Measurement uncertainty in calibration is the quantified range within which the true value of a measurand can reasonably be expected to lie, given the conditions and methods of the measurement. It’s not an admission of error — it’s a formal, standardized statement of confidence required under ISO/IEC 17025:2017 for every calibration activity. Without it, a calibration certificate tells you what the instrument reads, not how much you should trust that reading.
Every instrument acceptance decision depends on uncertainty. Whether you’re determining if a pressure gauge meets specification or deciding how often to recalibrate a reference thermometer, the uncertainty budget underpins the call. Get it wrong and compliant-looking equipment can still generate out-of-specification results on the production floor.
This guide walks through the full process: what measurement uncertainty means in a calibration context, how to classify and quantify each contributing source, how to calculate combined and expanded uncertainty using the GUM framework, and what a calibration certificate must include to satisfy ISO/IEC 17025:2017 clause 7.8.4.
Key Takeaways
- Measurement uncertainty is mandatory on calibration certificates under ISO/IEC 17025:2017 clause 7.8.4 — the certificate must state the expanded uncertainty value, coverage factor k, and approximate confidence level.
- Uncertainty sources split into two types: Type A (statistical analysis of repeated measurements) and Type B (certificates, specifications, or reference data).
- The international standard for uncertainty estimation is JCGM 100:2008 — the Guide to the Expression of Uncertainty in Measurement (GUM), maintained by the BIPM.
- ANSI/NCSL Z540.3-2006 requires a minimum 4:1 test uncertainty ratio (TUR): the calibration system’s expanded uncertainty must be no more than 25% of the instrument’s tolerance.
What Is Measurement Uncertainty in Calibration?
Measurement uncertainty is a parameter, associated with a measurement result, that characterizes the dispersion of values that could reasonably be attributed to the measurand (JCGM 100:2008, clause 2.2.3). In practice, it defines the confidence interval around every number on a calibration certificate. A pressure gauge reading of 100.0 psi with an expanded uncertainty of ±0.5 psi (k=2, ~95%) tells you the true pressure likely falls between 99.5 and 100.5 psi — not that the instrument is exactly right.
This distinction matters in regulated industries. An instrument that reads within tolerance, but whose calibration was performed with large measurement uncertainty, may not actually be fit for purpose. The result only carries meaning if the calibration uncertainty is small enough relative to the instrument’s tolerance — a relationship formalized as the test uncertainty ratio (TUR).
The concept eliminates the false precision of reporting a single calibration value with no context. Every measurement has inherent variability from multiple sources: the reference standard’s own uncertainty, environmental conditions, instrument resolution, repeatability, and drift since last calibration. Uncertainty quantifies all of them into a single defensible figure.
Understanding what uncertainty means for your calibration documentation is essential to using those documents correctly. Calibration certificates from ISO/IEC 17025-accredited labs report more than pass/fail for exactly this reason.
Type A vs. Type B Uncertainty Evaluation
The GUM classifies uncertainty evaluations into two types: Type A uses statistical analysis of repeated measurements; Type B covers every other method (JCGM 100:2008, clause 4). In a properly built uncertainty budget, both types carry equal weight. The distinction is how you calculate — not which type is more reliable.
Type A Evaluation
Type A uncertainty comes from repeated measurements under the same conditions. Take n readings, calculate the mean and standard deviation, and express the standard uncertainty as the standard deviation of the mean: uA = s/√n. Running 10 repeated measurements at a single calibration point is a typical Type A evaluation in laboratory practice.
Increasing n reduces Type A uncertainty, but with sharp diminishing returns. Going from n=5 to n=10 reduces the contribution by about 30%. Going from n=10 to n=100 reduces it by another 68%. Beyond 10–15 readings, the time cost rarely justifies the marginal reduction in uncertainty.
Type B Evaluation
Type B covers everything that isn’t statistical repetition: the calibration certificate of your reference standard, manufacturer accuracy specifications, temperature coefficients, drift since last calibration, and resolution. You assign a probability distribution to each source and divide by the appropriate divisor to convert the stated limit to a standard uncertainty.
Common distributions used in Type B evaluation:
- Normal distribution: use when a ±limit is given with a stated confidence level (e.g., ±0.01% at 95%) — divide by 2
- Rectangular (uniform) distribution: use when only a range is known with no stated probability — divide the half-width by √3 ≈ 1.732
- U-shaped distribution: use for sources dominated by a single sinusoidal cycle such as AC waveform residuals — divide by √2 ≈ 1.414
Most calibration guides treat Type A and Type B as roughly equivalent categories. In practice, Type B often dominates the budget for well-maintained reference standards — and that’s expected. The GUM explicitly states Type B evaluations can be just as reliable as Type A when based on sound data. The source of a contribution matters less than the rigor of how you quantified it.
Understanding how working and reference standards sit in the traceability chain clarifies why reference standard uncertainty is almost always the dominant Type B source in a calibration budget.
How to Build a Measurement Uncertainty Budget
An uncertainty budget is the structured list of all contributing sources, their magnitudes, distributions, and combined effect. ISO/IEC 17025:2017 requires documented procedures for estimating uncertainty, and a budget is the auditable record that shows the work. Building one follows six consistent steps regardless of the measurand.
Step 1 — Define the measurand and measurement model.
Write the measurement equation. What are you measuring, under what conditions, and using which method? This equation tells you which sources belong in the budget and which are irrelevant.
Step 4 — Apply sensitivity coefficients.
If a source isn't in the same units as the measurand, multiply by the partial derivative of the output with respect to that input. Temperature uncertainty in °C becomes pressure uncertainty in psi through the instrument's temperature coefficient.
Step 2 — Identify all significant uncertainty sources.
For most calibrations this includes: reference standard uncertainty (from its certificate), repeatability (Type A), instrument resolution, environmental conditions (temperature, humidity), and drift of the reference since last calibration.
Step 5 — Combine in quadrature (RSS).
Sum all contributions using root-sum-of-squares, assuming uncorrelated sources:
uc = √(u1² + u2² + u3² + ... + un²)
Step 3 — Quantify each source.
Express each as a standard uncertainty u in the units of the measurand. For Type A: u = s/√n. For Type B: apply the distribution divisor to convert from a stated limit to a standard uncertainty.
Step 6 — Apply a coverage factor.
Multiply by k to get expanded uncertainty: U = k × uc. For k=2 (~95% confidence, normal distribution), this is the default for most calibration certificates.
Calculating Combined and Expanded Uncertainty: A Worked Example
Consider a 0–100 psi pressure gauge with a ±0.25% FS tolerance. A 4:1 TUR requires your calibration standard’s expanded uncertainty to be ≤ 0.0625% FS. The table below shows a complete budget for this instrument, with the RSS calculation beneath it.
The TUR calculation guide covers how to determine your ratio from the uncertainty budget and what to do when you can’t achieve 4:1 with your current reference standard.
| Uncertainty Source | Type | Stated Value | Distribution | Divisor | Standard Uncertainty |
|---|---|---|---|---|---|
| Reference standard | B | ±0.04% FS | Normal (k=2) | 2 | 0.020% FS |
| Repeatability (n=10) | A | s/√n | — | — | 0.015% FS |
| Resolution | B | ±0.01% FS | Rectangular | √3 | 0.0058% FS |
| Drift (1-year interval) | B | ±0.02% FS | Rectangular | √3 | 0.0115% FS |
| Temperature correction | B | ±0.01% FS | Rectangular | √3 | 0.0058% FS |
uc = √(0.020² + 0.015² + 0.0058² + 0.0115² + 0.0058²)
uc ≈ 0.027% FS
U = k × uc = 2 × 0.027% = ±0.054% FS
This result appears on the calibration certificate as: “Expanded uncertainty: ±0.054% FS (k=2; approximately 95% confidence).” For non-linear measurement models, JCGM 101:2008 provides the Monte Carlo method as an alternative propagation approach.
How to Report Measurement Uncertainty on a Calibration Certificate
ISO/IEC 17025:2017 clause 7.8.4 specifies what a calibration certificate must include regarding uncertainty. For accredited laboratories this is non-negotiable: every calibration result must carry a reported measurement uncertainty. A compliant statement requires four elements: the numerical value, the unit, the coverage factor k, and the approximate confidence level.
Example of a compliant uncertainty statement:
Calibration uncertainty: ±0.054% of full scale (k=2; approximately 95% confidence)
ILAC Policy ILAC-P14:01/2013 adds that uncertainty must be reported in the same unit as the calibration result and clearly associated with each measured value when multiple points are reported across a range.
Many certificates from non-accredited labs state a tolerance — “calibrated within ±0.1%” — rather than a measurement uncertainty. These are different things. Tolerance is the specification the instrument must meet. Measurement uncertainty is a property of the calibration measurement itself: it tells you how well you know whether the instrument actually met that specification. A certificate with no stated measurement uncertainty cannot demonstrate metrological traceability under ISO/IEC 17025:2017, regardless of the equipment used.
Common Mistakes That Undermine Uncertainty Estimates
Persistent errors in uncertainty estimation don’t always produce obviously wrong budgets — that’s exactly what lets them survive audits for years. These are the patterns most commonly seen in laboratory reviews.
Omitting drift. Drift between calibration cycles is one of the largest Type B contributors for reference standards, yet it’s routinely absent from budgets. If your reference thermometer shows ±0.05°C drift over two years and you recalibrate annually, include ±0.025°C as a rectangular distribution: udrift = 0.025/√3 ≈ 0.014°C. NIST Technical Note 1297 recommends documenting why any source is considered negligible when it’s excluded.
Double-counting reference uncertainty. If you include the reference standard’s full expanded uncertainty from its calibration certificate, don’t also separately list its resolution, repeatability, and temperature contributions. The certificate value already incorporates those components.
Applying the wrong distribution to a tolerance limit. Using a normal divisor (÷2) on a manufacturer specification stated as a limit with no confidence level underestimates the standard uncertainty by roughly 15%. A tolerance limit with no stated probability is a rectangular distribution — divide the half-width by √3.
Not updating the budget across the calibration range. A budget built for 50% of full scale isn’t automatically valid at 5% or 100%. Sensitivity coefficients change, linearity errors grow at range extremes, and repeatability often degrades at the low end. Evaluate the budget at the worst-case calibration point, or at each point where the instrument is verified.
Treating the reference standard uncertainty as always dominant. For instruments with poor inherent repeatability or in facilities without environmental controls, the repeatability or temperature contribution can exceed the reference standard contribution. Review the full budget whenever equipment, lab conditions, or technicians change.
When an incomplete uncertainty estimate surfaces during a corrective action event, decisions about out-of-tolerance findings become much harder to defend. The uncertainty budget is the document that ties the decision to the data.
For comprehensive calibration services with fully documented measurement uncertainty budgets and ISO/IEC 17025 accreditation, request a quote from Micro Precision.
Additional Resources
FAQs
Measurement uncertainty in calibration is a quantitative parameter that characterizes the range of values within which the true value of the measurand can reasonably be expected to lie. It is evaluated per JCGM 100:2008 (GUM) by identifying all uncertainty sources, classifying them as Type A (statistical) or Type B (other means), and combining them to produce an expanded uncertainty value with a stated confidence level.
Type A uncertainty is evaluated by statistical analysis of repeated measurements, expressed as the standard deviation of the mean (s/√n). Type B uncertainty is evaluated by all other means: calibration certificate data, manufacturer specifications, temperature coefficients, or reference data. Both are equally valid in a GUM-compliant budget — the distinction is in the evaluation method, not the reliability of the result.
A coverage factor of k=2 is standard for calibration and corresponds to approximately 95% confidence for a normal probability distribution. Use k=3 for approximately 99.7% confidence. When Type A degrees of freedom are limited (fewer than 10 measurements), calculate effective degrees of freedom using the Welch-Satterthwaite equation to determine the correct k value.
Yes. ISO/IEC 17025:2017 clause 7.8.4 requires accredited calibration laboratories to report measurement uncertainty on all calibration certificates. The statement must include the expanded uncertainty value, the unit, the coverage factor k, and the approximate confidence level. A pass/fail result or tolerance statement alone does not meet this requirement.
The test uncertainty ratio (TUR) is the ratio of an instrument’s tolerance to the expanded uncertainty of the calibration process. ANSI/NCSL Z540.3-2006 requires a minimum TUR of 4:1, meaning the calibration system’s expanded uncertainty must be no more than 25% of the instrument’s tolerance. See our guide on how TUR is calculated and applied.
Reduce uncertainty by upgrading to reference standards with lower stated uncertainty, improving environmental controls (temperature stability, vibration isolation), increasing repeated measurements to reduce Type A contributions, shortening calibration intervals to minimize drift, and validating estimates through proficiency testing against external reference laboratories.
The GUM (Guide to the Expression of Uncertainty in Measurement, JCGM 100:2008) is the authoritative international framework for evaluating and reporting measurement uncertainty, maintained by the BIPM. ISO/IEC 17025:2017 requires accredited labs to apply GUM principles or equivalent methods when estimating and reporting measurement uncertainty.
Accuracy is a qualitative descriptor. Tolerance is a specification limit — the maximum permissible error for an instrument to remain fit for use. Measurement uncertainty is a quantitative, calculable parameter with units, a coverage factor, and a stated confidence level. An instrument can fall within tolerance while the calibration uncertainty is large enough to make that acceptance decision unreliable.
The Number That Makes Calibration Mean Something
Measurement uncertainty is what transforms a calibration result from a recorded reading into a defensible statement. Without it, you know what the instrument reads — you don’t know how much confidence to place in that reading or whether the calibration process was capable of making the judgment you’re relying on.
The framework is consistent regardless of measurand: identify your sources, classify each as Type A or B, convert to standard uncertainties, combine via RSS, apply a coverage factor, and report the expanded uncertainty with k and confidence level. Following JCGM 100:2008 and meeting ISO/IEC 17025:2017 clause 7.8.4 keeps your program auditable and your decisions defensible.
If your calibration program needs a documented uncertainty review or you need ISO/IEC 17025-accredited calibration certificates with NIST-traceable measurement uncertainty, contact Micro Precision.