Measurement uncertainty in the industrial laboratory – understand it, document it, master it with LIMS

What is measurement uncertainty – and why is it indispensable?

Whether in production, in the laboratory or in quality assurance: measurements provide the basis for decisions on releases, complaints and process optimisation. But how accurate are measurement results really? And how confident can you be that a measured value reflects reality?

This is where the concept of measurement uncertainty comes in. It quantifies the unavoidable dispersion of every measurement. Anyone who measures a carbon content of 0.42% with a spectrometer does not in fact receive “the value”, but rather a value plus or minus an interval – for example 0.42% ± 0.01%. Stating the uncertainty is not optional, but an integral part of the measurement result.

Three reasons make measurement uncertainty mandatory in the industrial laboratory:

• Comparability – only with documented uncertainty are values comparable across instruments, sites and time
• Basis for decisions – does the value really lie within tolerance, or only apparently? The answer depends on the uncertainty
• Standards compliance – ISO 9001, ISO/IEC 17025 and ISO/IEC Guide 98-3 require its systematic determination and documentation

Accuracy, precision, trueness – the terms cleanly distinguished

In everyday practice these terms are often used interchangeably – strictly speaking, however, they refer to different things. ISO 5725 defines them clearly:

Important: accuracy is a qualitative concept, measurement uncertainty a quantitative one. Anyone who says “our method is accurate” is making a statement that cannot be verified without a number. Anyone stating the measurement uncertainty as “± 0.03%” delivers something testable.

Two further terms belong to the same family and are closely tied to measurement uncertainty:

• Metrological traceability – every measurement has to be traceable, via an unbroken chain of calibrations, back to national or international standards (e.g. PTB, NIST). Each calibration step contributes its share to the total measurement uncertainty.
• Method validation – the proof that a test method is fit for its purpose. Among other things, this delivers the measurement uncertainty for the validated method.

Sources of measurement uncertainty – where it arises

Measurement uncertainty does not arise in a single place but is composed of numerous sources. Anyone wishing to assess it systematically has to know these sources:

In practice, a single source often dominates – frequently sampling or calibration. Anyone tackling that source improves the total measurement uncertainty most strongly.

Determining the measurement uncertainty: Type A and Type B

The international standard is the ISO/IEC Guide 98-3 (GUM – Guide to the Expression of Uncertainty in Measurement). It distinguishes two types of evaluation:

The individual standard uncertainties are then combined into the combined standard uncertainty u_c using the law of propagation of uncertainty (root sum of squares). Multiplied by a coverage factor k (typically k=2 for a confidence level of approximately 95%), this yields the expanded measurement uncertainty U.

This produces a statement such as: Carbon content: (0.42 ± 0.02)% at k=2. This notation is standards-compliant and audit-ready.

Measurement uncertainty in audits – what auditors look for

In audits under ISO 9001 or ISO/IEC 17025, measurement uncertainty is a systematically reviewed area. Auditors specifically want to know:

1. 1
   Has the measurement uncertainty been determined? For every relevant method, a documented measurement uncertainty has to be available – not just “roughly estimated”, but determined according to a verifiable procedure.
2. 2
   Is it acceptable for the intended use? A method with ± 0.5% uncertainty is unsuitable for a tolerance band of ± 0.1%. This fitness-for-purpose assessment has to be documented.
3. 3
   Is traceability demonstrated? Which reference materials were used? When was the instrument last calibrated? What was the uncertainty of that calibration?
4. 4
   Are measurement uncertainty and conformity assessment linked? If a measured value lies only just within tolerance, the measurement uncertainty has to feed into the release decision. This linkage is audit-relevant.

Anyone who can answer these four questions with documented evidence from the LIMS closes this audit block in minutes. Anyone having to reconstruct them from Excel files, instrument logs and handwritten notes loses hours – and credibility.

What role does a LIMS play in measurement uncertainty?

Measurement uncertainty is not a one-off project but a continuous obligation: every method, every instrument, every calibration contributes to the uncertainty – and this has to be maintained, documented and factored into evaluations on an ongoing basis. This is exactly where a LIMS comes in.

[FP]-LIMS supports measurement uncertainty management at several points:

• Method management with uncertainty stored alongside – every validated method carries its measurement uncertainty. Versioning ensures that the uncertainty valid at the time of measurement remains documented.
• Calibration management with calibration history – when was the instrument calibrated, against which standard, with what uncertainty of the calibration? All retrievable in the LIMS.
• Traceability across the entire chain – sample → instrument → calibration standard → primary standard. The metrological traceability is anchored in the database.
• Competence management – which operator is qualified for which method? Reduces the operator-related component of the uncertainty.
• Conformity assessment with tolerance logic – values outside the specified tolerance are flagged automatically (colour-coded: green/yellow/red). The measurement uncertainty can be included in the assessment.
• Audit trail – every change to method, calibration or measurement data is traceable with timestamp, user and reason.
• Reporting with uncertainty statement – test reports include the measured value and the expanded measurement uncertainty U with coverage factor k – audit-ready.

Practical example: OES analytics in the metals industry

A concrete example makes tangible why measurement uncertainty is decisive in industrial day-to-day work. A laboratory technician runs spark optical emission spectrometry on an OES spectrometer to determine the carbon content of a steel batch. The specification calls for 0.40% ± 0.05% C.

The spectrometer delivers: 0.42%. Within the specification, released? Not without further ado:

• Without measurement uncertainty: 0.42% lies between 0.35% and 0.45% – formally fine.
• With measurement uncertainty U = 0.02% at k=2: the true result lies, with 95% confidence, between 0.40% and 0.44%. Still in specification.
• With measurement uncertainty U = 0.06% at k=2: the true result lies, with 95% confidence, between 0.36% and 0.48%. The upper tolerance limit (0.45%) lies inside the uncertainty range – the release is no longer clear-cut.

In practice, we see at customers such as AGOSI (precious metal processing, ISO/IEC 17025-accredited since 2012) how this logic is mapped directly in the LIMS: method uncertainties are stored, the conformity assessment takes them into account, and the test reports contain all three values – measured value, uncertainty, conformity statement. The audit becomes routine instead of a special project.

Typical pitfalls in measurement uncertainty management

From over 30 years of practice with industrial laboratories we see recurring mistakes – and all of them are avoidable:

• Measurement uncertainty only “estimated globally” – without a documented, GUM-compliant determination path, this gets flagged in audits. Better to start pragmatically with Type B than not at all.
• Sampling not included in the uncertainty – a classic. The instrument uncertainty is determined carefully, sampling is left out – although sampling often delivers the largest share.
• Outdated method versions with old uncertainty – the method was changed, the uncertainty not re-determined. With versioning in the LIMS this does not happen.
• Calibration history distributed across Excel and instrument software – when the auditor asks, there is search effort instead of a LIMS query.
• Measurement uncertainty not in test reports – values without uncertainty are incomplete. Reports should include it as standard.
• Conformity assessment without considering the uncertainty – values close to the limit are formally released, although the uncertainty may breach the spec. Risk of complaints and liability.

Frequently asked questions on measurement uncertainty