Every piece of sales advice starts the same way: “define your ideal customer.” Cool. You can’t. You built something useful, you have a vague sense of who might want it, but if someone made you write down exactly who your customer is, you’d stall on the first line. That’s not a failure. For a new product, if you don’t know your ICP yet, that’s the normal starting point, and the usual advice abandons you right there.
So let’s do the opposite of the usual advice. Instead of defining your customer and then going to find them, you’re going to find people first and let them tell you who your customer is. Here’s how to find customers when you don’t know your ICP yet.
Why you don’t know your ICP yet, and why that’s fine
The clean ICP doc, with firmographics and personas and a tidy little diagram, is something you write at the end, not the start. Early on it’s fiction. You’re guessing. And acting confidently on a guess sends you chasing the wrong people for months.
The truth is that your real customer is hiding in the gap between who you think wants this and who actually pulls out a card. The only way to close that gap is contact with real people. So the goal of your first few weeks isn’t to sell. It’s to learn who leans in.
Start from the problem, not the person
You may not know who your customer is, but you know what problem you solve. Start there. Write down the problem in the exact words a frustrated human would use, not your marketing words. “I waste an hour every Monday copying numbers between spreadsheets.” “My team keeps missing replies because they’re scattered across four inboxes.”
Now go searching for those sentences. Reddit, X, Indie Hackers, support forums, review sites for the clunky tool you replace. You’re not looking for a demographic. You’re looking for anyone, anywhere, saying that sentence out loud. The people who show up are your first clue. Notice what they have in common after you’ve found twenty of them, not before.
Look at who already pays attention
If you have even a handful of signups or a few people who replied to a post, you’re sitting on data. Who actually engaged? Not the polite “looks cool” comments. Who asked a follow-up question, who tried the thing twice, who emailed you unprompted? Those people are leaning in, and leaning in is the only signal that matters this early.
Look for the pattern in the leaners. Maybe they’re all solo founders. Maybe they all run agencies with three to ten people. Maybe they’re all in one industry you never expected. That pattern, drawn from people who actually responded, is a far better ICP than anything you could have invented at your desk.
Have ten real conversations
Reach out to ten people who fit the loose pattern and just talk to them. Not a pitch, a conversation. Ask how they deal with the problem today, what they’ve tried, what made them give up on the last tool. You’re trying to hear the same frustration in their own words, again and again, until you can predict what they’ll say.
When you start finishing their sentences, you’ve found your ICP. Not because you defined it, but because you met enough of them to recognize the next one on sight. That recognition is worth more than any persona slide, because it’s built from real people instead of your assumptions about them.
Then write it down, lightly
Once you can recognize your customer, write a one-line version. “Solo founders who just launched a paid SaaS and have zero sales process.” Keep it loose enough to be wrong, because it will be, and you’ll revise it three more times. The doc isn’t the point. The point is that now, when you go hunting tomorrow, you know which complaints are worth your twenty minutes and which to scroll past.
This is the slow part of the work, and it’s the part we tried to make less lonely. Unbound Compute watches for the specific complaints you’re chasing and surfaces the people saying them, so you can spend your time having the ten conversations instead of hunting for who to have them with. You’ll still figure out your ICP the only way anyone does: by talking to it.