How to build an ICP from your best customers

An ideal customer profile is the short description of the company you should be selling to. Not a person, not a mission statement, just the traits that a company needs to have for your product to be an easy yes. Get it right and every other part of outbound gets simpler, because you stop wasting messages on companies that were never going to care.

Most founders build their ICP by guessing: they picture the customer they wish they had and write that down. There is a better source sitting right in front of you. This is how to build an ICP from the customers you have already won, as a founder doing your own outbound, with no research team.

What an ICP is, and what it is not

An ICP describes a company, not a job title. It is the firmographic and situational shape of an account that fits you well: their industry, their size, their setup, and the moment that makes your product relevant. It is deliberately narrow, because its whole job is to help you say no to the accounts that will waste your time.

A few things an ICP is not:

• It is not a buyer persona. A persona is the human you talk to (the ops lead, the founder, the finance manager). That matters, but it lives inside the account, and you find the account first.
• It is not your total addressable market. The market is everyone who could theoretically buy. The ICP is the slice that buys quickly, stays, and gets real value.
• It is not a wish. It is a pattern you can point at in your own customer list.

Why your best customers describe it better than you can

When you guess at an ICP, you describe the company you want. When you read it off your best customers, you describe the company that actually works. Those are rarely the same, and the gap is where months of wasted outbound hide.

So start with evidence. Pull the customers who bought without a long fight, stuck around, and got a result they would happily tell someone about. Ignore the ones who churned, haggled forever, or needed constant hand-holding. The winners share traits. Those shared traits are your ICP, and they beat any guess because a real customer already proved them out.

You do not need dozens of accounts for this. Even three or four strong customers will show a pattern, as long as you look at the ones who genuinely fit rather than the biggest logo you happened to close.

The fields to capture

Open a simple table, one row per best customer, and fill in these columns. Keep it plain. The point is to spot what repeats, not to write a report.

• Industry or niche. Not just "software" or "retail", but how narrow it really is. "Independent dental practices" tells you far more than "healthcare".
• Size band. Headcount, rough revenue, number of locations. Look for the range where you win, and be honest about where you start losing to complexity or price.
• The setup that makes them a fit. The tools, systems, or way of working that your product plugs into or fixes. This is often the sharpest filter and the one people forget to write down.
• Who feels the pain, and who signs off. The person the problem hurts is not always the person who approves the spend. Note both, because your message and your path in depend on it.
• The trigger event. What was happening at the company when they decided they needed you? A new hire, a new location, a growth spurt, a tool they outgrew. This is the field that connects your ICP to timing, and it is the one that turns a static list into a this-week list.

When you have filled the table, read down each column and circle what repeats. The overlaps are your ICP. Write it as one or two sentences you could say out loud.

A worked example

Here is a hypothetical to show the shape (not real data, just an illustration). Say you sell scheduling software and your four best customers look like this when you line them up:

• Industry: home-services businesses (plumbing, HVAC, electrical), not "SMBs" in general.
• Size: 10 to 40 staff, two or three locations. Below 10 they do it on paper, above 50 they buy a heavier platform.
• Setup: already using a basic booking tool or a shared calendar that is visibly straining, plus a field team that takes calls on the road.
• Pain vs sign-off: the office manager feels the daily chaos; the owner signs the cheque.
• Trigger: they opened a second location, or hired their first dispatcher, in the last few months.

Written as a sentence, the ICP becomes: "Home-services companies with 10 to 40 staff and two or three locations, already outgrowing a basic booking tool, that recently added a location or a dispatcher." That is something you can hunt for on purpose. "Small businesses that need scheduling" is not.

Common mistakes

Three errors quietly wreck most ICPs. They are easy to avoid once you know to look for them.

Making it too broad

A wide ICP feels safe because it keeps your options open. It is actually the expensive choice: it lets in accounts that fit poorly, and it gives you no way to prioritize. If your profile matches half the companies in a country, it is not a profile, it is a category. Narrow until it is almost uncomfortable, then check it against your best customers again.

Ignoring the trigger

Most ICPs describe a static company and stop there. But fit without timing just tells you who might buy someday. The trigger event is what makes an account worth contacting now, and it is usually hiding in plain sight in your own deals. Skip it and you have a good list with no sense of when to act on any of it.

Never updating it

Your ICP is a snapshot of what worked so far. As you close more deals, the pattern sharpens, and sometimes it shifts. The founders who keep a good ICP revisit it every quarter or so, add their newest wins, and quietly retire the traits that stopped predicting anything.

How to keep it alive

Treat the ICP as a living document, not a one-time exercise. A light habit keeps it honest:

• After each new win, add the account to your table and see if it fits the existing pattern or bends it.
• When a deal goes badly, note why. Bad-fit accounts teach you the edges of your ICP as clearly as good ones teach you the center.
• Every quarter, re-read the profile out loud. If it no longer sounds like your last ten customers, update it.

How to turn your ICP into a target list

An ICP is only useful when it becomes companies you can contact. To get there, take each trait in your profile and turn it into a filter, then find the accounts that match.

• Use verified, licensed business data and official platform APIs, plus public sources read like a human researcher would, to find companies that match your industry, size, and setup.
• Layer the trigger event on top. A company that fits your ICP and just hit your trigger is a reason to reach out today, not someday.
• Grow the list by finding lookalikes of your strongest customers: companies that share the exact combination of traits, not just the broad category.

A sharp ICP is the foundation of the whole process. If you want the wider playbook it sits inside, from profile to outreach, see our guide on finding customers who are ready to buy.

Where Digital Spoiler fits

Building the ICP is work you can do by hand, and it is worth doing. Keeping it current and checking the whole market against it every day is the part that eats your time. Digital Spoiler learns your profile from your best customers, watches the market every day for fit and buying signals, and delivers the companies most likely to buy to your Slack, email, or dashboard, each with the evidence and a suggested opening angle.

You can run the whole thing yourself. If you would rather it ran on its own, that is the point of the product. Have a look at pricing, or tell us what you sell and we will show you the companies that fit.

The short version

• An ICP describes the company to sell to, not a person or a wish.
• Read it off your best customers, not your hopes.
• Capture industry, size, setup, who feels the pain vs who signs off, and the trigger event.
• Avoid going too broad, skipping the trigger, or never updating it.
• Revisit it every quarter so it tracks who you actually win.
• Turn each trait into a filter, then find lookalikes and triggers.