"Anyone could use this" is not a market, because you cannot make a list of anyone. Narrow until you can name a group, find where they gather, and describe the trigger that makes the problem urgent for them this month. Narrow does not shrink your business; it makes the first hundred customers reachable. You widen later, from a position of having customers rather than guesses.
Someone asks who your product is for. You say it could be useful for a lot of people, and then you list three completely different kinds of company, and the conversation moves on. You have just described why nobody is buying.
The short answer is that a customer segment is only real if you can build a list of it. If you cannot write down twenty names and find their email addresses by Friday, you do not have a target market, you have a hope.
What an ideal customer profile actually is
An ideal customer profile is a description of the kind of organisation or person for whom your product is the obvious answer, specific enough to act on. Acting on it means two things: you can find them, and you can tell whether a given person qualifies.
That second requirement rules out most of what founders write down. "Small businesses that want to grow" fails it, because there is no test that separates a company that wants to grow from one that does not. "Companies using Shopify with more than fifty products and no support staff" passes, because you can check.
It is worth separating two things founders merge. The user is the person who opens the product and gets value from it. The buyer is whoever can authorise spending money on it. In a small company those are the same person, which is one of the underrated reasons small companies are easier to sell to, and in a large one they can be three levels apart with different problems.
When a definition feels right and nothing converts, this split is a frequent culprit. You have described a user precisely and pitched them faithfully, and they have no mechanism for buying anything. The fix is not better messaging to that person; it is deciding whether to sell to someone else, or to price low enough that the user is also the buyer.
Why breadth feels safe and costs you
Narrowing feels like throwing away revenue. If the product works for five kinds of buyer, picking one appears to discard four fifths of the market, which is an uncomfortable thing to do when you have no customers at all.
The arithmetic runs the other way. With a broad definition you cannot make a list, so you cannot do outreach. You cannot write a page that makes anyone feel recognised, so conversion is poor. You cannot pick a community to join, because your buyer is in nine of them. Breadth does not multiply your market; it removes every mechanism you have for reaching any of it.
The four segments you did not pick do not disappear. They wait. Companies that dominate broad markets almost always started in a narrow one, and widened once they had customers and a reputation to widen with.
There is a second cost that shows up later and hurts more. A product built for five kinds of buyer accumulates five sets of half-finished features, because each group asks for something different and none of it ever reaches the depth that makes a product obviously the right choice for anyone.
Narrow products feel thin to their founders and complete to their users. Broad ones feel comprehensive to their founders and inadequate to everybody, because every user arrives looking for the part that was built for them and finds a fifth of it.
Cut by situation, not by category
The instinct is to segment by industry, company size or job title, because those are the fields in a database. They are weak predictors of whether someone will buy, because they describe what a company is rather than what is happening to it.
Stronger cuts describe a situation:
- What they just did. Raised money, launched something, hired a first salesperson, moved off a tool.
- What they are missing. No marketing hire, no CRM, one engineer, no data team.
- What they are using. A specific stack you integrate with, or a competitor you replace.
- What is about to hurt. A renewal, a deadline, a rule change with a date attached.
- Who they are. A solo technical founder is a situation as much as a job title, and it predicts behaviour better than company size does.
The reason situational cuts work is that they carry timing. An industry does not become urgent in March; a company that just lost its only salesperson does.
Narrowing, worked through
Start with "companies that need better sales", which is not a segment because you cannot list it. Cut to companies with no salesperson, which is checkable. Cut again to companies with no salesperson where the founder is technical, which is findable, because those founders are visible on GitHub, in developer communities and on their own about pages.
Now add the trigger. Technical founder, no salesperson, and they shipped something in the last three months. That is a list you can build in an afternoon, a page you can write that makes someone feel recognised, and a reason the problem is urgent now rather than in general.
Notice that each cut removed people who might well have bought. That is the cost, and it is worth paying, because the version you cannot reach converts at zero regardless of how large it is.
The list test
Whatever definition you land on, run it through one check before you build anything around it.
- Write the definition in one sentence, with no adjectives that cannot be verified.
- Open whatever you have, a search engine will do, and try to name twenty real organisations or people who fit.
- For five of them, find a way to reach a specific human.
- Ask yourself what triggered the need this month rather than last year.
If step two takes more than an hour, the definition is too vague. If step three fails, the segment may be real and unreachable, which is the same as unreal for your purposes. If step four has no answer, expect a long sales cycle and a lot of polite silence, because nothing is forcing a decision.
Worth being explicit about the second step, because founders treat it as a formality. You are not estimating whether such people exist, you are opening a browser and writing down names. The exercise is diagnostic precisely because it is tedious: a definition that survives an hour of actually looking is one you can build a company's first year on.
If you finish with twenty names and no way to reach a human at any of them, you have found a real but unreachable segment. That happens most often with large organisations, where the person with the problem is four layers from anyone you can email, and the honest response is to pick a smaller kind of buyer for now rather than to spend a quarter on a doorway that does not open.
Use the evidence you already have
Most founders theorise about their ICP when they already have data pointing at it. Not much data, but more than zero, and observed beats imagined every time.
| Source | What it tells you | What to be careful of |
|---|---|---|
| People who already paid | The strongest signal you have | Small numbers; do not over-fit to one |
| People who tried and stayed | Which use case actually sticks | Free users may never have paid regardless |
| People who replied to outreach | Which situation resonates | Reflects your targeting, not the market |
| Support questions | Where the product is load-bearing | Loudest users are not always typical |
| Who your competitors sell to | A market that demonstrably exists | Their beachhead may be wrong for you |
The most useful exercise takes twenty minutes: list everyone who has engaged seriously so far and look for what they have in common that is not obvious. It is frequently not the industry. It is often a situation, like being the only technical person, or having just lost the person who used to do this.
When to change your mind
A definition is a hypothesis and should be revisited on evidence rather than on frustration. The signal to change is not that outreach is hard, because outreach is always hard. It is that a different group keeps arriving on its own, or that the people you targeted consistently understand the pitch and still do not have the problem.
The one thing not to do is change every few weeks. Each change resets the list, the copy and the community standing, and a segment abandoned before it has been worked properly has not been tested, it has merely been tried.
Where this advice is wrong
You are building genuine horizontal infrastructure
Some products really are for everyone who writes software, and forcing a vertical onto them distorts the roadmap. Even then, narrow the go-to-market rather than the product: you still need a first group to sell to, and it can be a beachhead without being a permanent constraint on what you build.
Your market is small enough to address entirely
If there are four hundred potential buyers in the world, segmentation is a waste of effort and you should simply contact all four hundred. The point of narrowing is to make an unreachable market reachable; a reachable one does not need it.
You have a customer already pulling you somewhere
If one kind of buyer keeps arriving unprompted, that is data and it outranks any analysis. Follow it for a while before deciding it is wrong, even if it was not the market you intended. Products are frequently pulled into their real market by the people who keep showing up.
Narrow is a starting position
The fear underneath all of this is that choosing a segment is permanent and that picking wrong forecloses the rest. It is not permanent, and you will almost certainly move, because the first definition is a hypothesis like everything else at this stage.
What narrowing buys you immediately is a list you can work, a page that makes someone feel recognised, a community you can be a member of, and a sentence that ends a conversation rather than prolonging it. Those are the mechanisms. Without them you are not addressing a large market, you are addressing none of it.
Once you have the definition, the next problem is turning it into an actual list of people and getting a message in front of them, which is covered in how to define an ICP for AI outbound and where your first 100 customers come from.
One practical way to hold both ideas at once: write the narrow definition down and treat it as who you are actively selling to this quarter, not as who is allowed to buy. Nobody is turned away. The list, the copy and the community choice all follow from the narrow version, and the wider audience continues to arrive on its own where it was going to anyway.
Key takeaways
- A segment you cannot build a list of is not a segment, whatever it feels like.
- The useful cut is rarely by industry or company size; it is by the situation someone is in.
- Your first customers are a beachhead, not a ceiling, and picking one does not forfeit the rest.
- The best evidence is who already bought or engaged, not who you think should have.
- If you cannot name the trigger that makes this urgent this month, expect long silences.