Three Measurement Mistakes Quality Teams Still Make

1. Treating resolution as accuracy

A balance that displays 0.001 g is not necessarily accurate to 0.001 g. Resolution is only the smallest change the display can show. Accuracy depends on repeatability, linearity, eccentricity, temperature sensitivity and drift, and is expressed through calibration error and uncertainty.

The opposite mistake also happens. If a process needs results to 0.001 g but the instrument shows steps of 0.005 g, no calibration can fix that. The instrument is simply unsuitable. Instrument selection should start from the tolerance, then work back to the required uncertainty, and only then to resolution.

2. Treating the “10:1 rule” as a law

The old rule of thumb said a reference should be ten times more accurate than the instrument it checks. In modern practice the comparison is the test uncertainty ratio (TUR): the tolerance of the instrument divided by the expanded uncertainty of the calibration. A TUR of 4:1 is the widely used benchmark in industry, for example in ANSI/NCSL Z540.3.

Two points matter more than the exact number:

• A good ratio does not remove the need to know the uncertainty. The ratio is calculated from it.
• When the ratio is low, results near the tolerance limit become uncertain. That is where a decision rule with a guard band is needed (see Decision Rules).

3. Expecting the calibration lab to decide the due date

ISO/IEC 17025 treats calibration status and calibration intervals as two separate responsibilities.

• The user owns the validity period. Clause 6.4.8 requires that equipment needing calibration, or with a defined period of validity, be labelled, coded or otherwise identified so the user can see its calibration status or period of validity. Clause 6.4.7 requires a calibration programme that is reviewed and adjusted as necessary. Both obligations sit with the organization that uses the instrument. It knows how often the instrument is used, in what conditions, and how critical the results are.
• The calibration lab reports; it does not set the interval. Under clause 7.8.4.3, a calibration laboratory may not put a recommended calibration interval on the certificate or the calibration label, unless this has been agreed with the customer. Where a customer asks for a due date on the label, the date reflects the customer’s decision, not a judgement the lab makes on its own.

Under ISO/IEC 17025 clause 7.8.4.3, an accredited laboratory may not put a recommended interval on a certificate or label unless the customer has asked for it. The due date belongs to the user’s own calibration programme.

A useful equipment label shows:

• the instrument ID
• the date of the last calibration
• the next due date, set by the user
• any restriction, for example “use 0–10 bar only” or “correction applies, see certificate”

A label never overrides events. An instrument that has been dropped or overloaded should be taken out of use and checked, whatever date the label shows.