Why Measurement Systems Fail Quietly

Every accept or reject decision, every process adjustment and every safety trip depends on a number. If the number is wrong, the decision is wrong, but the paperwork still looks correct. That is why measurement problems are dangerous: they do not announce themselves. A broken pump makes a noise. A biased gauge produces neat, confident data.

Two kinds of wrong decision

• False accept (also called consumer's risk): a part or batch that does not conform is measured as conforming and is shipped. The customer finds the problem, or nobody does until it fails in service.
• False reject (also called producer's risk): a conforming part is measured as nonconforming. It is scrapped, reworked or held. The cost is real but hidden in yield losses and rework hours.

Both risks always exist. A good measurement system makes them small and known. A poor one makes them large and unknown.

Worked case: the over-adjusted filling line

A filling line measured the net weight of each container on a checkweigher. Operators adjusted the filler every time a reading moved away from the target. The checkweigher had a repeatability problem: the same container gave readings that varied by several grams. The operators were reacting to measurement noise, not to the filler. Each adjustment added real variation to the process. When the checkweigher was repaired and the rule changed to "adjust only on a control chart signal", the actual fill variation dropped and give-away fell.

Worked case: the pressure test that passed

A fabrication contractor hydrotested a piping spool and recorded a pass. The pressure gauge was in calibration by date, but it had been dropped the week before and nobody reported it. Its reading was low by a meaningful amount. The spool had been tested below the required test pressure. Nobody knew until a later gauge check found the error and the as-found data was traced back. Every test recorded with that gauge since the drop had to be reviewed.

Why it is missed

Organisations often treat a valid calibration label as proof that the measurement is good. Calibration answers a narrower question: does this instrument read correctly against a reference, under laboratory conditions, at these points? It says nothing about the operator, the fixture, the method or the part. A calibrated micrometer used with poor technique on a warm, oily part can still give a poor measurement. The rest of this course treats the measurement system as the whole chain, and asks how to prove each link.