How do you find your ideal customer profile as an early-stage startup?

Your ideal customer profile isn't something you define in a spreadsheet before you talk to anyone — it's something you discover by getting uncomfortably close to a very small number of real people. The fastest path is to pick one user, serve them obsessively, and let the pattern that emerges tell you who your product is actually for. Everything else is hypothesis until contact with reality.

Start with one user, not a segment

Most founders try to define their ICP the way a McKinsey consultant would: demographics, firmographics, job titles, pain scores on a 1–10 scale. That approach assumes you already know enough to categorize. You don't — not yet. The more honest method is to find a single person who has the problem you're solving, who feels it acutely enough to act on it, and then treat that person as your entire market for a while.

Paul Graham's advice for B2B founders is concrete here: pick one user and work as if you're a consultant building exclusively for them. The goal isn't to close a deal or validate a hypothesis — it's to fit their needs so precisely that the product clicks. When it clicks for one real person with real constraints and real workflows, you've learned something that no amount of user-persona workshops could teach you.

This approach feels wrong because it's not scalable and it's not 'professional.' But that discomfort is exactly the point. Early startups don't need scalable customer acquisition strategies — they need a working product that someone urgently wants. One deeply satisfied user teaches you more about your ICP than a hundred survey responses.

Narrow the market until it's almost uncomfortably small

One of the most counterintuitive moves in early customer discovery is deliberately constraining who you serve. Facebook launched only for Harvard students. Not 'college students,' not 'young adults who want to connect online' — Harvard students specifically. That specificity made early users feel the product was built for them, not just available to them. That feeling of belonging is what drives the critical mass that lets you expand later.

The logic here is thermal: concentrate your energy in a small space to generate real heat before you try to spread warmth everywhere. A startup that tries to serve everyone in a broad market gets lukewarm traction across the board. A startup that dominates a narrow slice — geographically, by industry, by use case, by community — gets the kind of fanatical early adopters who recruit others and forgive rough edges.

When narrowing your market, ask: which subset of this broader group would suffer most if our product disappeared tomorrow? That's your beachhead. It's usually smaller than feels comfortable, and it's almost always more specific than your current targeting. The ICP you'll eventually write down in a sales playbook should be derived from this group — not invented before you meet them.

Go where the feedback is raw and unfiltered

The quality of the signal you get from a customer degrades with every layer you add between yourself and them. A founder sitting in a user's office watching them struggle with a workflow learns something irreplaceable. A founder reading a Typeform response learns almost nothing by comparison. This isn't just a nice principle — it's operationally important for ICP discovery, because the things that reveal who your ideal customer is are often the things users don't know how to articulate.

You're looking for friction, workarounds, and moments of genuine relief. Which users light up when they see a particular feature? Which ones adapt your product to uses you didn't design for? Which ones complain about the exact same thing in the exact same words? Those patterns, visible only through direct observation and conversation, are what a real ICP is built from. You're not surveying for satisfaction — you're watching for the signals that tell you 'this person's life is genuinely different because of what we built.'

Paul Graham's observation about direct user engagement is pointed: the feedback you get from your earliest users is the best you'll ever receive. Treasure that access. Large companies pay enormous sums to approximate what you can get for free by showing up at someone's office with a laptop and a genuine desire to understand how they work.

Let the users who pull hardest define the profile

Once you've done the unscalable work — the site visits, the hand-holding, the custom setup calls — you'll start to notice that some users adopt your product eagerly while others require constant pushing. This difference is enormously informative. The ones who pull hardest, who integrate your product into their workflow without being asked, who email you unprompted with ideas and referrals — those are your early ICP candidates.

Don't average them with the users who are lukewarm. Averaging is how you end up building for nobody. Instead, study the enthusiasts obsessively: what's their job title, their company size, their industry, their specific workflow, their existing tools, their internal organizational structure? Look for what they share. That intersection is the beginning of a real ICP — one grounded in behavior, not assumption.

Garry Tan has noted that one of the traps early founders fall into is mimicking big-company indifference to individual users in an effort to seem professional. The founders who discover their ICP fastest do the opposite: they're almost embarrassingly attentive. They know their first ten users by name, by company, by the specific problem that brought them to the product. That intimacy isn't a phase to move past quickly — it's the foundation everything else is built on.

Build what you would urgently want, then find others like you

There's a practical shortcut to ICP discovery that doesn't get enough credit: build something you yourself urgently need. When you are the user, your feedback loop collapses from weeks to seconds. You know immediately when something is wrong because it's wrong for you. You know when it's right because you actually want to use it.

This isn't just a productivity hack. It gives you immediate access to a social network of potential early customers — the people in your professional circle who share your problem. If you've built a tool that solves your own workflow problem, there's a reasonable chance your former colleagues, co-workers in adjacent roles, or people in your professional community have the same problem. Those people are your first outreach list. They trust you enough to give honest feedback, and they're motivated to engage because the pain is real for them too.

From this starting point, the ICP emerges organically: you learn which of your contacts respond urgently versus politely, which ones recruit others without being asked, and which ones push back on assumptions you didn't know you were making. That's how a rough profile sharpens into something you can eventually operationalize — not from a template, but from accumulated contact with people who genuinely needed what you built.

“Pick a single user and act as if they were consultants building something just for that one user.”

— Paul Graham, source

The one thing to do

This week, find one person who has the problem you're solving, sit with them while they work, and let what you observe rewrite your assumptions about who your product is actually for.

Frequently asked questions

How many customers do I need to talk to before defining my ICP?

There's no magic number, but depth matters more than breadth at the earliest stage. Sitting with five users in their actual work environment will teach you more than surveying fifty. Start with one, go deep, then look for patterns as you add more.

What if my product could serve many different types of customers?

That's exactly the situation where you need to narrow deliberately. Pick the segment that has the sharpest pain and the most urgency, serve them exclusively for a while, and expand from a position of strength rather than diluting your energy across everyone.

How do I know when my ICP is 'right'?

When customers in that profile adopt the product without heavy hand-holding, refer others unprompted, and complain about specific missing features rather than fundamental confusion about what the product does — that's a signal you've found real fit.

Should I write down my ICP before I start talking to users?

You can write a working hypothesis, but treat it as something to disprove rather than confirm. The ICP you discover through direct engagement will almost always be more specific and more accurate than the one you invented before talking to anyone.

Sources

More playbook answers · Growth Prophet home