Measurement uncertainty in ISO 17025
Measurement uncertainty describes the spread of values that you can reasonably attribute to a measured quantity, based on the information available. This gives a measurement result context: how much confidence does the number give for your application? The RvA explanatory notes on ISO 17025 (Dutch Accreditation Council) cover measurement uncertainty and reporting. Use them alongside the laboratory scope and the explanation of ISO 17025.
As a buyer, you do not need to calculate the full uncertainty budget yourself. You do need to understand which result the laboratory reports, and whether the related uncertainty suits your decision. A number with many decimal places is not, in itself, proof of an accurate measurement. You can check the chain behind a number on metrological traceability and at a calibration laboratory. You can read how to support a pass or fail decision with that uncertainty on the statement of conformity.
Measurement uncertainty, measurement error and tolerance
These three terms answer different questions. Keep them separate when you discuss a measurement report.
| Term | Meaning in practice | Example question |
|---|---|---|
| Measurement error | Difference between a measured value and a reference value | Which deviation was found in the comparison? |
| Measurement uncertainty | Uncertainty that belongs to the result | Which spread remains after assessing the measurement influences? |
| Tolerance | Permitted limits set by the specification | What does the drawing, client agreement or product requirement accept? |
A known deviation may call for a correction. Some uncertainty usually remains after that. The tolerance is not determined by the measurement uncertainty: it comes from the application or specification.
Where does uncertainty come from?
An instrument is only one part of the measurement chain. The method and the conditions can also affect the result. The table below is your own starting list for a conversation with the laboratory. It is not a complete uncertainty budget.
| Possible contribution | What you want to be able to find |
|---|---|
| Reference or calibration | Relevant certificate, measurement point and uncertainty |
| Repeatability | Results of suitably repeated measurements |
| Resolution | The smallest step the instrument can display |
| Environment | Influence of, for example, temperature on the measurement |
| Sample and preparation | Variation or changes before and during measurement |
| Method and calculation | Substantiated assumptions and corrections used |
Ask how the relevant contributions were assessed and combined. A laboratory may not simply add up all the numbers from different documents. Their relationship, distributions and any dependencies are part of the technical justification.
Simple fictional calculation example
Suppose a laboratory for a length measurement has determined a combined standard uncertainty of 0.04 mm. In this example, it uses a coverage factor k = 2. The expanded uncertainty is then 2 × 0.04 = 0.08 mm.
In this simplified example, a result can be reported as 20.00 mm ± 0.08 mm, with k = 2. The real justification belongs to the method and the report. This is only an explanation of the final calculation step. It is not a worked-out validation or a suitable uncertainty budget for a real measurement.
NIST explains that k = 2 can give roughly 95% coverage under suitable assumptions. That percentage does not follow automatically from every number with ”±”. The distribution and reliability of the estimate are decisive.
Does the result suit your decision?
Suppose a fictional client accepts a dimension between 19.95 and 20.05 mm. The central value of the example lies within that range, but the expanded uncertainty is relevant to the decision. So agree in advance which decision rule the laboratory uses when it reports “conforms” or “does not conform”.
A decision rule describes how uncertainty is taken into account in such a statement. Choose one that fits the standard, regulation or agreement that applies to the order. The laboratory and the client must both know which conclusion is requested. A single measured value and a statement of conformity are different outcomes.
When an uncertainty is reported, also ask about the unit, the coverage factor and the justification. Compare results only when it is clear that the same quantity and conditions are meant.
Frequently asked questions
Does measurement uncertainty always have to be as small as possible?
It must suit the intended use. A much smaller uncertainty can cost extra time or money without improving your decision. Agree the required suitability in advance.
Is k = 2 required for every ISO 17025 measurement?
No, do not treat that as a universal rule. The applicable technical framework and the justification determine the approach. The NIST explanation is background to the concept. It does not replace the RvA requirements for a specific discipline.
Does a calibration certificate prove that my whole measurement process is correct?
No. The certificate gives information about the calibration that was carried out. The application, environment, method and measured object can have additional influences.