Measurement uncertainty is one of the most misunderstood requirements in ISO/IEC 17025 — and one of the most frequently cited during accreditation assessments. The core concept is straightforward: every measurement result is an estimate, not an exact truth, and the uncertainty tells you how wide the interval around that estimate might reasonably be. ISO/IEC 17025 requires laboratories to identify and, where relevant, report this uncertainty alongside their results.
This article explains what measurement uncertainty is, why the standard requires it, how to build an uncertainty budget, and how to express uncertainty in reports — with particular attention to calibration laboratories, where uncertainty reporting is most rigorous.
What Measurement Uncertainty Actually Means
When a laboratory reports that a sample contains 12.4% THC, or that a reference mass weighs 100.0012 g, that number is the best estimate the measurement process can produce. But the measurement process itself — the instrument, the method, the analyst, the environment, the reference standard — introduces variability. Measurement uncertainty quantifies that variability and expresses it as a range: the true value is likely to lie within that range, at a stated level of confidence.
A result expressed with uncertainty looks like this: 12.4% ± 0.3% (k=2, approximately 95% confidence level). The ± 0.3% is the expanded uncertainty. The k=2 is the coverage factor, which corresponds to roughly 95% confidence assuming a normal distribution of errors.
Uncertainty is not the same as error. Error is the difference between a measurement and the true value — something you can only know if you have a perfect reference. Uncertainty is a statement about the dispersion of possible values given what you know about your measurement process. You calculate uncertainty from what you know, not from what you cannot know.
Why ISO/IEC 17025 Requires It
ISO/IEC 17025 requires laboratories to identify all contributions to measurement uncertainty and, where relevant, report uncertainty alongside results. The rationale is practical: a test result without uncertainty information is incomplete. A customer making a compliance decision — is this pesticide residue above or below the action limit? — needs to know whether the measurement is precise enough to support that decision. Without uncertainty, the result is a number without context.
For calibration laboratories, reporting measurement uncertainty is essentially always required. For testing laboratories, the requirement applies where uncertainty is relevant to the validity or application of the test result — which in practice covers most quantitative measurements. Testing labs sometimes argue that a method-prescribed result already “contains” uncertainty implicitly; assessors generally expect explicit documentation of the uncertainty budget even if expanded uncertainty is not routinely reported on certificates.
The Two Types of Uncertainty Components
ISO/IEC 17025 follows the internationally harmonized approach set out in the Guide to the Expression of Uncertainty in Measurement (GUM), which distinguishes two types of evaluation:
Type A Evaluation
Type A uncertainty components are estimated from statistical analysis of repeated observations. If you run the same calibration ten times and calculate the standard deviation of the results, that standard deviation (or the standard deviation of the mean, depending on how results are averaged) is a Type A component.
Type A evaluation is straightforward but requires sufficient repeat measurements to be statistically meaningful. A standard deviation based on three replicates is far less reliable than one based on twenty.
Type B Evaluation
Type B uncertainty components come from any source other than repeated observations — reference certificates, published specifications, calibration data, expert judgment, or prior knowledge. Examples include:
- The stated uncertainty of a certified reference material (CRM)
- The resolution of a digital display (typically ±0.5 of the last digit, treated as a rectangular distribution)
- The calibration certificate uncertainty of a reference standard
- Temperature coefficient of a resistor across the measurement range
- Known bias from a matrix effect
Type B evaluation requires the analyst to convert the stated uncertainty or specification into a standard uncertainty using an assumed probability distribution (normal, rectangular, triangular, etc.).
Building a Measurement Uncertainty Budget
An uncertainty budget is a structured list of all significant uncertainty sources, their individual contributions expressed as standard uncertainties, and their combination into a combined standard uncertainty. The process has five steps:
1. Define the measurand
State precisely what you are measuring — the quantity, the unit, and the conditions under which the measurement is made. The uncertainty is specific to this measurement definition.
2. Identify uncertainty sources
Systematically list every factor that can cause variation in your result. Common sources for a laboratory measurement include:
- Repeatability of the measurement (Type A)
- Reproducibility across analysts, days, or instruments (Type A or Type B)
- Calibration uncertainty of the reference standard or CRM used (Type B)
- Instrument resolution (Type B)
- Environmental effects — temperature, humidity, vibration (Type B)
- Sample preparation steps (Type B or Type A)
- Matrix effects (Type B or Type A from spike recovery studies)
Use a cause-and-effect (Ishikawa) diagram or a systematic sweep of the measurement equation to ensure completeness. A missing source is a hidden bias.
3. Quantify each source as a standard uncertainty
Convert each contribution to a standard uncertainty (u), which is equivalent to one standard deviation:
- For Type A components: u = s/√n (standard deviation divided by square root of number of replicates)
- For Type B rectangular distributions: u = a/√3 (half-width divided by √3)
- For Type B normal distributions: u = stated uncertainty / coverage factor k
4. Combine the components
Combine all standard uncertainties using the law of propagation of uncertainty. For independent, additive sources, the combined standard uncertainty is the square root of the sum of the squares of the individual standard uncertainties:
u_c = √(u₁² + u₂² + u₃² + …)
If the uncertainty sources are not additive (e.g., they are multiplied in the measurement equation), apply the appropriate sensitivity coefficients.
5. Calculate and report expanded uncertainty
Multiply the combined standard uncertainty by a coverage factor k to obtain the expanded uncertainty U:
U = k × u_c
For most laboratory measurement uncertainty statements, k = 2 (approximately 95% confidence assuming normality). Report the result as: result ± U (unit), k = 2.
Reporting Uncertainty in Test and Calibration Certificates
ISO/IEC 17025 requires that calibration certificates always include measurement uncertainty. For test reports, uncertainty must be included when it is relevant to the validity or interpretation of the test result.
A compliant uncertainty statement should include:
- The expanded uncertainty value and unit
- The coverage factor k
- The confidence level (e.g., approximately 95%)
- A brief statement that uncertainty was calculated in accordance with recognized methodology (e.g., the GUM)
Avoid vague statements like “results are accurate to ±5%.” Instead, be specific: “The expanded measurement uncertainty is ±0.8 mg/kg, calculated at k=2 (approximately 95% confidence).”
For calibration laboratories, uncertainty must also be traceable through the calibration chain — meaning the uncertainty of your reference standards propagates into and contributes to the uncertainty of the calibrations you perform. This is why calibration lab accreditation demands particular attention to traceability and the uncertainty budgets of reference standards.
Practical Tips for Getting Uncertainty Right
- Start with repeatability studies: Collecting repeat measurement data under realistic conditions is the fastest way to anchor your budget with defensible Type A components.
- Don’t neglect small contributors: While it is acceptable to exclude sources that contribute less than 1% of the combined uncertainty, document why they were excluded rather than simply omitting them.
- Update your budgets when methods change: A new instrument, new analyst, or new sample preparation step may change the uncertainty. Treat uncertainty budgets as living documents tied to your method validation records.
- Use the ISO/IEC 17025 documentation kit as a template framework for uncertainty budget worksheets — but populate every cell with data from your own measurements.
Use our free accreditation readiness checklist to confirm that your uncertainty documentation is assessment-ready before submitting your application.
Frequently Asked Questions
Q: Is measurement uncertainty required for qualitative tests? A: No. Measurement uncertainty applies to quantitative results — where a numerical value and unit are reported. For qualitative tests (e.g., presence/absence microbiology), the concept of measurement uncertainty does not apply in the same way, though concepts like detection probability and misclassification rates serve an analogous function.
Q: What is the difference between measurement uncertainty and method precision? A: Precision (repeatability, reproducibility) is one component that feeds into a measurement uncertainty budget. Uncertainty is broader — it combines precision with contributions from calibration standards, reference materials, environmental factors, and all other known sources of variability. Precision alone understates the true uncertainty.
Q: Can I use an empirical “top-down” approach instead of a component-by-component budget? A: Yes. The top-down approach estimates combined uncertainty directly from interlaboratory comparison data, proficiency testing results, or long-term precision studies that encompass all major variation sources. It is particularly common in chemical testing labs. It is acceptable under ISO/IEC 17025 provided the data adequately represents the range of conditions under which the method is used.
Q: How do I handle uncertainty when my measurement result is near a specification limit? A: This is a conformance decision. ISO/IEC 17025 requires that when a customer requests a statement of conformance (pass/fail), the lab must clearly state which decision rule was applied — i.e., whether uncertainty was taken into account and how. The lab’s role is to report the result and uncertainty and apply the agreed decision rule; it is not to hide uncertainty to make a borderline result look cleaner.
Q: How often should measurement uncertainty budgets be reviewed? A: Review uncertainty budgets whenever there is a significant change to the measurement process — new equipment, method modification, change in reference material supplier, or evidence from proficiency testing that performance has shifted. At minimum, they should be reviewed as part of the periodic method revalidation cycle.