How to Estimate & Report Measurement Uncertainty under ISO/IEC 17025:2017
A comprehensive metrological manual for Quality Managers, technical signatories, and analytical chemists. Master Type A statistical evaluation, Type B probability distributions, combined uncertainty, and audit-compliant reporting.
1. Introduction to Measurement Uncertainty in Laboratory Testing
No physical or chemical measurement is absolute. Whenever a chemist measures the concentration of lead in drinking water, the protein percentage in milk, or the tensile strength of an alloy, the measured numerical value is an estimate of the true value of the measurand, inherently influenced by physical variations in instruments, environment, volumetric glassware, and analytical handling.
In modern metrology, Measurement Uncertainty (MU) is defined by the Guide to the Expression of Uncertainty in Measurement (JCGM 100:2008 - GUM) as a parameter, associated with the result of a measurement, that characterizes the dispersion of the values that could reasonably be attributed to the measurand. Rather than a sign of doubt or error, rigorous evaluation of uncertainty is the hallmark of analytical competence and professional integrity.
2. ISO/IEC 17025:2017 Clause 7.6 Mandates
In ISO/IEC 17025:2017, evaluation of measurement uncertainty is governed under Clause 7.6:
Identify All Contributions
Laboratories shall identify the contributions to measurement uncertainty. When evaluating measurement uncertainty, all contributions that are of significance, including those arising from sampling, shall be taken into account using appropriate methods of analysis.
Calibration Laboratories
A laboratory performing calibrations, including of its own equipment, shall evaluate measurement uncertainty for all calibrations and maintain complete metrological traceability.
Testing Laboratories
A laboratory performing testing shall evaluate measurement uncertainty. Where the test method precludes rigorous evaluation, an estimation shall be made based on understanding theoretical principles or practical experience of method performance.
3. Why Measurement Uncertainty Matters in Decision Rules
Measurement uncertainty is not an abstract academic exercise. Its primary real-world significance arises when evaluating conformity against legal and commercial specifications under ISO/IEC 17025 Clause 7.8.6 (Reporting statements of conformity):
The Guard Band & Consumer/Producer Risk
Suppose a statutory regulation sets a strict maximum acceptable limit of 1.0 mg/l for a chemical contaminant in wastewater. If your laboratory measures 0.98 mg/l with an expanded uncertainty of ± 0.05 mg/l, the true concentration could be as high as 1.03 mg/l.
Under modern accreditation guidelines (ILAC-G8), a laboratory cannot declare an unqualified "Pass" without applying a predefined decision rule agreed with the customer, accounting for the expanded uncertainty band around the specification limit.
4. Type A Evaluation: Statistical Analysis of Series of Observations
Type A evaluation is the method of evaluating uncertainty by the statistical analysis of a series of independent repeat observations obtained under defined repeatability or reproducibility conditions.
Mathematical Formulas for Type A Uncertainty
Critical Rule: If an individual routine sample is tested only once, the standard uncertainty due to repeatability is the experimental standard deviation s (degrees of freedom ν = n - 1). If the reported test result is the arithmetic average of n replicate determinations, the standard uncertainty is the standard error of the mean: s / √n.
5. Type B Evaluation: Non-Statistical Probability Distributions
Type B evaluation evaluates uncertainty components through scientific judgment, historical data, calibration certificates, manufacturer tolerance sheets, and reference handbook physical constants.
To combine a Type B component with other sources, its quoted limits must be converted into a 1-sigma standard uncertainty by assuming an underlying probability distribution:
Divide by √3
Used when an uncertainty interval is given as ±a with no knowledge of the distribution (e.g. manufacturer volumetric flask tolerance ±0.15 ml, digital display resolution ±0.01 g).
Divide by √6
Used when values near the midpoint are known to be distinctly more probable than values near the extremes ±a (e.g. filling a volumetric flask meniscus at nominal laboratory temperature).
Divide by k
Used when taking expanded uncertainty U from an accredited calibration certificate with stated coverage factor (typically k = 2 for 95% confidence).
6. Standard Uncertainty: Bringing Components to a Common Scale
Every identified uncertainty source—whether Type A or Type B—must be converted into a standard uncertainty u(x), representing one standard deviation of that input parameter.
Crucially, all standard uncertainties must either share the exact same physical measurement units (e.g., grams or mg/l), or be expressed as dimensionless relative standard uncertainties u(x) / x (often expressed as percentage %).
7. Combined Standard Uncertainty (uc): Law of Propagation
When individual input quantities x1, x2, ..., xn are mutually independent (uncorrelated), the combined standard uncertainty uc(y) is calculated using the square root of the sum of squares:
For Additive Models: y = c₁x₁ + c₂x₂
For Multiplicative / Ratio Models: y = (x₁ · x₂) / x₃
Where ci = ∂f/∂xi represents the partial derivative (sensitivity coefficient), indicating how changes in input quantity xi scale the final result y.
8. The Coverage Factor (k) and Confidence Levels
The combined standard uncertainty uc corresponds to a single standard deviation (~68% coverage for a normal distribution). In commercial testing and compliance reporting, a 68% confidence level is inadequate.
Laboratories multiply uc by a numerical multiplier called the coverage factor (k) to define an expanded interval enclosing the true value with a high level of confidence:
Corresponds to an approximate 95% level of confidence (95.45% exactly for a normal distribution with infinite degrees of freedom). This is the universally mandated default for NABL and ISO/IEC 17025 test reports.
Corresponds to an approximate 99.73% level of confidence, used in extreme-consequence nuclear or aerospace metrology where false conformity risks catastrophic failure.
9. Expanded Uncertainty (U = k · uc)
The final metric reported to customers is the Expanded Uncertainty (U):
Expanded uncertainty defines the symmetric half-width of the confidence interval around the reported measurement result: [ y - U, y + U ].
10. Reporting Measurement Uncertainty under ISO/IEC 17025 Clause 7.8.3.1
When reporting measurement uncertainty on a Certificate of Analysis, laboratories must adhere to strict metrological expression rules:
Standard Reporting Syntax
Mandatory footnote statement required by NABL 141 and GUM:
"The reported expanded uncertainty of measurement is stated as the combined standard uncertainty of measurement multiplied by the coverage factor k = 2, which for a normal distribution corresponds to a coverage probability of approximately 95%."
The Two Significant Figures Rounding Rule
Rule 1: The expanded uncertainty U should be rounded to at most two significant figures (e.g. ± 0.046 mg/l, not ± 0.046382 mg/l).
Rule 2: The numerical value of the test result must be rounded to the same number of decimal places as the uncertainty (e.g. 214.2 mg/l, matching the tenths digit of 4.6 mg/l).
11. Practical Worked Example: Analytical Balance & Volumetric Flask
To illustrate the entire evaluation workflow, consider a practical gravimetric determination in a testing laboratory: preparing a standard chemical solution by weighing 1.0000 g of reference salt on a 4-decimal analytical balance and dissolving it in a 100.0 ml Grade-A volumetric flask.
Hypothetical Laboratory Data & Uncertainty Budget Table
| Uncertainty Source | Quoted Value | Distribution | Divisor | Standard Uncertainty u(x) |
|---|---|---|---|---|
| Balance Calibration Cert (U) | ± 0.0004 g (k=2) | Normal | 2 | 0.00020 g |
| Balance Repeatability (s, 10 reps)* | s = 0.00015 g | Normal (Type A) | 1 | 0.00015 g |
| Flask Calibration Tolerance | ± 0.10 ml | Triangular | √6 ≈ 2.449 | 0.0408 ml |
| Temperature Variation (±4°C) | ± 0.084 ml (expansion) | Rectangular | √3 ≈ 1.732 | 0.0485 ml |
*Note on repeatability divisor: This budget assumes one routine weighing per analytical sample, applying the historical repeatability standard deviation directly as urep = s / √1 = s (divisor 1). If an analytical procedure specifies that the reported result is the arithmetic average of multiple independent weighings, the appropriate divisor must reflect the actual measurement design by dividing by √n.
Step 1: Calculate Relative Uncertainties for Mass and Volume
Combined relative uncertainty for Mass: urel(m) = √[(0.00020/1.0000)² + (0.00015/1.0000)²] = 0.00025 (0.025%)
Combined relative uncertainty for Volume: urel(V) = √[(0.0408/100)² + (0.0485/100)²] = 0.000634 (0.0634%)
Step 2: Combined Relative Standard Uncertainty for Concentration C = m / Vuc,rel(C) = √[ (0.00025)² + (0.000634)² ] = 0.000681 (0.0681%)
Step 3: Expanded Uncertainty at Nominal C = 10.000 g/luc(C) = 10.000 × 0.000681 = 0.00681 g/lU = 2 × 0.00681 = 0.01362 g/l rounded to 2 sig figs: 0.014 g/l
12. Common Calculation & Reporting Errors Cited by NABL Assessors
During laboratory assessments under ISO/IEC 17025, technical assessors frequently issue non-conformances for arithmetic and conceptual errors:
1. Adding Expanded Uncertainties Directly
Adding U values directly (linear addition) instead of converting to standard uncertainties and performing root-sum-of-squares addition under the law of propagation.
2. Mismatched Decimal Places
Reporting a measurement result like 14.285 mg/l ± 0.12 mg/l. The measured result has three decimal digits while the uncertainty has only two. Both must be rounded to the same decimal position (i.e., 14.29 mg/l ± 0.12 mg/l).
3. Using Repeatability s Instead of Mean Uncertainty
Confusing the standard deviation of an individual run s with the standard error of the mean s / √n when reporting an averaged duplicate analysis.
13. Maintaining Uncertainty Budgets & Audit Documentation
Under NABL 141 guidelines, laboratories must maintain formal, version-controlled Uncertainty Budgets for every test parameter on their accredited scope:
- Detailed mathematical measurement model and sensitivity equation.
- Documented rationale for all assumed probability distributions (rectangular, triangular, normal).
- Re-evaluation triggers: Mandatory recalculation whenever an instrument is recalibrated, a balance is relocated, a reference standard changes, or a significant method modification occurs.
- Annual review during internal audits and management reviews under Clause 8.5 & 8.9.
14. How Laboratory Information Management Software (LIMS) Supports Uncertainty Workflows
Evaluating and maintaining measurement uncertainty manually across dozens of test scopes across multiple analysts creates significant audit vulnerability. Modern laboratory management software provides structured operational support:
Centralized Uncertainty Budget Storage
Store version-controlled uncertainty budgets linked directly to parameter test methods, calibration records, and instrument asset IDs.
Automated Decimal Matching & Rounding
Automatically rounds expanded uncertainty to two significant figures and synchronizes decimal places of measured test results upon report generation.
Decision Rule & Guard Band Flagging
Flags borderline compliance results when the expanded uncertainty band overlaps statutory specification limits under ISO/IEC 17025 Clause 7.8.6.
Instrument Calibration Guard Rails
Prevents test results from being approved if the analytical instrument has an expired calibration certificate or missing uncertainty budget link.
Explore ISO/IEC 17025 software management capabilities in our ISO 17025 Laboratory Software Overview.
Frequently Asked Questions on Measurement Uncertainty
Direct answers to common technical and audit questions regarding ISO/IEC 17025:2017 Clause 7.6.
Is every testing laboratory required to report measurement uncertainty on all test reports under ISO/IEC 17025:2017?
No. Under ISO/IEC 17025:2017 Clause 7.8.3.1(c), testing laboratories must evaluate measurement uncertainty, but are only required to report it on the Certificate of Analysis when it is relevant to the validity or application of the test results, when requested by the customer, or when the uncertainty affects conformity to a specification limit (decision rules under Clause 7.8.6).
What probability distribution should be used for manufacturer equipment tolerances when no distribution is specified?
When a manufacturer specification or calibration certificate states a tolerance limit as ±a without specifying a probability distribution or confidence level, metrological guidance (GUM / JCGM 100:2008) requires assuming a rectangular (uniform) distribution, where the standard uncertainty is calculated as u = a / √3.
Why is k=2 commonly chosen as the coverage factor?
For a normal (Gaussian) distribution with a sufficiently large effective degrees of freedom, a coverage factor of k = 2 corresponds to a coverage probability (level of confidence) of approximately 95.45% (conventionally stated as approximately 95%). If degrees of freedom are small, the Student's t-distribution via the Welch-Satterthwaite equation must be applied.
How many significant figures should be reported for expanded measurement uncertainty?
According to NABL 141 and international metrological conventions, the numerical value of the expanded uncertainty U should be rounded to at most two significant figures. The numerical value of the measurement result must then be rounded to the same number of decimal places as the uncertainty.
Maintain Audit-Proof ISO 17025 Quality Records
Connect equipment calibrations, uncertainty budgets, control charts, and Section 7.11 data integrity records within ExactLIMS cloud software.