Fifteen customers, and no two rows the same
What follows is drawn from client work and de-identified. It was a company of three people, a quarter of a million in turnover, profitable, no investors and no debt. She opened by telling me that an ICP was going to be difficult.
If she laid her customers out in a table, one row each, with columns for industry, business and point of contact, every row would be different.
Fifteen customers, spread across transport, food processing, motor retail, chemicals, hospitality and clothing. The largest group in any single industry was two.
"Everyone is uniquely different," she said. "There's no two customers that are the same."
So I asked the question I ask next. If you were to describe your ideal customer to me, what would that look like?
She answered it in one sentence, without pausing.
An ideal client is somebody willing to spend on ESG compliance in the next three to six months, because they either have a customer demanding it or a regulator demanding it.
That is a usable ICP. It is more precise than most companies ever write down. And it contains no industry, no company size, no job title and no revenue band.
The columns were the reason it felt impossible.
Why does my ICP feel too broad?
A profile feels broad when it stops ruling anything out.
One founder posted to ask how other people had defined theirs, because their own felt too broad.
They had started where everyone starts. Company size bands, until they noticed that "a 50-person fintech and a 50-person agency are totaly different beasts".
Then they went further, and "even built an ideal customer profile template doc with all the standard fields - firmographics, technographics, pain points, whatever - but it still feels like we're guessing".
Every field the template asked for was filled in. The finished document still could not say who to call on Monday.
Firmographics and demographics describe a company accurately. However, two companies can match on size, sector, funding stage and job title while one is three weeks from signing and the other will never buy in this category at all.
What makes it even harder is that the most-used metric for judging ICP fit, response rate to outreach, doesn't accurately measure it. A poor response rate does not prove the profile is the broken part.
A dead channel is a result for three things at once: the offer, the profile, and the channel. A poor response rate, therefore, only means that the existing combination of these 3 variables doesn't work.
Know which of the three you are changing.
The dimension your template leaves blank is the one that predicts a purchase
Our name for a profile built properly is ICP ArchitectureICP Architecture Building targeting precision across five dimensions: Firmographic (company attributes), Demographic (individual attributes), Psychographic (mindset), Behavioral (intent signals), Temporal (timing triggers)., and it has five dimensions: firmographic, demographic, psychographic, behavioural and temporal.
Almost every profile I read is complete on the first two and empty on the last three.
The first two are the ones you can fill in from a database, alone, in an afternoon. Company size, sector, funding stage, job title, seniority: all of it is purchasable, and none of it requires you to have spoken to a customer.
The last three cannot be bought. Psychographic covers what the buyer believes and is accountable for. Behavioural covers what they have actually done that indicates intent. Temporal covers what has changed recently enough to open a budget.
Buying signals and intent data are the industry's names for the fourth of those, and both are sold as products. Deciding which signal means somebody can now spend money in your category is the part nobody can sell you.
Go back to that customer table. Her sentence carried all three and skipped the first two entirely. Willing to spend is psychographic. In the next three to six months is temporal. A customer or a regulator demanding it is the trigger event that makes the timing real.
Getting this wrong is a PositionPosition The architecture phase, holding strategy and execution together: the ideal customer profile defined to the level of title, company stage, industry and trigger event, the message mapped to a problem the buyer is accountable for, the channels where that buyer congregates, and the live execution carrying the message through them. The gate clears when an ICP-matched prospect raises a qualified hand. Primary failure mode: volume into broken positioning. problem, the earliest of the five gates our revenue-debt whitepaper measures, and that is why it surfaces so late. Nothing goes visibly wrong at the point the profile is written. It goes wrong later, in a pipeline nobody can close.
The trigger event is the part that gets dropped, because it is the only part you cannot buy. You learn it by asking customers what was happening the quarter before they signed.
An unfinished profile does not stay a documentation problem. Imprecise targeting has a price: the gap between what you actually pay to acquire a customer and what you would pay if every lead matched your profile.
It is paid in sales hours long before it reaches a media budget, and a review of the media budget will never find it.
Eight clients is a target worth £40,000 or £400,000
The size of a revenue target is set by which accounts a company chooses.
Those contracts ran from £5,000 a year to £50,000 a year. The average was £20,000. Every one was a three to five year deal.
Her goal for the rest of the year was eight new clients. At the average, that is £160,000.
Now put the spread back in. Eight clients at the bottom of her range is £40,000. Eight at the top is £400,000.
The goal she had written down was the same number in both cases.
Now attach the term length.
The difference between a £5,000 client and a £20,000 client is £15,000 a year. On a three year contract that is £45,000 of committed revenue. It comes from one decision about who to approach, taken before anybody wrote a word of copy.
Eight of those decisions is £360,000 over the term.
Eight signatures fill eight slots on the same three to five year terms, whether each one is worth £5,000 or £20,000. The £360,000 is not deferred to next year. The contracts that took those slots run for years, and you do not get to re-decide inside them.
This is the number to redo with your own book. Take your customer list, put your annual contract value next to each name, and sort it. If the spread is wide, your ICP is deciding a larger number than your pipeline target does.
What a finished ICP has in it
A finished ideal customer profile passes The List TestThe List Test Validation test for ICP specificity: Can a researcher build a list of 100 target accounts using only your ICP documentation without asking clarifying questions? If not, ICP isn't specific enough.: can a researcher build a list of 100 target accounts using only your ICP documentation, without asking clarifying questions?
Most profiles fail it on the first account.
A profile that passes carries five fields, and only the first is on the standard template.
- The attributes. Ranges a stranger can filter on: a revenue band, a headcount band, a sector list, and the geographies you can actually serve. "Mid-market" is not a range.
- The trigger, with its window. What has to have happened, and how recently. New regulation, a funding round, a named executive hire, a contract renewal date, a customer of theirs asking them for something. Ninety days after a funding round is a different account from eighteen months after it.
- The disqualifiers. The attributes that look like a match and are not. Ours includes companies where the person with the problem reports to the person with the budget through a third function, because those deals take a year and we are not built to carry them.
- The two people. Who has to agree, and who signs.
- The channel expression. The profile written in the targeting options of the channel you will actually use.
The trigger is the field that turns a list into a queue. Without it you have a set of companies; with it you have an order to work them in.
The two people is the field that decides whether the deal closes. Most profiles name a job title and stop there.
A job title hides whether this person can say yes on their own. Someone can own the budget line and sign within their threshold and still be unable to commit, because the purchase is a question about direction rather than cost, and direction is somebody else's to settle.
So the qualifying question is two questions: whose agreement would this need, and who signs.
On your last three wins, write down who was in the room at the decision and who was not. Then ask the same of every open opportunity above your average deal size.
Where nobody can name the signer, you have found a targeting gap that is already costing money.
You learn the disqualifiers from losses rather than wins, which is why almost nobody has them.
And the channel expression is where good profiles die. An advertiser for an immigration practice found that their profile "pretty much doesn't exist in Facebook's interests".
The nearest the platform would allow was newlywed within three or six months, on "less than 100k" of available audience. The profile was right and the channel could not express it.
Should I niche down, or reach everybody?
Byron Sharp makes the serious case against narrow targeting. In How Brands Grow, he and the Ehrenberg-Bass Institute argue that brands grow by reaching all the buyers in a category rather than a chosen segment of them, and that much targeting buys less reach at a higher price.
We answer that case at length in Lead generation, because it is really an argument about how to count demand.
Sharp is talking about who should see the message. An ICP decides who gets a seller's afternoon, a bespoke proposal, a second meeting. Those are different budgets, and only one of them is cheap.
Reach is close to unlimited. Selling days are not. A year holds about 261 weekdays; take off eight public holidays and the leave your team actually takes, and one person is left with something over two hundred. You cannot buy one more.
So an ICP is a rationing decision.
A solo founder with "25 paying customers, ~$5k MRR" asks "Niche down or stay horizontal?". The real question is which customers get their own week.
The expensive part is reversing your own decisions
Rewriting an ideal customer profile is a morning's work that you can do without help. Sorting a customer list by contract value takes ten minutes.
The difficulty sits somewhere else. You are the person who chose the current profile.
The disqualifiers are a list of your own decisions to reverse. The trigger event is learnable only by going back to customers and asking what was going on before they bought. Often that means hearing that the reason was never the one in your pitch.
And the accounts the finished profile rules out are usually accounts someone on your team is currently working. Narrowing an ICP is a conversation about a named opportunity in the pipeline, not a slide.
You are paying someone else to carry that difficulty.