For a while we did what design agencies do: research, interfaces, a neat handover, an invoice. Some of that work shipped. A lot of it did not, and we watched good design quietly die between the file and the release.
Then a jeweller asked for something we could not hand over.
What he actually needed
He needed the rate to be right at the moment of sale, old gold to be valued and credited without a calculator, and a set of books that would survive an audit. None of the retail software he could buy did the first two, and the third followed from them.
We could have designed a beautiful interface for a system that did not exist. Instead we built the system.
What changed for us
Once you run the thing, the incentives invert. A shortcut you take in month one is a support call you take in month six. We stopped optimising for the handover and started optimising for the year after it.
That is now how we work for clients too: we would rather argue about scope before we agree about price than deliver exactly what was asked for and watch it fail.
What it became
DiamondOar runs jewellery businesses today. It is also the reason every product after it exists — once you have built and operated one real product, the second is a different kind of problem.