Most founders reach this point with a few customers, a few more prospects who went quiet, and a mentor who keeps telling them to "niche down". Nobody explains how to choose the niche, or what to do with the customers who do not fit it.
The question comes up repeatedly on Hacker News, and the typical version runs like this: a founder closed one pilot in a specific niche, cannot repeat it, and watches horizontal inbound dry up at the same time. "Narrow your ICP" sounds right but gives them nothing to do on Monday morning.
This article sets out a practical way to build an ideal customer profile from a small customer base, and to test whether it is right before you rebuild your marketing around it.
Use a working definition: who you will turn away
An ICP is useful only when it changes decisions. If yours describes "B2B companies with 50 to 500 employees", it will not stop you from pursuing a bad-fit deal or writing a homepage that speaks to everyone.
A better test is whether your ICP tells you which inbound lead you will politely decline this week, which accounts your outreach skips and which audiences your ads exclude. If it does none of that, you have a market description, and a market description does not guide a small team.
The cost of a loose ICP shows up in revenue. In the Fullcast and Ebsta 2025 update, strongly ICP-fit accounts showed 5.1 times higher lifetime value than low-fit accounts, with two times lower churn and four times the likelihood of expansion. The same research found that 64% of CROs revisit their ICP once a year or less, and that only 23% of accounts truly fit the ICP. These are vendor figures without a stated sample, but the direction matches what most founders feel: bad-fit customers gradually take over the roadmap.
Start from the customers who got value fastest
With a small base, do not average all your customers together. Pick the two or three who reached value fastest and with the least hand-holding, and study them closely.
For each one, answer four questions. What happened in their business that made them look for a solution at that moment? Who felt the pain day to day, and who signed? What were they using before, whether a spreadsheet, an agency or a competitor? And why did they choose you over doing nothing?
Replies to the founder on Hacker News point in the same direction: dig into why the first pilot happened, meaning the specific problem, the urgency and the budget, then ask that customer to introduce you to peers with the same pain. Introductions from a happy customer are the cheapest way to test whether the pattern repeats.
April Dunford's positioning framework, as summarised by Wynter, starts in a similar place: with the alternatives your best-fit customers would otherwise use, and the segment where your particular strengths create disproportionate value. Your ICP is that segment, described precisely enough to act on.
Describe the trigger and the buying group, not only firmographics
Industry, size and region tell you where to look. They do not tell you when an account is ready to buy. Add two things to every ICP.
The first is the trigger, meaning the event that makes the problem urgent. Examples include a new compliance deadline, a funding round that brings a hiring plan, a migration away from an old system, or a new head of department with a mandate to change things. Triggers are what make outreach timely instead of random.
The second is the buying group. Forrester's State of Business Buying 2026 reports that the average buying decision involves 13 internal stakeholders and nine external participants, and that procurement acts as a decision-maker 53% of the time. Even in a smaller company, the person who loves your demo is rarely the only person who decides. Write down who signs, who can block the deal and what each of them needs to see.
The LinkedIn B2B Institute's 95:5 rule adds timing to the picture: it holds that only about 5% of B2B buyers are in market at any given time. Your ICP describes who should buy, and the trigger describes when.
Test it in the market, then let it drive everything
A written ICP is a hypothesis. Test it with a small outbound run of 20 to 30 accounts that match it closely, and with a version of your homepage headline written only for them. If replies and calls improve, keep going. If they do not, the weak point may be the trigger, the buyer or the message, so change one variable at a time.
Once it holds, use it everywhere. The 6sense study summarised by CustomerThink found that buying groups filled four of five shortlist spots on day one and bought from their top-ranked vendor 77% of the time. If your ICP does not shape your website copy, ad targeting and content, buyers in that segment may build their shortlist without you on it.
A tested ICP also prepares the company for growth. Techstars lists the things a founder should be able to define before handing off sales: the ideal customer, the problem solved, the positioning, the recurring objections and what a successful sale looks like. The ICP comes first on that list because every other item depends on it.
What to do this week
- Pick your two or three fastest-to-value customers and write one page on each: trigger, daily users, signer, previous alternative and reason for choosing you.
- Write your ICP as one paragraph that includes a trigger and a buying group, plus one sentence describing who you will turn away.
- Ask your happiest customer for two introductions to peers with the same problem.
- Run a 20-account outbound test against the new ICP and track replies and meetings separately from your usual pipeline.
- Rewrite your homepage headline for the ICP alone and compare demo requests over the next month.
Niching down can feel like giving up revenue, but for a small team it is how you get enough repetition to learn what works. Choose the customers you serve best, describe the moment they need you, and let that description decide where you say yes and where you say no. If you want help turning a tested ICP into targeting, messaging and campaigns, Stolos Digital works with B2B teams on exactly that.