Every ICP Guide Says "Analyze Your Customer Data." You Have Three Customers.

Here's how founders actually build an ideal customer profile and 2–3 buyer personas — as a guessing discipline, not a data exercise. And why the fictional persona with the stock photo is not useful.

A detail without an evidence is a decoration. Photo by Jessica Lewis/ Unsplash

Go read any guide on creating an ideal customer profile. I'll wait. They all give you the same first step: "Start by analyzing your existing customer base. Look at your best customers and find what they have in common."

Lovely. One small problem. You're a founder. You have three customers. One is your former colleague, one found you through a podcast, and one you're still not entirely sure why they bought. "Analyze the patterns in your customer base" is advice written for a company with 500 accounts and a RevOps team — which is exactly who those guides are written by, and for.

So let's do the honest version: how do you build an ICP and buyer personas when you don't have the data yet, why should you bother at all, and how do you keep the whole exercise from becoming the laminated fiction nobody ever looks at again?

I've done this with 500+ startups. Here's what actually works.

First: Why bother, when you have no customers to analyze?

Because the ICP isn't a summary of your past. At your stage, it's a discipline for your guessing.

Without an ICP, here's what founder-led sales looks like: you talk to anyone who'll take a meeting. The agency owner, the enterprise VP, the solo freelancer, the government contact your uncle knows. Every conversation is different, so nothing you learn compounds. Your pitch mutates weekly. Your product roadmap gets yanked in four directions by four completely different kinds of buyer. And six months in, you're exhausted and can't say what you've learned, because you were never running the same experiment twice.

An ICP — even a wrong one — fixes that. It's a hypothesis stated clearly enough to be tested: "I believe my best customer is X kind of company with Y problem." Now every sales conversation is a data point for or against the same hypothesis. Now the learning compounds. Now a "no" teaches you something, because you know what it's a "no" to.

That's the real reason to write it down. Not because the ICP will be right — yours won't be, at first, and I'll get to that — but because a written guess can be corrected, and a vibe cannot.

ICP vs. personas, in one line each

People tangle these up, so let's untangle them fast.

Your ICP describes the company: what kind of organization gets the most value from what you sell, and can actually pay for it. One ICP. Singular.

Your buyer personas describe the humans inside that company: the two or three specific people who have to say yes, and what each of them is afraid of. Two or three personas. Not seven.

The line I use with founders: the ICP tells you which door to knock on. The personas tell you who answers, and what to say to them. You need the door first. Personas without an ICP are characters in search of a story.

(If you're B2C, the distinction mostly collapses — your "company" and your "human" are the same person, so you're really just building one to three sharp personas. The method below still applies; skip the firmographics and go straight to the human.)

How to build your ICP with almost no data

You have more data than you think. It's just not sitting in a CRM — it's sitting in your memory and your inbox. Here's the sequence.

Start with the problem, not the demographics. The textbook ICP starts with industry, company size, revenue. Wrong order. Start with: whose hair is on fire? Write one sentence: "The companies that need this most are the ones struggling with ______, usually because ______." If you can't finish that sentence sharply, no amount of firmographic filtering will save you — you don't have an ICP problem, you have a positioning problem.

Mine the conversations you've already had. You've talked to dozens of people about your product, even pre-revenue — discovery calls, demo requests, the friends who asked good questions, the prospects who ghosted. Go back through them and sort them into two piles: the conversations where the other person leaned in (asked about pricing, asked "how soon," described their pain unprompted) and the ones where they were politely interested. Now describe the lean-in pile: what did those companies have in common? Size, stage, situation, trigger? That description is your first ICP draft — built from behavioral evidence, not spreadsheet data.

Add the "can they actually buy" filter. Loving your product is not the same as buying it. Your ICP needs the boring qualifiers: do they have budget for this category, is there someone empowered to decide, and is there a reason to act now rather than someday? A company with the perfect problem and no budget is not your ideal customer. It's your ideal fan. That's a different story.

Then — and only then — add the demographics. Industry, headcount, geography, funding stage. These aren't the ICP; they serve as a starting point to help you find other companies similar to those in your “Lean-In” stack. Keep it to three or four attributes. An ICP with eleven criteria isn't precise — it's misleading. No one will ever meet all of these criteria, and you'll use that as an excuse not to reach out. Write the whole thing in three sentences, not a slide deck. If it doesn't fit on an index card, it won't get used.

How to build 2–3 personas that don't become fiction

Now the humans. And here's where I want to avoid a tradition: the fictional persona. "Marketing Mary, 34, drives a Subaru, values work-life balance, enjoys podcasts." Complete with a stock photo. You've seen these. Maybe you've made one. Nobody has ever closed a deal because they knew Mary's favorite podcast.

The problem with fictional personas isn't that they're detailed. It's that they're invented — a creative-writing exercise that feels like customer research but contains no information. A detail without an evidence is a decoration.

Here's the founder's version instead: base each persona on a real person you have actually spoken to.

Not a composite, and not an archetype. Choose a real human — with the name changed if you like — whose words you can quote. For most B2B startups, you need two or three, and they usually map to roles like these:

The person in pain. The one who lives with the problem daily and will use your product. What you need to know about them: how they describe the problem in their own words (steal those words — they're your best sales copy), what they've already tried, and what would make their Tuesday better. This persona is built from your best discovery conversations. Quote them.

The person with the budget. Often not the same human. They don't feel the pain directly — they feel the cost of it, or they don't. What you need to know: what they're measured on, what makes a purchase feel safe to them, and what would make approving this look smart to their boss. If you've never talked to one of these people, that's not a persona gap — that's a sales-process gap. Go have that conversation.

The person who can ruin the whole deal. The skeptic, the technical evaluator, the "Dave" who shows up in the final meeting. What you need to know: what they're protecting (their time, their systems, their reputation), and what would turn them from a blocker into a neutral. You usually meet this person late. Write down what the last one said.

For each persona, you need maybe five lines: who they are, what they're afraid of, how they describe the problem, what a win looks like for them personally, and one real quote. Five lines you'll actually reread before a call beats five pages nobody opens after the workshop.

The part every guide skips: your first ICP is wrong

Here's the thing I most want you to internalize, and it's the same thing I tell founders about their first ten customers: the Day 1 ICP is a guess. The post-ten-customers ICP is data. The gap between them is the whole point.

You will write an ICP, sell against it, and discover that your actual best customers are slightly — or wildly — different from your hypothesis. The companies you thought were perfect will stall in procurement. Some segment you dismissed will start pulling your product out of your hands. This is not the ICP failing. This is the ICP working — it's the experiment producing a result.

So plan the revision from the start. Every quarter, look at your real pipeline and ask: who actually bought fastest, paid happily, and got value? Rewrite the index card to match reality.

Reject your ICP twice before you trust it.

The founders who fail aren't the ones who write a flawed ICP. They're the ones who write it once, then set it in stone, and defend it against all evidence.

Write the guess down. Test it in every conversation. Correct it without sentimentality. That's the entire method, and no template on the internet does the work for you.

An ICP at the founder stage isn't a data-analysis exercise — you don't have the data. It's a guessing discipline: a written hypothesis about which company gets the most value from you, sharp enough to be tested in every sales conversation. Build it from the conversations you've already had (find the lean-in pile), add the "can they actually buy" filter, keep it to three sentences. Then build 2–3 personas based on real people you've spoken to — the person in pain, the person with the budget, the person who can ruin the whole deal — five lines each, with real quotes, no stock photos. And expect to be wrong: the Day 1 ICP is a guess, the post-ten-customers ICP is data. Revise quarterly. A written guess can be corrected. A vibe cannot.

Not sure who your ideal customer actually is — or suspect the one you wrote down six months ago is wrong? That's one of the most common things I work through with founders. Book a free 30-minute strategy call →