Pipettes, balances, temperature probes, centrifuges… The reliability of laboratory results directly depends on the control of its measuring equipment. Calibration establishes a relationship between the instrument's readings and reference values, along with their uncertainties. It does not automatically constitute a conformity verification: this requires predefined specifications and a decision rule. Choosing the traceability route, uncertainties, maximum permissible errors, statement of conformity: behind calibration lies a much more fundamental question: what measurement does the laboratory really need?

In a laboratory, metrology can sometimes be perceived as a series of obligations: having equipment calibrated, retrieving its certificate, checking compliance, and scheduling the next due date. But this approach misses the essential point. Metrology is not simply about having an instrument calibrated. It is about ensuring that this instrument provides sufficiently reliable information for its intended use.

A pipette used to dispense a few microliters, a probe monitoring a thermal chamber, or a centrifuge used in a biological protocol do not meet the same requirements. Even before talking about calibration, you must therefore start by defining your needs.

When equipment needs to be calibrated, three main traceability routes can be considered.

Performing an in-house calibration requires, in particular, demonstrating personnel competence, having traceable reference standards, controlling environmental conditions, documenting methods, evaluating uncertainties, and monitoring the validity of results. In other words, doing it yourself does not mean doing it more simply.

Using an accredited laboratory provides recognized competence for activities covered by its scope of accreditation, provided that the service and certificate are indeed issued under accreditation. However, this does not mean that the user laboratory can delegate all of its metrological thinking. The service provider knows how to calibrate. The client, on the other hand, knows the actual use of their equipment. It is therefore up to them to precisely express their needs.

Three questions are particularly essential: what uncertainty is actually required? Does the service need to be performed under accreditation? Is a statement of conformity expected and, if so, against which specifications? Because systematically asking for "the best" service is not necessarily relevant.

The lower the target uncertainty, the greater the resources required may be—and the more expensive the service can become. The goal is therefore not to achieve the lowest possible uncertainty, but rather an uncertainty compatible with the use of the instrument.

This reflection leads directly to the question of maximum permissible errors, or MPE. How do you decide at what level of error equipment becomes unacceptable? Several approaches exist. The laboratory can use the specifications provided by the manufacturer. It can also rely on the values defined by a standard or a technical framework.

These solutions have the advantage of being simple and immediately available. However, they do not necessarily reflect the actual use of the equipment. A manufacturer's specification generally indicates the expected performance of a product. A standard may set performance levels corresponding to a specific use, method, or context, which do not necessarily cover the actual conditions of use of the equipment in the laboratory. Yet the laboratory's need may be different.

The most relevant approach is therefore, whenever possible, to define MPEs based on the actual impact of measurement error on the analytical process. If a pipette delivers a volume with a 2% error, does this error significantly alter the final analytical result? And at what error threshold does the result become scientifically or clinically problematic? It is this information that should ideally determine the expected performance of the equipment.

No measurement is perfectly exact.

Each result is therefore associated with a measurement uncertainty, which characterizes the dispersion of the values that could reasonably be attributed to the measurand. In metrology, its estimation is based in particular on the principles of the GUM: the Guide to the Expression of Uncertainty in Measurement.

Different sources of uncertainty are identified and then combined: repeatability, instrument resolution, reference standard, environmental conditions, method used, etc. The result can then be associated with an expanded uncertainty, obtained by multiplying the combined standard uncertainty by a coverage factor. A factor k close to 2 often corresponds, under certain distribution and degrees-of-freedom assumptions, to a coverage probability of approximately 95%. However, this uncertainty is not just a figure to include on a certificate. It helps us understand the level of confidence that can be placed in the measurement.

The question becomes particularly interesting when an equipment needs to be declared compliant or non-compliant. One possible decision rule is to take uncertainty into account by applying a guard band before declaring compliance. However, other rules may be adopted depending on the accepted level of risk, contractual requirements, or the applicable standards.

However, the closer the result is to the tolerance limit, the more critical the decision becomes. Let's imagine a limit value set at 55 and a calibration result exactly equal to 55.

If the uncertainty surrounding this measurement is taken into account, the situation is not as binary as a simple "pass" or "fail". The interval associated with the result can then cover values located on either side of the specification limit. The decision therefore depends on the rule applied and the risk of false acceptance or false rejection that the laboratory is willing to take. The statement of conformity then also becomes a matter of risk and confidence level. This approach makes it possible to go beyond a purely binary reading of the certificate and integrate risk into the metrological decision.

This is probably one of the most important points for the end user.

Let's take a micropipette. It is calibrated under controlled conditions, using a defined method, by a qualified operator, and in a controlled environment. In the user's laboratory, conditions may be different: room temperature, liquid temperature, nature of the sampled liquid, pipetting technique, user, frequency of use, etc. Furthermore, the drift of the instrument over time is not necessarily reflected in the uncertainty reported during its calibration.

Depending on the application and the criticality level of the measurement, it may therefore be necessary to supplement the certificate's information with contributions related to actual operating conditions in order to estimate the measurement uncertainty in use. The same applies to a temperature probe: an instrument calibrated in a liquid bath may behave differently when subsequently used in air.

The certificate is therefore an essential basis, but it does not tell the whole story of the measurement.

In-sourcing is sometimes seen as a way to reduce costs. However, this is not automatically the case. Properly carrying out your metrology requires reference equipment, its own calibration traceability, procedures, competent personnel, time, uncertainty calculations, method validation, and monitoring mechanisms.

The real benefit of in-sourcing often lies elsewhere: control and responsiveness.

A laboratory with the necessary skills can quickly check an instrument, adapt its methods to its activity, and better understand the behavior of its equipment inventory. However, if the skills or resources are not available, outsourcing can be much more relevant. The right question is therefore not "which solution is the cheapest?", but rather: which organization allows for the effective management of metrological risk?

This approach also leads to questioning fixed intervals. When regulatory or standard requirements allow it, why automatically maintain a 12-month interval if documented history shows that the equipment drifts very little?

Conversely, why wait a year when equipment shows significant drift after just a few months? Monitoring the equipment fleet allows intervals to be tailored based on actual data. Comparisons between similar instruments, reference instruments, repeated calibrations, intermediate checks, drift tracking, or interlaboratory comparisons can all contribute to this monitoring.

A comparison can also be organized between two laboratories. This bilateral comparison can provide useful information on the compatibility of their results, provided that the protocol, the stability of the compared artifact, and the uncertainties are properly controlled. However, it does not necessarily have the same scope as an interlaboratory comparison or a proficiency test organized under a formalized framework.

Behind all these technical questions, a fairly simple principle ultimately emerges. The laboratory must not simply endure its metrology. It must remain the decision-maker.

It is up to the laboratory to determine what it measures, why it measures it, what error is acceptable, what uncertainty is required, how often its equipment should be checked, and what level of confidence it wants to associate with its decisions. The calibration certificate then ceases to be merely documentary proof to be archived, but becomes a source of information enabling a better understanding and control of its equipment. Because metrology is not just about demonstrating that an instrument has been calibrated.

It consists in demonstrating that the measurement produced is sufficiently reliable to make the right decision.