How to create an ideal customer profile
An ideal customer profile (ICP) describes the type of company that gets the most value from your product and brings you the most value in return. To create one, study your strongest existing customers, find the traits they share (industry, size, region, technology, buying triggers), check those traits against lost deals and churn, write them as testable criteria, and review them every quarter.
| Component | Question it answers | Example (Northwind Analytics) |
|---|---|---|
| Industry | Where do we win most often? | Mid-market wholesale distributors |
| Company size | Which headcount or revenue band? | 50 to 500 employees |
| Geography | Where can we sell and support? | US and Canada |
| Technology | What must they already run? | A cloud ERP |
| Trigger | Why would they buy now? | Hiring ops or supply chain roles |
| Buying roles | Who signs, who uses it? | VP Operations, Head of Procurement |
| Disqualifiers | Who should we not chase? | Under 20 employees, fully custom in-house systems |
What an ideal customer profile is
An ideal customer profile (ICP) is a short, specific description of the companies you should sell to: the accounts that buy quickly, get real value, stay, and are worth the effort of winning. It is written at the company level, not the person level.
That is the main difference between an ICP and a buyer persona. The ICP says which companies to target (for example, US distributors with 50 to 500 employees running a cloud ERP). A persona describes the people inside those companies (a VP Operations who owns stock accuracy and reports to the CFO). You need both, but the ICP comes first, because a well-written persona at the wrong company still does not buy.
A useful ICP has three qualities:
- Specific. "Mid-market B2B companies" is not an ICP. "Wholesale distributors in the US with 50 to 500 employees" is.
- Testable. Every criterion can be checked against a real company, ideally from public or purchasable data.
- Evidence-based. It comes from customers you already have, not from the market you wish you had.
How to create an ICP in six steps
- List your strongest customers. Pull closed-won deals from the last 12 to 24 months and pick the ones that score well on several measures: fast sales cycle, good contract value, high usage, low support load, renewal or expansion. Ten to thirty accounts is enough to start.
- Describe them with the same fields. For each account, record industry, headcount band, revenue band if known, country, business model, key technology, and what was happening at the company when they bought (new funding, a new leader, a hiring push, a failed tool).
- Look for patterns. Group the accounts and see which traits repeat. You are looking for the two or three attributes that most of your strongest customers share and that most other companies do not.
- Check against the losses. Run the same exercise on lost deals and churned customers. A trait that shows up equally among wins and losses does not belong in the ICP. A trait that shows up mostly among losses becomes a disqualifier.
- Name the buying roles. For each pattern, note who championed the deal, who signed, and who used the product. This links the ICP to your personas and tells prospecting which titles and seniority levels to search for.
- Write it down as criteria. Use the template in the table above: industry, size, geography, technology, triggers, buying roles and disqualifiers. Keep it to one page so sales, marketing and anyone building lists can apply it the same way.
If you have no customers yet, start from the problem you solve and the companies that feel it most acutely, write a hypothesis ICP, then treat the first 20 to 50 conversations as the evidence that confirms or rewrites it.
A worked example
Maya Chen runs revenue operations at Northwind Analytics (northwind.co), which sells inventory software and a supplier portal. Her sales team had been chasing "any company with a warehouse".
She pulled the deals Northwind had won over the past two years and kept those that closed in under 90 days and renewed. The pattern was clear: most were wholesale distributors, most had between 50 and 500 employees, nearly all already ran a cloud ERP, and many had posted operations or supply chain roles in the months before they bought. The lost deals skewed toward very small firms and companies with fully custom in-house systems.
Her finished ICP fits on one page and matches the table at the top of this answer. It also gave her two personas: the VP Operations who feels the stock-accuracy pain, and the Head of Procurement who manages suppliers. These numbers and traits are illustrative, but the method is the same for any company.
Turning your ICP into a prospect list
An ICP only pays off when it changes who you contact. The practical test is whether each criterion can become a filter: industry, employee range, country, job function and seniority map directly onto most B2B data tools. Triggers such as hiring are harder, because they change week to week.
If you work with an AI agent, you can hand it the ICP in plain language. With Leadsets, an agent connected over MCP turns criteria like "VP Operations at US distributors with 50 to 500 people" into people search filters, and every result comes back with match reasons, which makes it easy to spot where the ICP is too loose. The scrape_jobs tool reads careers pages for hiring signals, and the free ICP builder skill in the skills library interviews you for about five minutes and saves your answers as reusable filters. The free plan includes 1,000 records with no card, enough to test whether an ICP returns the companies you expect.
From there, the usual next steps are building the list itself (how to build a prospect list) and scoring inbound and outbound leads against the same criteria (how to qualify B2B leads).
Common mistakes
- Making it too broad. If almost every company qualifies, the ICP is not doing any work. Narrow it until it excludes most of the market.
- Copying the total addressable market. The market is everyone who could buy. The ICP is the subset most likely to buy, succeed and stay.
- Using untestable traits. "Innovative culture" cannot be filtered. "Hired a head of RevOps in the last six months" can.
- Ignoring disqualifiers. Knowing who to skip saves as much time as knowing who to chase.
- Writing it once. Products, pricing and markets change, so the customers who fit best change too.
How often to update it
Review the ICP every quarter, and sooner after a pricing change, a new product line or a move into a new region. Compare the latest wins and losses with the written criteria and adjust. Timing signals such as hiring or new leadership are worth layering on top of a stable ICP; see how to use intent data for sales for how teams combine fit and timing.
Questions and answers
What is an ideal customer profile?
An ideal customer profile (ICP) is a description of the type of company that gets the most value from your product and is most valuable to you as a customer. It usually covers industry, company size, geography, technology, buying triggers and disqualifiers. It describes companies, not individual people.
What is the difference between an ICP and a buyer persona?
An ICP defines which companies to target, while a buyer persona describes the people inside those companies who influence or make the purchase. Teams usually define the ICP first and then write personas for the roles involved in buying. Both should be based on real customer evidence.
What should an ideal customer profile include?
A practical ICP includes industry, headcount or revenue band, geography, key technology, the triggers that make a company ready to buy, the buying roles involved and clear disqualifiers. Each criterion should be testable against real company data. Keeping it to one page helps sales and marketing apply it consistently.
How many ideal customer profiles should a company have?
Most early-stage companies do well with one ICP, and larger companies often have two or three tied to distinct products or segments. More than that usually means the profiles are too loose to guide prospecting. If two ICPs share most criteria, merge them.
How do you create an ICP if you have no customers yet?
Start with the problem you solve and the companies that feel it most acutely, and write a hypothesis ICP from that. Use your first 20 to 50 sales conversations to test it, noting who engages, who buys and who drops out. Rewrite the ICP as soon as real wins give you evidence.
Can AI help build an ideal customer profile?
Yes. An AI assistant can analyse your closed-won and lost deals for shared traits, and when connected to a B2B data source it can turn the finished ICP into search filters and check which companies match. The judgement about which customers count as ideal should still come from your own revenue and retention data.
Go deeper