The first customer
One payment is a real fact about the world and a much smaller one than it feels like on the day.
I remember the sensation better than the amount. There was no launch and no first day, only a first person who paid me: small books, personal filings, the sort of work that arrives because somebody heard you could do it. That was 2014, on my own account, while I was also working for other people.
The sensation is that something has been settled.
Nothing has been settled. What has happened is narrower and stranger than it feels, and it took starting a company twelve years later, registered in Ontario under its own name, to make me look at the first-customer moment properly. Actually, a company was the wrong prompt. The prompt was noticing that the feeling had been identical both times, and that it had told me nothing either time.
What one payment actually proves
Before anyone pays, everything you believe about the work is a hypothesis. It might be a well-informed hypothesis. You may have done the work for an employer for a decade. It is still an untested claim, because the claim is not “I can do this”, it is “someone will part with money for me to do this rather than doing something else with the money”.
After the first payment, exactly one thing has changed: you know that claim is not impossible.
That is a real change and it is not nothing. Impossible and rare are different categories, and moving between them is the only move a first customer can make for you. What you have learned is that at least one person, in one situation, facing whatever alternatives they happened to know about, on whatever day they happened to be deciding, preferred this.
Every one of those clauses is doing work, and every one of them is a reason the observation does not generalise.
They were selected on proximity and tolerance
The next mistake is to treat the person as a market.
I have watched people do it and I have done a version of it. The first customer becomes the model customer. The thing they asked for becomes the roadmap. Their particular constraint, which may have been idiosyncratic to the point of being unique, becomes a design principle, and the second year of the business gets built around a sample of one that nobody has examined for how it was drawn.
And it was drawn terribly. Your first customer is almost never a random selection from the people you hope to serve. They are the person closest to you who had the problem, which means they are selected on proximity and tolerance, not on need. They tolerated an unfinished thing, an unclear price, and someone learning in front of them. That tolerance is the least representative property a customer can have.
The error I actually made, which is the other one
I want to be honest about the direction of my own failure, because the essay so far reads as a warning against enthusiasm and my problem was the reverse.
For years I treated early customers as anecdotes. I would not update on them. I kept waiting for enough of them to constitute evidence, in the way a sample becomes evidence, and a small practice does not produce that. You will not reach a defensible sample size. Ever. The book is fifteen people and then thirty and the composition changes underneath you the whole time, so the statistical patience I was so proud of was just a way of never concluding anything while continuing to work.
So the actual position is uncomfortable. Overreading one customer builds the wrong thing quickly. Underreading them builds nothing slowly. There is no sample size that dissolves the choice, and anybody offering you a threshold at which anecdote becomes data is selling a number they made up.
What has stayed with me is not the conclusions. It is the questions those first people asked, which I can still recall almost verbatim, long after every inference I drew from them turned out to be wrong.