile describes the company or account that buys and stays. A buyer persona describes the person inside it who decides or uses the product. You need the first before the second is useful.
What if I have no customers yet?
Use the people who said yes to a call, a pilot or a waiting list, and the problems they described. Mark the profile as a hypothesis and test it faster.
How often should I update my ICP?
Revisit it every quarter, or after every ten new customers, whichever comes first. Update it when the test numbers say a trait does not predict replies.
How specific should an ICP be?
Specific enough that two people reading it would pick mostly the same companies from a list. If it would include half the market, it is too broad.

Try it on your own market

Sluice quotes the worst-case price before anything runs and charges only for lookups that found something, so finding out costs close to nothing.

Get started