How to define your ICP (ideal customer profile)

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 Architecture, 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 Position 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 Test: 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.

1. 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.
2. 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.
3. 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.
4. The two people. Who has to agree, and who signs.
5. 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.

Run this

Work through this in one sitting, with your customer list open.

1. List every customer. Name, annual contract value, contract length, industry, how they first came to you.
2. Sort by contract value. Note the ratio between your largest and your smallest, and hold that ratio next to your target for new clients.
3. Take the top three. For each, write what was happening in their business in the ninety days before they signed. If you do not know, call them and ask. This is the trigger.
4. Look for the shared trigger, not the shared attribute. Two of your three may be in different sectors and have had the same thing happen to them.
5. Write your disqualifiers from your last five losses. For each loss, name the attribute that made it look like a fit.
6. Name the two people. For your last three wins: who had to agree, and who signed. Where those were different people, write down how you found that out.
7. Run The List Test. Hand the document to someone who does not work on sales and ask for 100 accounts. Count how many questions they have to ask you.

Thresholds worth holding:

• The List Test passes at zero clarifying questions. Every question your researcher has to ask you names a field the document is missing.
• Write down your largest-to-smallest contract ratio. The company in this chapter ran 10x, on contracts of £5,000 and £50,000. We publish no benchmark to compare it against, so read your own: the wider it is, the more of your revenue number your ICP is deciding for you.
• No trigger field means you have a description, not a profile.

Diagnostic questions:

• Can you name the event that preceded your last three wins?
• Can you name an account type you have decided not to pursue this year?
• Does your profile survive being typed into the targeting options of the channel you actually use?

Frequently asked questions

How do I define my ICP if all my customers are different?

Stop looking at what they are and look at what happened to them. Customers in unrelated industries frequently share a trigger: a regulatory deadline, a funding round, a new executive, a renewal date. Sort your customers by contract value, take the top three, and establish what changed in each business in the ninety days before it signed.

What is the difference between an ICP and a buyer persona?

An ICP describes the account you want: the company, its conditions and its timing. A persona describes a person inside it. You need the ICP first, because a persona attached to the wrong account type is a well-researched description of somebody who was never going to buy. Personas are covered in Why buyer personas fail, in the $0 to $100k Playbook.

Do I need TAM, SAM and SOM to define an ICP?

Market sizing and an ICP answer different questions. TAM, SAM and SOM size a market you might one day serve; an ICP decides which accounts you contact this quarter. The sizing exercise will not narrow anything, because every band it produces still contains companies that cannot buy yet. Do the profile first and size afterwards, if an investor asks.

How many customers do I need before I can define an ICP?

Three closed deals is enough to start, because you are looking for a repeated trigger rather than a statistically sound sample. Write the profile from those three, mark it as provisional, and revise it after the next five. Waiting for a large sample means spending a year targeting on assumptions instead.

What if our product really is for everyone?

A product can be usable by everyone and still never be bought by everyone at once. An ICP answers a narrower question: who has a reason to act this quarter, and the authority to pay for it. Sell to that group first. Breadth of use is an argument for broad reach, never for undirected spend.

What is the Targeting Tax?

The Targeting Tax is 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 wasted sales time as well as media spend. It appears as a line item nowhere, so a budget review will never find it.

Who should own the ICP, sales or marketing?

Whoever is accountable for the pipeline number should own it, because the ICP decides that number before either team touches it. In practice it is written jointly and signed off by one person. Where the two functions hold different profiles, the symptom is sales calling the leads unqualified, which is covered in Marketing and sales alignment.

Who is the economic buyer, and how do I find them?

The economic buyer is whoever can commit the spend without needing anyone else's agreement, which is often not the person who owns the budget line. Find them by looking backwards: on your last three wins, write down who was in the room when the decision was made and who was not. The name that keeps appearing is the one your profile should carry.

Key frameworks

ICP Architecture. Building targeting precision across five dimensions: Firmographic (company attributes), Demographic (individual attributes), Psychographic (mindset), Behavioral (intent signals), Temporal (timing triggers).

The 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.

The Targeting Tax. The cost premium paid for imprecise targeting. Calculated as actual CAC minus theoretical CAC with 100% ICP-match leads.

Economic Authority. The power to commit spend without needing anyone else's agreement. It is identified by neither job title nor budget line alone: controlling a budget is not the same as having uncontested authority to spend it. Where the two come apart, the person whose agreement is actually required holds Economic Authority, whatever the org chart says.

Position. 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.

Sources

1. Founder asking publicly how others had defined their ICP, because their own felt too broad, July 2026. Quoted verbatim, including the source's own spelling. https://www.reddit.com/r/SaaSSales/comments/1v8p8nk/how_did_you_define_your_icp_ours_feels_too_broad/
2. Advertiser for an immigration practice, on a profile that no targeting option could express, August 2026. https://www.reddit.com/r/PPC/comments/1vlesrd/for_an_immigration_law_firm_is_broad_targeting/
3. Solo founder with twenty-five paying customers, asking whether to narrow, August 2026. https://www.reddit.com/r/SaaS/comments/1vm06gt/25_paying_customers_5k_mrr_niche_down_or_stay/
4. Practitioner on why a channel result is only ever a result for one combination of offer, profile and channel, August 2026. https://www.reddit.com/r/SaaS/comments/1vhudsq/we_tested_this_channel_doesnt_work_for_us/
5. Sharp, B. How Brands Grow: What Marketers Don't Know, Ehrenberg-Bass Institute for Marketing Science, Oxford University Press.
6. PacedRevenue, "The Revenue Debt You Cannot See" (v1.0, 29 May 2026; author revision 21 August 2026): the Position gate definition, its trigger-event clause and its failure mode. https://pacedrevenue.com/whitepapers/revenue-debt-you-cannot-see/
7. PacedRevenue framework registry: ICP Architecture, The List Test, The Targeting Tax and Economic Authority, definitions as held for the glossary and schema.

Hassaan Ahmad, Managing Partner and Chief Revenue Officer. Published 4 September 2026.

Related reading: Lead generation · Marketing and sales alignment · Full funnel optimization

Start with the free read, or go straight to the scored one. The PACED diagnostic is free and names the gate most likely to be binding. The PACED Review scores every gate of your engine against your real numbers: £2,499, and it stands alone whether or not you go further.