Why Does Your ICP Keep Describing Companies That Never Buy?

Why does your ICP keep describing companies that never buy?

Because most ideal customer profiles describe what a company is rather than what a company is doing. Industry, headcount, and funding stage tell you a company could plausibly need you. They tell you nothing about whether anyone there is currently feeling the problem you solve, and that second thing is what actually creates a deal.

I see this constantly. A founder writes an ICP that reads like a database query, builds a list from it, gets almost no response, and concludes that outbound does not work for their market. The list was not wrong exactly. It was just a list of companies that match the shape of a customer without any evidence that they are in motion.

Over six years I have taken on more than 70 projects for more than 25 clients, and the pattern in the ones that went well was almost never firmographic. It was that something had just changed for them. That is the thing worth building your profile around, and this piece is about how to do it deliberately rather than by luck.

What is a first-order signal, and why is it not enough?

A first-order signal is a stable fact about a company. Industry, employee count, revenue band, location, funding stage, the software they run. These are easy to find, easy to filter on, and they are the entire content of most ICP documents I get shown.

The problem is not that they are wrong. It is that they are shared by thousands of companies and they do not change. A signal that was true last year and will be true next year cannot tell you that now is the moment. If your entire targeting is first-order, you are choosing who to talk to but not when, and timing is most of what determines whether a message lands.

First-order signals do real work as filters. They are how you avoid wasting time on companies that could never buy at your price or could never use what you make. What they cannot do is prioritise, and a list of ten thousand plausible companies with no ordering is functionally the same as no list at all.

What counts as a second-order signal?

Something that just changed, and that plausibly creates or exposes the problem you solve. A new role hired. A platform migration underway. A pricing page that was rewritten. A product launch that suddenly needs supporting content. A key person leaving. These are not descriptions of a company. They are descriptions of a company in motion.

The reason they work is that they carry an implied deadline. A company that hired its first marketing person has a person who needs to show results in a quarter. A company that just raised has a plan it told somebody about. A company that just migrated its site has a list of things that broke. Each of those is a real reason for a conversation to happen this month rather than someday.

The best second-order signals are specific to what you sell, which is why generic buying-intent lists rarely work well. For my own practice, a company publishing a lot of new content that is invisible to answer engines is a far stronger signal than that company being in software and having fifty employees. One describes a problem. The other describes a demographic.

How do you find second-order signals without buying a data tool?

By looking at what companies publish about themselves, which is more than most people assume. Job postings tell you what a company decided to invest in. Public changelogs and blogs tell you what they shipped and what they are pushing. Their own pages tell you when positioning changed, if you saw the old version.

The manual version is not glamorous and it works. Pick twenty companies that fit your first-order filter, check each one for the three or four things you decided count as signals, and note which ones show movement. That is an hour of work for a list of twenty genuinely warm accounts, which is a better use of an hour than sending three hundred untargeted emails.

Tools like Clay and Apollo exist for the part of this that scales, and they are worth reaching for once you know which signals matter. What they cannot do is decide what counts as a signal for your business. If you automate before you have that answer, you are just producing the same weak list faster. I have written about finding the first ten customers, and the manual approach is not a temporary stage you graduate from. It is where the signal definitions come from.

How many signals should an ICP actually have?

Three or four second-order signals, sitting on top of two or three first-order filters. More than that and nobody can hold it in their head while looking at an account, which means it will not get used. Fewer and you are back to guessing.

The structure I would write it in is a sentence rather than a table. Something like: a software company between twenty and two hundred people, selling to businesses, that has recently hired a marketing lead, changed its pricing page, or launched a product it needs to explain. That sentence is usable during a live conversation, which is the real test of whether an ICP document is doing its job.

Keep the signals independent of each other. If three of your four signals are all downstream of the same event, you have one signal wearing four hats, and your list will be narrower than you think. The point of multiple signals is that different companies reveal motion in different places.

How do you know a signal is real and not a coincidence?

Check it backwards against deals you already have. Take your last ten customers, including the ones who said no late, and ask what was happening at each company in the month before the conversation started. If a signal shows up in seven, it is worth building on. If it shows up in two, you invented it.

This is the step almost everyone skips, and it is the only one that separates a real profile from a plausible one. It also has the advantage of costing nothing, because the data is your own history rather than a purchased list. If you have too few deals to do this, you are not ready for a formal ICP and you should be having more conversations instead.

The same exercise works even better on losses. Companies that matched your profile, took meetings, and then went quiet are telling you that your signals select for interest rather than for need. That distinction is expensive to learn slowly. I have written about running win and loss interviews, and the ICP is one of the first things those conversations should change.

What do you do about accounts that match on paper but never move?

Stop working them and stop deleting them. They are not bad accounts, they are accounts without a trigger yet, and the right response is to put them somewhere you will look again rather than to keep sending them messages that have no reason to exist.

A simple watch list is enough. A sheet of companies that pass your first-order filter, revisited monthly against your signal list. Most of them will show nothing for a long time and then one will hire a head of growth, and that is the day the account becomes real. The cost of maintaining that list is far lower than the cost of either chasing or forgetting.

What you should not do is lower your standards to make the list feel productive. The temptation, when a quarter is quiet, is to widen the profile until there are enough names in it. That produces activity and not pipeline, and it teaches you the wrong lesson about your market because the failures are all explained by bad targeting you chose deliberately.

How does this change what you say when you reach out?

It gives you a first sentence that could only be written to that company. Not a compliment about their website, but an observation about the thing that just changed and what it usually creates. That is the difference between a message that feels researched and one that feels generated, and buyers can tell instantly.

It also changes the shape of what you offer. If the signal is a new marketing hire, the useful thing is probably something that makes them look competent in their first quarter. If the signal is a migration, the useful thing is a specific list of what breaks. The signal tells you not just who to talk to but what would actually be welcome.

I sell from Bengaluru to buyers in other countries, with fixed-fee pricing and most projects between one thousand and ten thousand dollars. At that size there is no room for a long nurture sequence, so the first message has to earn the conversation on its own. Signals are what make that possible, and they matter more the smaller your deal size is. They also change when it makes sense to stop doing this yourself, which I have covered in deciding when founder-led sales should end.

What should you do next?

Pull your last ten opportunities, won and lost, and write down what had changed at each company in the month before you spoke. Do not theorise first. Just write what was actually happening, and then look for the repeats. Whatever appears three or more times is the beginning of a real signal list.

Then rewrite your ICP as one sentence containing both halves: the filter that says who could buy, and the signals that say who is in motion right now. Test it by taking twenty companies through it by hand this week. If the sentence is hard to apply to a real company in under a minute, it is still too abstract and needs another pass.

If you are staring at a list that looks right and produces nothing, that is usually a signal problem rather than a messaging problem, and it is worth diagnosing before you rewrite another email. Reach out if you want to talk it through.