Maladière, Neuchâtel.

There is one room and it has one purpose. A steel bench, a 400 × 250 × 70 mm granite surface plate ground flat to a grade that means nothing to a keyboard and everything to the habit of working on it, and a window that faces north over the lake side of the town.

Every board that leaves goes across that plate first. It is clamped, a probe is brought down onto one keycap at a time, and the board is asked what it thinks the depth is while the bench tells us what the depth actually is. Eighty-seven keys, about four minutes, and the difference between the two numbers is the certificate.

The order of operations

Cases arrive machined and anodised from a shop in the valley that has been doing watch cases for forty years and finds this work slightly beneath them. Boards are populated and reflowed in Bienne. Everything else — the plate, the gasket, the switches, the caps, the measurement, the box — happens here, and the whole of it fits in one room because the company is not large and there is no advantage in pretending otherwise.

Why the certificate is not a marketing document

ENTREFER sold sensors before it sold keyboards, and it still does. A sensor that goes into a watch-testing bench arrives with a sheet saying what it was measured against and by whom, because a measuring instrument whose own accuracy is a matter of opinion is an ornament. The keyboards inherited the habit. It costs about four minutes a board and it is the only part of this business that would look completely unremarkable to anyone else in this valley.

What we get wrong

Roughly one board in fourteen goes back to the bench, and the cause is almost always the same: a switch that has not seated fully in its socket, which lifts the magnet a few hundredths of a millimetre and moves that key's actuation point by more than the rest of the board. It is not a difficult fault to find once you are looking for it. It took us most of a year to stop blaming the sensors.