How to Create an Ideal Customer Profile (ICP): Step by Step

Short answer: you create an ideal customer profile (ICP) by starting from your best current customers, figuring out what they have in common across four layers (firmographics, pain or trigger, buying behavior, and their channels), and testing that against both your CRM data and a handful of customer interviews. Then you pour it into a profile with clear inclusion and exclusion criteria, and you use it to sharpen your targeting, content, and sales conversations. Count on half a day to a full day of work for a first version, then refine it every quarter.

This is the operator version: the step-by-step plan itself. If you first want to know what an ICP precisely is and how it differs from a target market, read our companion what is an ideal customer profile (ICP). Below we assume you know that basis and simply want to get to work.

Before we zoom in, here is the full plan at a glance. Each step builds on the previous one: you climb from your best existing customers up to a profile you dare to test and adjust every quarter.

What you need before you start

For a first ICP you do not need expensive tools, but three things at hand:

• Access to your CRM or billing, so you can sort your customers by revenue, margin, and tenure.
• A list of your 10 to 20 best customers. Best here means: high value, stay long, little hassle, and ideally also: the ones you would want to clone.
• Two to three hours to make calls. An ICP that comes from data alone misses the why. Schedule a few short conversations.

Step 1: Select your best existing customers

How you do it: export your customer list and score each customer on three axes: total value (revenue or margin), tenure (how long they stay), and ease (how much support or discussion they cost). Set the top 10 to 20 aside. This is your gold layer, not your biggest customer and not your newest.

Example: a B2B software company that steered on revenue thought enterprise was its ideal customer. When it sorted on margin and tenure, mid-sized manufacturing companies (50 to 250 employees) turned out to be far more valuable: smaller deals, but three times longer tenure and hardly any support.

Common mistake: basing your ICP on the customers who signed the fastest. Signing fast says nothing about value; it often only says something about price. Look at who stays the longest and delivers the most.

Step 2: Map the firmographics

How you do it: for each top customer, note the hard characteristics: industry, company size (employees and revenue), region, business model, and the technology they use. Look for the pattern. Often you see one or two industries and a clear size band recurring.

Example: the firmographics of the software company above became concrete: “manufacturing and wholesale companies in the Benelux, 50 to 250 employees, revenue between 10 and 50 million euros, still working with loose Excel processes.”

Common mistake: staying too broad. “SMEs in Flanders” is not a firmographic, that is a phone book. If your profile fits more than half the market, it is not a profile.

Step 3: Determine the pain and the trigger

How you do it: firmographics tell you who, pain tells you why now. For each top customer: what concrete problem did you solve, and which event drove them to start searching? That event (the trigger) is gold for your timing: a new executive, a growth spurt, a merger, a new regulation.

Example: the shared pain was “we are losing oversight of our inventory and making too many manual errors.” The recurring trigger: a growth jump where the number of orders doubled in a year and Excel gave out.

Common mistake: filling in the pain from your own product (“they need our tool”). That is wishful thinking. Customers do not buy a tool, they buy a solution to a pain they name that way themselves. Use their words, not those of your pitch.

Step 4: Analyze the buying behavior

How you do it: map how your best customers buy. Who takes the initiative, who decides, who signs? How long does the cycle take, what budget was involved, and through which steps did it run? In B2B, one person rarely buys; you sell to a buying committee of several roles.

Example: among the top customers, the operations manager started the search, but the CFO signed. Average cycle: four months, budget around 25,000 euros per year. That changes everything about how you build your content and outreach: you have to serve both the operational pain and the financial business case.

Common mistake: confusing the ICP with the person. The ICP describes the company; the roles within that company belong in your buyer personas. Keep those two separate. Then develop your personas with our step-by-step plan for an effective B2B buyer persona.

Step 5: Find where these accounts are

How you do it: an ICP you cannot find back in the real world is theory. Note for each top customer: on which channels are they active, which trade media do they read, which events do they attend, which LinkedIn groups, which tools give away their profile? This makes your ICP immediately operational for targeting.

Example: the manufacturing companies turned out to be findable via a specific industry federation, two trade fairs per year, and LinkedIn filters on job title plus company size. That immediately became the basis of their first 10 B2B leads.

Common mistake: skipping this step because the profile “feels finished.” Without findability, your ICP stays a nice document that no one uses.

Step 6: Test with interviews and sharpen

How you do it: call three to five of your best customers. Ask: why did you choose us, what was the real problem, who helped decide, and what almost made you drop out? The data from steps 1 to 5 gives you the skeleton; the interviews give you the nuance and the exact wordings. Work in exclusion criteria: which characteristics predict a bad customer?

Example: the conversations revealed that companies without an internal “driver” always got stuck during implementation. That became a hard exclusion criterion: no deals without a point of contact who is responsible for the project.

Common mistake: making your ICP once and then freezing it. Markets shift. Put a quarterly moment in the calendar to test your profile against your most recent won and lost deals.

Worked example: an ideal customer profile

This is what the end result looks like when you go through the six steps. This is deliberately concrete; an ICP of five vague bullets does not work.

ICP: Growing manufacturing and wholesale company in the Benelux

• Firmographics: manufacturing or wholesale, 50 to 250 employees, revenue 10 to 50 million euros, headquarters in Belgium or the Netherlands, still working largely with Excel and loose systems.
• Pain: loses oversight of inventory and orders, makes manual errors that cost money, cannot scale without hiring more people.
• Trigger: doubling of order volume in twelve months, or a new operations lead who wants to put things in order.
• Buying behavior: operations manager starts, IT evaluates, CFO signs. Cycle around four months, budget around 25,000 euros per year, buying committee of three to four roles.
• Where to find them: industry federation, two trade fairs, LinkedIn (function plus company size), trade media around supply chain.
• Exclude: no internal project owner, fewer than 50 employees, pure services company without physical inventory.

With this profile, marketing knows exactly who to target, content knows which pain to play, and sales knows immediately whether a lead fits. That cuts noise, and noise is expensive. An agency that applies this rigorously in account-based marketing achieved on the order of 189 percent more MQLs at a 22 percent lower monthly cost; read the ABM case study.

Common mistakes (summarized)

• Too broad. If your profile fits half the market, it steers nothing. Dare to narrow.
• Wishful thinking. Your ICP is not who you dream of reaching, but who already extracts the most value today.
• Confusing ICP with persona. The ICP describes the company; the persona describes the people in it.
• No exclusion criteria. An ICP that only includes is a wish list. Daring to say no makes it sharp.
• Only data or only gut. Numbers say which customers are valuable, conversations say why. You need both.
• One-off and then forgotten. Sharpen every quarter based on won and lost deals.

Checklist: is your ICP ready to use?

• Based on at least 10 real top customers, scored on value and tenure
• Firmographics so sharp they fit less than a fifth of the market
• Pain and trigger in the words of the customer, not of your pitch
• Buying behavior mapped: who starts, who decides, who signs, what budget
• Concrete channels where you find these accounts
• At least three customer interviews worked in
• Clear exclusion criteria written down
• A quarterly moment scheduled to adjust

Want to know whether a won account is worth targeting? Calculate it quickly with the free value-per-lead calculator, then consistently qualify your leads afterwards with a scoring model based on your ICP.

What this means for your visibility in AI

A sharp, well-structured ICP does not only make your marketing better, it also makes you a better source for AI answers. When you describe your pain, trigger, and audience explicitly and in customer language on your site, both search engines and language models recognize precisely who your solution is for. That is exactly the logic behind GEO for B2B: making clear who you serve, so you get mentioned when exactly those accounts ask ChatGPT or Google.

Want to turn your ICP into a running lead machine? Check out our approach to lead generation, or get in touch and together we build the profile and the campaigns that go with it.