Proof

An analytical balance in a laboratory is a precise instrument and nobody trusts it.

Before a run, the operator places a certified mass on the pan and reads what the balance says. After the run, they do it again. The certified mass is a check weight, and it exists for one reason, which is that a balance can drift without giving any indication that it has drifted. A drifting balance does not produce an error. It produces a number, in the usual format, to the usual number of decimal places, at the usual speed.

The discipline around the check weight is the part worth attention.

If the check at the end of the run fails, the run is not the only thing in question. Everything measured since the last successful check is in question, because there is no way to know when during that interval the drift occurred. This is why the check happens at both ends. The before-check is not confirming the instrument works. It is establishing a boundary, so that when a later check fails there is a known point to fall back to.

And the check weight itself is only worth something because of what sits behind it. It carries a calibration certificate that traces it to a higher standard, which traces to a national standard, which traces to the international definition of the kilogram. Break any link in that chain and the check weight becomes an object of approximately known mass, which is to say a paperweight.

The thing this apparatus is built to prevent is a specific and otherwise unpreventable failure, which is a measurement that is wrong in a way the measuring instrument cannot detect. You cannot verify a balance by weighing something on it twice. You cannot verify it by weighing something heavier. The only thing that works is an external reference with an unbroken chain back to something that is not your balance.

Proof is the third discipline in the Process layer and the last of the three Ps. Permission decided what the system may reach. Practice was the work of finding where it touches the job. Proof is the evidence that any of it worked, produced in a form that convinces somebody who was not there, does not trust you, and has the standing to ask.

Chapter seven ended on a question it deliberately did not answer. Everything in that chapter is asserted by the people doing it. The team says the routine changed, the reps are happening, the corrections are being captured. All of it may be true and none of it is evidence. The scraper's answer was to re-blue the surface and count the contact points, which works because the measurement is independent of the claim.

This chapter is about what the equivalent is when the surface is a production system, and it starts with a number almost nobody reports.