How to build an ICP that actually converts

An ICP is not a market segment

Most ICPs I see are useless, and the reason is always the same. They are too broad. "Mid-market B2B SaaS companies in Europe" is not an ICP. It is a market. It does not tell a rep which company to open, what to say, or why today rather than next quarter.

A real ICP does one job. It points at a specific set of accounts and gives you a reason to reach each one. If your ICP does not exclude anyone, it is not helping anyone. The whole value is in what you leave out.

I have booked over 2000 meetings for B2B clients, and the ones that worked all started here. Not with a tool, not with copy, with a tight definition of who and why. Everything downstream, the list, the offer, the email, inherits its quality from the ICP. A weak ICP poisons all of it.

The right 200 beat the wrong 5000

Founders treat list size as a strength. It is the opposite. Five thousand loosely-relevant accounts is a liability. You cannot personalise to them, you burn your domain sending to them, and most replies you get are the wrong people wasting your time.

Two hundred accounts that genuinely fit, where you know the buyer and the trigger, will out-book the five thousand every time. Fewer accounts means you can actually look at each one. You can find a real reason to reach out. You can write something only that company would receive. Targeting beats volume, and the ICP is where you make targeting happen.

Start from your best existing customers

The single best source of truth for your ICP is already in your business. It is your best customers. Not all of them, the best ones. The accounts that closed fast, paid without friction, got value quickly, stayed, and referred you.

Pull your last ten to twenty closed-won deals and look at them honestly. What do the good ones share that the bad ones do not? Do not build the ICP from who you wish you sold to. Build it from who actually buys and sticks. Assumptions feel good and lie to you. Closed-won data does not.

Look at the losses too

Your closed-lost and churned accounts are just as useful. They tell you who looks like a fit on paper but never converts, or signs and leaves three months later. Those patterns become your negative indicators, the red flags that disqualify an account even when the firmographics match. A good ICP says who to chase and who to skip.

Layer one: firmographics

Firmographics are the static facts about a company. They are the floor of your ICP, not the ceiling. They narrow the universe to companies that could plausibly buy, but on their own they never tell you that an account is ready now.

Keep this layer tight. The four that matter most:

• Industry or vertical. Be specific. "Software" is too broad. "Ecommerce brands selling physical products" is a target.
• Company size. Headcount or revenue band. The size where your product fits and the buyer has budget but not so much process that nothing moves.
• Region. Where you can sell, support, and speak the language and the buying norms.
• Tech stack. Technographics. What they run often predicts whether you are relevant. A detected tool is a fact, not a guess.

If you cannot describe your ICP's firmographics in two or three sentences, it is still too broad. Cut until it hurts a little.

Layer two: the buyer, not just the company

Companies do not reply to emails, people do. So the ICP has to name the human. A company can be a perfect fit and you still go nowhere because you wrote to the wrong person.

Name the role you are selling to, and be honest that there is usually more than one. There is the person who feels the pain day to day, the person who signs, and sometimes a blocker who can kill it. For outbound, you mostly care about the one who feels the pain, because they reply.

Pain personas and value personas read differently

The same truth lands two ways depending on who reads it. An operator, a Head of Ecommerce or a VP of Sales, responds to what they are actively losing right now. A leadership persona, a CMO or a founder, responds to what they could unlock. Same product, different frame. Your ICP should note which persona you are writing to, because it changes every line of copy after it.

Layer three: buying signals, the part most people skip

This is where ordinary ICPs become ICPs that convert. Firmographics tell you who could buy. A signal tells you why they might buy now. A signal is something observable happening at the account that makes your offer relevant this week instead of someday.

Most ICP models fail not because they pick the wrong accounts, but because they have no sense of timing. They find good-fit companies and contact them all at random moments. Signals fix the timing. They tell you which of your good-fit accounts to call first, today.

Not all signals are equal

Some signals are strong enough to open an email with. Some are weak background context. Here is roughly how I rank them, and what each one tells you.

The rule is simple. Always reach out on the strongest signal you actually have for that account. If all you have is firmographics, the account is on the list but it is not ready to send yet. Go find a reason first.

Score the accounts, then sort

Once you have the layers, you need a way to rank. You do not need a complicated model. You need a simple, consistent score so you contact the best accounts first instead of working the list in the order it was scraped.

I think of it as three multipliers. Fit, how well the firmographics and persona match. Signal, how strong and recent the trigger is. Strategic value, how much this account is worth if it lands. Multiply, not add, because a zero on fit should kill the score no matter how loud the signal is. Sort the list by that number and work top down.

Narrow until it feels too narrow

Every instinct tells you to widen the ICP so you do not miss anyone. Resist it. The broader the ICP, the more generic the messaging, the more wasteful the targeting, and the less your outreach knows who it is talking to.

A narrow ICP feels risky and is actually safer. You can personalise. You can find the trigger. You protect your domain because you only send to accounts worth sending to. If you are worried two hundred is too few, remember that two hundred well-chosen accounts you can speak to precisely will beat thousands you can only spam. You can always expand later from a position of knowing what works.

Weak ICP vs sharp ICP, side by side

The mistakes that quietly kill an ICP

Most bad ICPs fail in one of a few predictable ways. Watch for these.

• Too broad. The number one killer. If it excludes no one, it guides no one.
• Firmographics only. Static fit with no signal means you have a list with no sense of timing.
• Built from assumptions. Not validated against who actually closes and stays.
• No buyer named. A perfect-fit company and you still wrote to the wrong person.
• Never updated. Built once, then frozen while your market and your wins move on.

The ICP is the input to everything else

Here is why this matters more than any subject line. The ICP feeds the list. The list feeds the offer, because the offer has to solve the specific pain of that specific buyer. The offer feeds the copy, because the first line is the signal you targeted in the ICP. Reporting only makes sense against the segment the ICP defined.

Get the ICP right and the rest of the system has a chance. Get it wrong and you are doing excellent work aimed at the wrong people, which is just expensive noise. This is the highest-leverage decision in the whole motion, and it takes hours, not weeks. Targeting beats volume, and the ICP is where targeting is decided.

Keep the ICP alive

An ICP is not a document you write once and file. It is a living definition that gets sharper as replies come in. The accounts that book, the angles that land, the signals that convert, all of that feedback should flow back into the ICP. Every campaign teaches you something about who really buys. The teams that win are the ones that tighten the definition every month instead of defending the one they wrote on day one.

How I build an ICP

I start from your best customers and the patterns in your closed-won data. I set tight firmographics, name the exact buyer and the frame they respond to, and build the list around observable buying signals so every account carries a reason to reach out now. I score and sort so the best accounts get contacted first, write negative indicators so we skip the traps, and feed replies back in so the definition keeps sharpening. Then it becomes the input to the offer, the copy, and all-bound outreach across email, LinkedIn and calls. The whole thing is documented, and it stays yours.

FAQ