Measurement uncertainty (GUM), explained for calibration labs

No measurement is exact. Uncertainty is how you state, honestly and numerically, how much doubt surrounds a result — and an ISO/IEC 17025 certificate without it is incomplete.

Uncertainty is doubt, quantified

When you report 10.000 V, you are really saying “our best estimate is 10.000 V, and the true value is very likely within a small range around it.” Measurement uncertainty is that range, expressed so anyone can judge whether the result is fit for their purpose. The internationally agreed method for working it out is the GUM — the Guide to the Expression of Uncertainty in Measurement.

Two kinds of component: Type A and Type B

Type A — from your own data

Type A components are evaluated statistically. Take repeated readings, calculate the standard deviation, and the standard uncertainty of the mean follows from it. Repeatability is the classic example: it falls straight out of the measurements you just made.

Type B — from everything else

Type B components are evaluated from other information: the reference standard’s calibration certificate (its CMC), instrument resolution, drift, temperature effects, manufacturer specifications. Each is converted to a standard uncertainty using its distribution — divide by √3 for a rectangular spec limit, by 2 for a certificate quoted at k=2, and so on.

Combining the components

Each component is multiplied by a sensitivity coefficient (how strongly it affects the result) and then combined by adding them in quadrature — the square root of the sum of the squares. The result is the combined standard uncertainty, u_c. Quadrature, not simple addition, because independent errors partly cancel rather than all stacking up at once.

From combined to expanded uncertainty

A single standard uncertainty corresponds to roughly 68% confidence — not enough for a certificate. Multiply u_c by a coverage factor k (about 2 for ~95% confidence) to get the expanded uncertainty U. The exact k comes from the effective degrees of freedom via the Welch–Satterthwaite formula; for budgets dominated by well-known Type B terms, k = 2 is the usual reported value.

Result = measured value ± U, at a stated coverage (e.g. k=2, ~95%). That ±U is the heart of the certificate.

Why it decides conformity

Uncertainty is not paperwork — it changes pass/fail. If a tolerance limit sits inside the result’s ±U band, you cannot honestly claim a clear pass or fail without a stated decision rule (such as guard-banding). ISO/IEC 17025 expects the uncertainty to be evaluated and, where relevant, taken into account in the statement of conformity. Skipping it doesn’t make the doubt disappear; it just hides it.

Make the budget repeatable

The hard part isn’t the maths — it’s applying the same, defensible budget to every calibration without copy-paste errors. Software that stores each component, its distribution and sensitivity, and computes u_c, the effective degrees of freedom and U automatically keeps every certificate consistent and audit-ready.

More from the Cali blog: browse all calibration & metrology guides →