A watch comes back from a service with a printout saying it runs at +2 seconds a day. You wear it for a week and it is a minute ahead. Nobody has made a mistake.

The two numbers were produced by different methods asking different questions, and knowing which is which saves a lot of unnecessary worry and a certain amount of unnecessary correspondence with watchmakers.

What a timing machine actually measures

A timing machine listens. The watch sits in a cradle against a microphone, and the machine picks up the sound of the escapement and works out, from the spacing of those sounds, how fast the balance is currently swinging.

Three things follow from that, and all three matter.

It measures a few seconds, not a day. The seconds-per-day figure it prints is an extrapolation, the rate the watch would run at over a day if it carried on exactly as it is doing right now.

It measures one position at a time. A careful reading is taken in several positions in turn, and a service report may show all of them, but each line is a single orientation held still.

It measures a watch doing nothing. Sitting in a cradle, wound to a chosen state, at room temperature, undisturbed.

That is a controlled measurement, and control is the point. It is exactly what you want when you are adjusting a watch, because it isolates the variable you are working on.

What your wrist measures

Wearing the watch measures the opposite. It integrates everything: every position the watch passed through, every hour of a mainspring winding and unwinding, your body temperature during the day and the room’s overnight, and whatever the week happened to contain.

That is uncontrolled by design, and representative as a result. It is the only measurement that answers the question an owner actually has, which is whether the watch will be right when they look at it on Thursday.

Why the two disagree

Once the methods are set out, the disagreement stops being mysterious.

  • The bench figure is one position, your week is all of them. Gravity moves the rate, often by more than the rate itself. A watch adjusted to a fine figure lying flat can run differently on its side, and your nights are spent in a position somebody chose without measuring it.
  • The bench figure is usually taken near full wind. Your watch spends part of every cycle further from that.
  • A few seconds of listening cannot see a week. Anything intermittent, anything that depends on how the watch is moved, simply is not in the sample.
  • Temperature. A cradle on a bench is a steadier place than a wrist and a windowsill.

A service report and a week on the wrist can differ by several seconds a day with both being accurate accounts of what they measured.

Which one should you believe

Both, for their own purposes.

The bench figure is the right tool for adjustment. If a watchmaker is regulating a watch, they need the controlled number, and the spread across positions tells them things about the movement that a weekly average never could.

The lived figure is the right tool for ownership. If you want to know what your watch does in your life, only wearing it and measuring will tell you, and no amount of bench precision substitutes.

They are also good at catching different things. A timing machine can reveal a problem while the watch is still keeping tolerable time, because it hears the escapement directly. A lived rate can reveal a problem the bench missed, because it includes conditions the bench excluded.

What this means in practice

If your watch comes back from a service at a fine figure and then keeps worse time than that on your wrist, the first thing to look at is not the watchmaker’s work. It is which position it spends the night in and whether it is staying wound. Resting position and overnight accuracy covers the first of those, and it is a change you can make in a second.

If you have adjusted for those and there is still a large gap, that gap is genuinely worth mentioning, because a substantial difference between controlled and lived performance is information a watchmaker can use.

Measuring the lived rate

You do not need a machine for the wrist side of this. How to measure watch accuracy without a timegrapher is the whole method, and the calculator does the arithmetic.

Ratekeeper keeps the series rather than the single figure, which is what turns a set of readings into something a bench printout can be compared against. It is free for your first watch.