I mend domestic mechanical clocks at one bench in Lyon and write down what each movement actually needed, so the next one of its kind costs less guessing and fewer ruined parts.

This is one room on the first floor of a building in the ninth arrondissement of Lyon, with a bench one metre ten long under a north window and a wall board behind it where movements run on test. I have worked at it since 2015. What comes through the door is domestic and mechanical: French mantel movements under glass domes, weight-driven wall regulators, the odd carriage clock, and longcase movements that arrive in a cardboard box with the weights wrapped in a towel. I do not touch watches. The bench is too big, the loupe is too weak and the patience is the wrong shape for them.

Every job gets a written record before, during and after: what the pivots measured, which holes were bushed and with what, the free length and thickness of the mainspring, the shake at each bearing when the plates went back together, and the rate over fourteen days on the board. The record exists because my memory is confident and wrong. These pages are that record, tidied up and stripped of anything that would identify whose clock it was. The numbers are copied straight out of the job book, including the ones that show me doing something twice.

I came to this late. I was thirty-four and doing something else entirely when I started taking movements apart in a spare room, and most of what I know arrived through spoiled parts. Nothing written here is instruction. It is a log of decisions made at one bench with one set of tools, and several of those decisions were wrong at the time and are marked as such. There is nothing for sale on this site, no workshop named, no supplier named, and no clock owner named.