In quality control, production, or research, comparing analytical results is part of everyday practice.
But what happens when two professionals — each confident in the reliability of their work — observe a discrepancy between their measurements?
It’s a common situation, often leading to tension, misunderstanding… or rushed decisions.

Example: Two laboratories measure the alcohol content of the same spirit sample. One reports 40.1% vol, the other 39.8% vol.
Who’s right? Are the results truly incompatible — or simply different within the range of measurement uncertainty?

Before drawing conclusions, it’s essential to ask the right questions. Behind every result lies a method, an instrument, an operator, a degree of uncertainty—and sometimes even a different definition of the parameter being analyzed.

1. Understanding the Source of Discrepancies: The Right Questions to Ask

• Were both measurements taken on the same sample, at the same time?
• Were the sampling and storage conditions equivalent?
• Are the units the same?

Once these points have been checked, the investigation can continue with the following questions:

• Is the definition of the parameter the same?
• Are the analytical methods and the precision of the equipment the same?
• Are these methods appropriate for the analysis?
• Are the instruments properly calibrated, with verified traceability?

If the answers to the previous questions do not explain the observed discrepancy, attention must then turn to measurement uncertainty, with two essential questions to ask:

• Are the measurement uncertainties known, calculated, and documented?
• Are the results compatible in light of the stated uncertainties?

Answering these last two questions requires a clear understanding of how to interpret measurement uncertainty, an aspect often underestimated.
Yet these answers determine the ability to make an objective decision, one that can have significant financial or commercial consequences: product rejection, production stoppage, quality.

2. Example of Two Compatible Results Despite a Discrepancy

Two different results can be considered equivalent if their uncertainty intervals overlap.

Example :

• Result 1: 40,1 % vol. ± 0,2 % → Interval: [39,9 ; 40,3]
• Result 2: 39,8 % vol. ± 0,2 % → Interval: [39,6 ; 40,0]
• Overlapping zone: [39,9 ; 40,0]

To interpret a discrepancy correctly, it is essential that results are accompanied by their measurement uncertainty and that this uncertainty is both known and well understood.

3. But what kind of uncertainty are we really talking about?

Not all uncertainties are created equal. To objectively compare two analytical results, relying solely on the method’s repeatability is not enough.
The relevant uncertainty is an expanded uncertainty that accounts for all sources of variability—including those identified through interlaboratory studies (1), not just internal testing.

It’s possible to be precise and repeatable, yet still inaccurate.

That’s why it’s essential that:

• the operator is properly trained in the method,
• the measurement system is regularly checked to detect any drift,
• uncertainties are documented, justified, and traceable.

4. When results are calculated from multiple measurements

In some cases, the evaluation of uncertainty becomes more complex, especially when the result is based on several combined measurements.

Example

A result discrepancy on pure alcohol volume, between the delivery note and the reception → if the results are accompanied by an uncertainty, it will be a valuable aid for decision-making

5. When regulations impose a fixed limit: what to do in case of a discrepancy with the regulatory value

Managing a discrepancy becomes particularly delicate when a regulatory threshold is involved. this is the case, for example, with spirits benefiting from a designation of origin, which require minimum or maximum thresholds for certain parameters.

Example:

My internal control, or the one carried out by an external laboratory, indicates an alcohol content of 39.8 % vol. ± 0.2 %, while regulations set a minimum of 40.0 % vol..

The associated uncertainty interval [39.6 ; 40.0] does include the regulatory limit, but the central value remains below that threshold.

Am I compliant with the regulation?

 What happens in the case of an official inspection?

1 If the value found by the inspection body is 39.8% vol. with a measurement uncertainty of ± 0.2% vol., the interval in which the true value lies is therefore between [39.6 ; 40.0]. There is a non-zero probability that its “true value” is indeed 40.0% vol. → The product can be considered compliant within the bounds of uncertainty.

2 However, if the same result is reported but with an uncertainty of ± 0.15% vol., the interval in which the true value lies is then [39.65 ; 39.95]. In this case, the 40.0% limit is no longer covered by the interval → The product will be deemed non-compliant.

🔴 If the uncertainty associated with your measurement result is not at least as low as that of official laboratories, it is wise to apply a safety margin. This ensures that, in the event of an inspection, your results will remain compliant within the uncertainty limits of the official laboratory.

Official inspection laboratories are required to operate under ISO 17025. This standard requires laboratories to evaluate measurement uncertainty and to provide it whenever it is relevant for interpreting the result—especially in cases of regulatory compliance.

6. Conclusions on measurement uncertainty

In the event of a dispute over an analytical parameter—whether with a supplier, a customer, or an official inspection body—it is crucial to have results accompanied by their uncertainty, whether they come from your own laboratory or an external provider.

If this expertise is not available in-house, it is recommended to use an ISO 17025-accredited laboratory for analyses subject to regulatory or contractual requirements. Upon request, such a laboratory will provide results along with their measurement uncertainty, ensuring both reliability and credibility.

A well-calculated, documented, and justified uncertainty is a technical defense tool in the event of disagreement over an analytical parameter.

Not knowing your uncertainties is giving others the power to decide for you.

7. In case of dispute: interpret differences with rigor (and humility)

A discrepancy is not necessarily a problem. It may reflect:

• a difference in methodological approach,
• natural variation in the product,
• or controlled variability within acceptable uncertainty limits.

The goal is not to “find someone to blame,” but to understand and improve data comparability.

 Collaborer rather than oppose: resolving an analytical discrepancy is often an opportunity to exchange on practices, strengthen processes, and build mutual trust.