Projects

In 2023 I built a product brand properly. Custom branding, two colourways, roughly three hundred units sourced direct from a Chinese manufacturer and paid for up front. It sold. It also lost money on almost every unit, and it took five or six months to clear the stock I had already paid for.

The number that killed it was the margin multiplier. Landed cost to sale price ran at about three times. That sounds healthy until you start subtracting. Advertising takes a cut. Platform fees take a cut. Returns take a cut, and the return rate ran near ten percent because the product had electrical defects I could not see from photographs. At three times, there is no room for ads, fees, returns and one mistake at the same time.

Everything else that went wrong was a symptom of that one number. I have seen the same pattern in other people's stores since. They describe a marketing problem, or a supplier problem, or a platform problem. Underneath it is almost always a multiplier that was never high enough to survive the real cost of selling.

Total loss came to around seven thousand dollars including ad spend. I count that as tuition rather than waste. But only because I can now state the rule in one sentence, which I could not do before.

Cost to price has to be five times or better. Below that, the first mistake is the last one.

The idea was a channel that did not depend on me being in front of a camera. Scripted, produced, and entirely anonymous. On paper that scales. In practice it removed the only thing that was actually working in my favour.

Faceless content competes purely on production and volume. I was never going to win on either against people doing it full time with teams. What I do have is specific experience and a willingness to say what I think. None of that survives being stripped out of the frame. I was playing away from my strengths and calling it a strategy.

The second problem was cost per datapoint. Every experiment on that channel was slow and expensive to run. Weeks of production to learn one thing about what an audience responded to. Compare that to writing, where the same question gets answered in an afternoon. When learning is expensive, you learn less, and you learn it later.

I killed it rather than parking it. Parking a project keeps it on the list and keeps it taking up attention. Calling it dead frees both. The two rules it produced still bind what I build now, which is more than most successful projects leave behind.

Kill it or commit to it. A parked project still charges rent on your attention.

The premise is narrow on purpose. Any chatbot can summarise a book in seconds, so a summary is worth nothing. What nothing on the market does well is turn a book into a system you actually run. The insights, plus workbooks, prompts and a tracker that make you do the thing the book describes.

The first version covers one book. Someone bought it in July, organically, through a marketplace listing, for a little under thirteen dollars net. I want to be precise about what that means, because it is easy to inflate. It means the product can be bought. It does not mean there is a market. One sale is a signal that the mechanism works end to end, and nothing more than that.

The temptation after a first sale is to mass-produce. Twenty more books, immediately. That would be building on a sample size of one, and the failure mode is obvious in hindsight and invisible at the time.

The real constraint has never been production. I can build these faster than I can sell them, which is exactly the trap. Every project on this page failed or stalled on distribution, and this one will too unless the demand question gets answered before the catalogue grows.

A first sale proves the thing can be bought. It does not prove anyone is waiting for the next one.