We started by selling our time

The first thing this company had was a services page. Capabilities, an inquiry form, a description of what we could do for you. It is the obvious way to start, because it is the only model that produces revenue in week one without asking anyone to believe a thesis.

It also works. Clients pay, problems get solved, and the feedback is immediate in a way that no internal project ever is. Nothing about the decision was wrong given what we knew. What became clear over the following months is that the model has a ceiling built into its shape, and the ceiling is not the rate.

What a services business cannot do

Every engagement starts at zero. Not literally — you carry judgment and habits between clients — but the artifact does. The thing you built for the last client belongs to them, sits in their repository, and solves their version of the problem. The next engagement begins with an empty directory and the same conversation you had four months ago.

The knowledge stays in your head rather than on a shelf. That feels like expertise accumulating, and in a limited sense it is, but it is the worst possible storage medium: it does not survive attention moving elsewhere, it cannot be handed to anyone, and it cannot be executed by a machine. A system that improves only through the operator getting more experienced is a system with one very fragile component.

And the asset you build is a reputation rather than a machine. Reputation is real capital and it compounds slowly, but it is entirely dependent on you continuing to show up. Stop for four months and it decays. Stop for four months with a working system and the system is still there.

The part worth keeping

Client work is the cheapest market research that exists, and it is the only kind where someone pays you to receive it. You learn which problems are urgent rather than interesting, which workflows are genuinely fragmented, and where people are already spending money badly. Those are the selection criteria we still use.

It also teaches the difference between a problem someone will describe at length and a problem someone will pay to remove. Those are not the same set, and the gap between them is where most product ideas die. Nothing about a whiteboard reveals which side of the line you are on.

What changed

The decision was to stop selling hours and start producing owned systems, accepting that revenue would go to nothing for a while. That is an easy sentence to write and a difficult one to act on, because services revenue is immediate and system revenue is speculative, and the gap between them has to be funded from somewhere.

The honest reason it was possible is that the alternative was worse in a specific way: a services business that runs for five years leaves you with five years of experience and no asset. Every hour spent on client work is an hour not spent building the thing that makes the next hour cheaper. That trade is fine at the start and ruinous if it becomes the model.

What we did not understand yet was how much machinery sits between deciding to build owned systems and actually being able to. That took the following year, and most of it was not the products.