What Should Happen When Calibration Finds Equipment Out of Tolerance?

An out-of-tolerance result is not simply a maintenance finding. It is evidence that, at the time of calibration, the equipment did not meet its defined acceptance criteria. The immediate question may be whether to adjust or repair the instrument, but the more consequential question is whether results obtained before discovery remain suitable for their intended use.

A controlled response must distinguish three matters that are often combined:

• the calibration result and its uncertainty;
• the conformity decision against the applicable acceptance limits and decision rule; and
• the potential effect on measurements made since the equipment was last known to be acceptable.

The right response is proportionate to risk, but it should be technically supported and recorded. Treating every out-of-tolerance finding as proof that all prior work is invalid is usually unjustified. Treating it as irrelevant because the equipment has now been adjusted is equally difficult to defend.

Contain the equipment and preserve the evidence

Remove the equipment from service, or clearly identify it as restricted, as soon as the out-of-tolerance condition is confirmed. This prevents further uncontrolled use while the laboratory decides what work, if any, can continue under an approved alternative arrangement.

Before adjustment, repair or reset, preserve the as-found evidence. The as-found condition is central to any retrospective assessment. Record the calibration points, measured errors or corrections, uncertainty, environmental conditions where relevant, reference standards used, method revision, operator and date. If the instrument has a configuration that affects its result, such as range selection, probe type, software version, scaling factors or a connected sensor, retain that information too.

Adjustment may be appropriate, but it should not obscure the original condition. After adjustment or repair, perform an as-left calibration or verification sufficient to demonstrate the restored state. The as-left result answers whether the equipment is now fit for return to service. It does not answer whether earlier results were affected.

Confirm what “out of tolerance” means

The laboratory must first establish the basis of the finding. A measured error of +0.82 °C is not inherently out of tolerance. It becomes a conformity issue only when compared with defined acceptance limits using a stated or otherwise agreed decision rule.

For example, a process thermometer may have an acceptance limit of ±0.75 °C at 100 °C. If its measured error is +0.82 °C with an expanded uncertainty of 0.10 °C, the conformity decision depends on the decision rule. With a shared-risk rule based on simple acceptance at the tolerance boundary, it may be accepted. With a guard band that requires the result, including the relevant uncertainty allowance, to fall within a reduced acceptance interval, it may be rejected.

The retrospective investigation must use the same technical basis. Do not describe an item as out of tolerance without retaining the applicable specification, acceptance limits, decision rule and calibration result. These determine both the significance of the finding and the possible extent of prior impact.

Assess the period and pattern of possible nonconformance

The last satisfactory calibration date is a useful starting point, not automatic proof that the instrument remained acceptable until that date. The assessment should consider evidence of stability and the likely onset of the condition.

Relevant evidence may include:

• previous calibration history and drift trend;
• intermediate checks, check standards or control charts;
• damage, overload, contamination, battery failure, repair or other events;
• user observations, complaints and anomalous results;
• environmental exposure outside the equipment’s intended conditions; and
• the magnitude, direction and measurement-point dependence of the as-found error.

A stable balance that exceeds its acceptance limit by a small amount after a full interval presents a different case from a pressure transducer with a sudden zero shift following an overpressure event. In the first case, historical drift and intermediate checks may support a limited impact assessment. In the second, the review period may need to begin at the event date or, if that date cannot be established, extend to the last demonstrated satisfactory condition.

The investigation should also consider where the equipment was used. A deviation at one point does not necessarily affect all applications. A digital multimeter found high by 0.15 % on a 10 V range may have no material relevance to continuity checks, but it may affect calibrations of voltage sources performed on that range.

Evaluate the impact on previous results

ISO/IEC 17025:2017 clause 7.10 requires a laboratory to evaluate the significance of nonconforming work, including an analysis of the impact on previous results where appropriate. An out-of-tolerance instrument can trigger this process, but the investigation should be driven by potential measurement impact rather than by paperwork alone.

Identify the affected jobs, batches, certificates, test reports or internal decisions within the justified review period. For each relevant use, compare the equipment’s potential error with the measurement method, uncertainty budget and decision being supported.

Suppose a reference pressure gauge was found to have a +0.12 bar error at 10 bar. If it was used to set a process at 10.0 bar with an allowable range of 9.0 to 11.0 bar, that error may not change the operational decision. If it was used as a reference in a calibration with a ±0.10 bar acceptance limit, it may be decisive. The finding is the same; the impact is not.

Where exact reconstruction is possible, recalculate affected results using the as-found correction or a justified estimate of error over the review period. Where the timing of drift is uncertain, a conservative bound may be more appropriate. State assumptions explicitly. A retrospective correction should not create false confidence if the equipment’s behaviour between calibrations is unknown.

Possible dispositions include no demonstrable impact, result remains acceptable with documented rationale, result requires qualification or amendment, repeat measurement is needed, or customer notification is required. The laboratory should define who has authority to make these decisions and when the customer or other recipient must be informed.

Record the technical decision, not only the corrective action

The investigation record should allow an independent reviewer to understand why the laboratory reached its decision. It should link the out-of-tolerance result to the equipment history, affected work, technical evaluation, any communication, corrective action and effectiveness check.

Corrective action may address a specific cause, such as mishandling or an unsuitable calibration interval. It may also identify that the acceptance limits were poorly matched to intended use, that intermediate checks were ineffective, or that users selected an inappropriate range. Extending or shortening an interval should follow evidence of performance and risk, not the fact that an out-of-tolerance finding is inconvenient.

A well-managed out-of-tolerance event does not presume widespread failure. It establishes what was known, what may have been affected, what evidence supports the decision, and whether the controls around the equipment remain suitable for the work it supports.

References

• ISO/IEC 17025:2017, *General requirements for the competence of testing and calibration laboratories*, clauses 6.4, 7.7 and 7.10.
• ILAC G8:09/2019, *Guidelines on Decision Rules and Statements of Conformity*.
• JCGM 106:2012, *Evaluation of measurement data - The role of measurement uncertainty in conformity assessment*.