Your First Ten Customers: Where Do You Actually Find Them?

Where do you actually find your first ten customers?

In conversations you have personally, with people you can already reach, about a problem you can describe better than they can. Not in a channel. The first ten customers are not a marketing problem, and treating them as one is the most common reason a good product spends a year with three users and a content calendar.

This is the least comfortable answer, which is why it keeps getting replaced by a plan involving ads, SEO, and a launch date. Those things work later. They do not work at ten, because none of them can tell you the thing you most need to know, which is whether anyone will pay.

Here is how I would go about it, what to say, and how to tell when you have actually found a channel rather than a coincidence.

Why is this a sales problem rather than a marketing problem?

Because marketing scales a message that already works, and at ten customers you do not have one yet. Every channel you could build is a machine for repeating something. If the something is wrong, a better machine just distributes the wrongness faster and charges you for it.

Paul Graham made this argument directly in his essay Do Things that Don't Scale, where he writes that you cannot wait for users to come to you, and that you have to go out and get them. That essay is about startups, but the logic applies to anyone selling something new, including a service business, including mine.

He also warns that big launches and partnerships usually do not work for generating early growth. I would put it more bluntly: a launch is a way of finding out whether people who already know you will act, which is useful information, and it is not a customer acquisition strategy.

Who should you talk to first?

The people closest to the problem who are also closest to you. Concentric circles, worked outward: people you have worked with, people who have asked you about this exact thing before, people one introduction away, then strangers who have publicly described the problem.

Start with the ones who have already complained about the problem in front of you. A person who has said the thing out loud is worth thirty people who match your demographic filter on paper, because their existence proves the problem is felt rather than theorized. Go back through your inbox and your notes and find those people first.

The instinct to skip this circle because it feels like cheating is worth naming. It is not cheating. It is the only group where you get an honest answer quickly, and an honest no from someone who likes you is more valuable at this stage than a polite maybe from a stranger.

My own practice ran on this for years. Across 70 plus projects for 25 plus clients over 6 plus years, referral and direct conversation did more than any paid channel I tried, which is a pattern I wrote about in why referrals beat paid ads for a freelance practice.

What do you actually say in those conversations?

Ask about the last time the problem happened rather than whether they have the problem. What did you do, how long did it take, what did it cost you, what did you use instead. Past behavior is evidence. Stated interest is politeness.

Then, and this is the part people avoid, ask for money. Not in a slick way, just directly: here is what I would build or do, here is what it costs, would you pay for that. The conversation changes character the moment price enters it, and everything before that moment is unverified enthusiasm.

You are listening for three things: whether they have tried to solve it before, whether they have a budget line it could come from, and whether they can name the person who would have to approve it. A prospect who cannot answer the third question is not a prospect yet, they are research.

What does doing things that do not scale look like in practice?

It looks like doing the customer's work for them. Graham's essay describes the Collison installation, named after Stripe's founders, who when someone agreed to try the product would say give me your laptop and set it up right there rather than sending a link. That is the whole idea in one gesture.

The same essay points at Airbnb's founders going door to door in New York, recruiting new users and helping existing ones improve their listings, and at Wufoo sending handwritten thank-you notes to every new user. None of these are growth tactics. They are ways of removing every possible reason for a person to not start.

Translate that into your own situation and it usually means: do the onboarding yourself, do the migration yourself, sit on the call while they use it the first time. It feels unscalable because it is, and that is fine, because you are not trying to reach a thousand people. You are trying to make ten people succeed so completely that they will tell someone.

Which channels are worth touching at this stage?

The ones where the conversation is already happening and you can join it as a person rather than a brand. Communities where your buyer complains: a Slack group, a Discord server, a subreddit, a LinkedIn comment section, an industry forum, a WhatsApp group of operators. Show up with answers, not with a link.

Outbound email works at this stage precisely because you are sending ten of them rather than ten thousand, so each one can reference something real about that specific company. The thing that makes cold outreach work at ten is the thing that makes it impossible at scale, and that is the correct trade for now.

What I would not do yet is build. No ads, no long-form SEO program, no podcast, no newsletter. Every one of those is an investment in repeating a message, and you are still finding out what the message is. Graham's contained fire idea, illustrated in his essay by Facebook launching only at Harvard, is the better instinct: dominate a space small enough that word travels inside it without you.

How do you know when a channel is actually working?

When the same thing works twice with people who do not know each other. One customer from a subreddit is a story. Three customers from that subreddit, arriving through the same motion, is a channel. Until you have the second and third, you are looking at a coincidence with a narrative attached.

Measure conversations, not impressions. At this stage the only numbers that mean anything are how many real conversations you started, how many reached a price discussion, and how many said yes. Everything else is a proxy for those, and proxies are exactly what you cannot afford to trust when the sample is ten.

Write down where each of your first ten came from, in one line each, on the day it happens. You will not remember accurately in three months, and this list is the single most valuable document you will own when you eventually decide where to spend money. The version of this that bit me was losing deals I had not tracked properly, which I wrote up in what three lost deals taught me about my discovery pipeline.

When should a founder stop doing this personally?

Later than feels comfortable, and specifically not until you can describe the sale as a repeatable sequence. If you cannot write down what you say, in what order, and what objections come back, you cannot hand it to anyone, and hiring a salesperson to discover it for you is an expensive way to postpone the learning.

The signal to look for is boredom. When the tenth conversation feels like the ninth, and you can predict the objection before it arrives, you have found the pattern. That predictability is the asset. It is what a hire receives, what a landing page describes, and what an ad campaign amplifies.

Selling from Bengaluru to buyers in other countries has made this lesson unavoidable for me, because the shortcuts that rely on being in the room are not available and the sequence has to be genuinely written down. That constraint turned out to be useful. Narrowing who you sell to makes the sequence easier to find, which is the argument I made in niching down versus staying a generalist.

What should you do next?

Make a list of twenty named people, not companies, who have shown you evidence they have this problem. Evidence means something they said or did, not something you infer from their job title. If you cannot get to twenty, that gap is itself the finding, and it is better to learn it this week than after a quarter of building.

Then have five conversations before Friday, ask about the last time the problem happened, and get to a price question in every one. Write down where each person came from. After ten, look at the list and see whether any source appears more than twice. That repetition, if it exists, is your first channel.

I work with founders on exactly this stage, mostly on fixed fees, with most projects landing between 1,000 and 10,000 dollars. If you have a product and a blank space where the first customers should be, reach out and let's chat.