A certificate showing a result outside the acceptance criterion raises two questions: what happened to the instrument, and which decisions may have depended on its measurements? The finding does not automatically make every earlier reading wrong. Preserve the equipment in the condition in which it arrived, then map its use and assess possible effects in context.
This article describes a way to organize that assessment in general industrial settings. It does not set tolerances, acceptance criteria, or a universal rule for deciding which product, process, or record to review. Each organization must apply the specifications and metrology procedure relevant to the instrument’s actual function.
Identify the criterion the instrument failed
“Out of tolerance” can compress several separate questions: what quantity was calibrated, which points in the range were tested, under what conditions, what error was observed, and which limit was used for comparison. Before looking for consequences, read the full report. Confirm the instrument identifier and configuration, tested range, units, results before any adjustment, and criterion used to classify them.
If the certificate says only “nonconforming” without giving the criterion or relevant test points, request those details from the laboratory or whoever defined the plan.
The International Vocabulary of Metrology (VIM) defines calibration as an operation that relates values provided by standards to instrument indications, including their uncertainties, and uses that relationship to obtain a measurement result. A certificate is therefore more than a pass-or-fail label: its data can show how the indication varied across the calibrated points. BIPM/JCGM, VIM3 2.39.
To compare those results with a specification, ILAC’s document on decision rules says the specification and the criterion used to issue a statement of conformity need to be defined. Its scope is laboratory work that reports conformity, not every production decision. ILAC G8:09/2019, §§2.3 and 2.6.
The report also involves three distinct ideas: calibration, verification, and adjustment. Calibration provides information about the relationship between an indication and a reference; verification checks performance against specified requirements; adjustment changes the measurement system. The VIM cautions that calibration and adjustment are not equivalent. BIPM/JCGM, VIM3 2.39, note 2.
Metrological traceability also applies to a measurement result, rather than serving as a general promise about an instrument: it requires a documented calibration chain that contributes to the result’s measurement uncertainty. BIPM/JCGM, VIM3 2.41. The acceptance criterion should therefore be found in the specification, procedure, or requirement that actually governs the use. People on the plant floor often say “tolerance,” but the report should identify the requirement being applied.
ILAC/OIML cautions against using that term to mean the instrument’s maximum permissible error; write down the specific specification before comparing results. ILAC/OIML G24:2022, §3.13.
A deviation at one point can decide whether an instrument is suitable for one application and have little bearing on another. The relevant factors include the portion of the range used, required resolution, acceptable uncertainty, and the role of the measurement. An error in a thermometer used for a monitoring alarm may matter differently from the same error in a measurement that determines a process mixture.
You cannot tell which case matters more without understanding the risk and the decision made from each reading.
Record the condition as found, often called “as found,” and keep it separate from any result obtained after adjustment or repair. If the supplier reports “as left” values, confirm whether they describe the instrument after an intervention. Those later values help assess the repaired equipment; they do not replace evidence of how it arrived at the laboratory.
Preserve the as-found condition before intervention
When a result fails the acceptance criterion, avoid changing the instrument until you have saved the available information and agreed how to carry out the assessment. Keep the complete certificate, data for each point, reported error, stated uncertainty, environmental conditions if provided, reference-standard identification, and calibration date. If settings, firmware, corrections, or accessories can affect the result, record them before intervening.
Adjustment changes the instrument and may change its behavior. The VIM defines adjustment as a set of operations intended to make a measurement system provide prescribed indications, and notes that recalibration normally follows. The before and after conditions answer different questions: the first helps explain the earlier period; the second shows how the instrument performs after the work. BIPM/JCGM, VIM3 3.11.
Ask the laboratory for whatever information is needed to interpret the results before relying on a summary conclusion. You may need to know whether the out-of-limit result appeared at one point only, whether errors showed a sustained increase or decrease, whether the laboratory repeated measurements, or whether it applied a correction. Do not invent those details or treat an “as left” certificate as indirect proof of the original error.
If the report lacks them, note the limitation and ask for clarification. The JCGM’s GUM describes uncertainty evaluation in relation to a measurement model and its inputs, so a single summary figure does not describe every condition of use. JCGM 100:2008, sections 4 and 5.
Preserve the link between the instrument and historical measurements as well. Export and retain original records; if you recalculate any value, do so in a copy. Keep timestamps, units, identifiers, configuration, and source system. If you later apply an estimated correction for analysis, retain the original data and the transformation separately, along with its rationale and version.
This lets someone reconstruct what was observed and what was added during the later assessment.
NBS Handbook 145 includes a section on evaluating product potentially affected by measuring equipment found out of tolerance and says the evaluation results and actions should be recorded. It is a historical reference about metrological quality assurance in its own context, not a current general standard for every manufacturer. NIST, NBS Handbook 145, §5.6.
ILAC/OIML’s G24 guide, in its 2022 edition, helps assess and review calibration intervals; it does not set a predefined retrospective window for impact analysis. ILAC/OIML G24:2022, sections 5 and 6. The practical documentation point is to retain enough evidence for another person to review the assessment. The scope of any action depends on local procedures and obligations.
Bound the measurements and period that may be affected
The practical question is not “how long did the deviation last?” when the evidence cannot establish that. Ask instead: “Which results were obtained with this instrument, under what conditions, and which decisions may have relied on them?” Start with the equipment identifier and use records: work orders, line, station, product or process, installation and removal dates, calibration dates or periods, and configuration changes.
If the instrument measures a process variable, include the complete loop where relevant: sensor, transmitter, conversion, input channel, display, and any calculation that turns a signal into a reported quantity.
Calibration history helps bound dates; by itself, it does not show when the error began or how it changed between checks. Comparing successive results may show apparent stability or drift. Changes in use, impact, repair, contamination, overload, or environmental conditions can alter that pattern. ILAC-G24 describes methods for setting and reviewing recalibration intervals using historical data and equipment evaluation.
Its purpose is to guide calibration programs, not to supply a universal rule for reconstructing every measurement made before an out-of-tolerance result. ILAC/OIML, G24:2022, scope and methods.
Set an initial window using facts that can be checked: the last known result within the criterion, the date of the out-of-limit result, maintenance records, intermediate checks, and relevant changes. Then identify which readings were used to accept a batch, adjust a machine, regulate flow, calculate consumption, trigger an alarm, or close an investigation. Archived readings do not all carry the same consequence.
Some show a trend; others trigger quality, safety, or cost decisions. Sorting records by use helps focus the analysis.
Next, compare the “as found” error with the actual operating range. If the certificate shows a deviation at a point that was never used, that may reduce the relevance of that point, but it does not prove the instrument was correct at the points that were used. If values are available, assess the calibration relationship, direction of error, and relevant uncertainty.
Do not extrapolate beyond the points and conditions supported by evidence without explaining why.
Metrological traceability is a property of a measurement result. The BIPM/JCGM defines it as a relation to a reference through a documented, unbroken chain of calibrations, each contributing to measurement uncertainty. Having that chain does not guarantee that uncertainty is suitable for the use or that no errors occurred. BIPM/JCGM, VIM3 2.41. So, alongside the reference-standard chain, preserve operational traceability: which instrument produced each result and where it was used.
One concerns metrological references; the other concerns an item’s history. Confusing them can leave unanswered which plant decisions depended on the measurement.
Assess impact with data, uncertainty, and clear limits
A retrospective assessment is not a matter of subtracting the observed error from every old reading. To correct a measurement, you would need evidence that the relationship found during calibration also represents the conditions when that reading was taken. The instrument may have changed during the period; temperature, range, load, position, or reading method may have differed.
If that information is unavailable, present scenarios or bounds as such, not as certain corrected values.
A calibration result includes uncertainty. The VIM links calibration to uncertainties associated with the standard values and indications; the JCGM’s GUM gives principles for evaluating and expressing measurement uncertainty from a model and its inputs. Neither source provides one uncertainty value for every instrument or application. BIPM/JCGM, VIM3 2.39; JCGM 100:2008, GUM.
For each group of results, compare the available information with the decision that depended on it. If a reading was compared with a limit, identify that limit, the recorded reading, available uncertainty, and the instrument’s possible influence. Do not turn this into an automatic rule such as “if the error fits inside the band, everything is acceptable”; the decision depends on the criterion defined for the use.
ILAC-G8 addresses decision rules for statements of conformity within accredited laboratory assessment. It helps explain why the rule and decision risk should be explicit, but it is not a general standard for every factory decision. ILAC G8:09/2019, purpose and scope.
If the plausible interval could change the conclusion, leave the result undetermined until a competent person assesses the options. Options might include expanding the analysis, repeating a measurement with a suitable method, reviewing an earlier decision, consulting the customer, or documenting why the effect does not change the outcome. These are examples, not universal recommendations. The right choice depends on the process, intended use, risk, and internal rules.
NBS Handbook 145 gives a historical example of assessing effects on product examined with out-of-tolerance equipment and recording the results. But the text belongs to a NIST handbook for a specific metrological assurance program; it does not establish that every deviation requires product withdrawal or that a particular period must be reviewed. NIST, NBS Handbook 145, §5.6.
Keep evidence, hypotheses, and decisions distinct: “the certificate found an error at these points” is a fact; “last week’s reading was wrong” requires an assessment of use, history, and measurement model.
Keep calibration, adjustment, and repair separate
Calibration characterizes the relationship between an instrument’s indications and reference values; adjustment changes its response. Note 2 to VIM3 2.39 says that calibration “should not be confused with adjustment.” BIPM/JCGM, VIM3 2.39, note 2. Note 3 to VIM3 3.11 says that a measuring system normally needs recalibration after adjustment. BIPM/JCGM, VIM3 3.11, note 3.
First preserve the as-found condition and assess possible historical effects. Then decide whether to adjust, repair, remove, or keep the instrument out of service under the plant procedure. After an intervention, check the result against the criteria defined for its intended use.
An acceptable “as left” reading describes the condition after the intervention and the tests performed; it does not erase the “as found” results or prove the instrument was correct throughout the earlier period.
If the cause is unknown, do not assume adjustment resolves it. There may be a problem with mounting, wiring, configured range, environment, wear, overload, method, or reference. Changing the zero may hide an offset without fixing nonlinearity or hysteresis, where the response depends on whether the same measured value is approached from increasing or decreasing values.
If the intervention alters the equipment before you collect data needed for diagnosis, you may lose information about the failure. ILAC-G24 includes methods for reviewing recalibration intervals using calibration results and intermediate checks; a program review should use data and risk, rather than follow automatically from one result. ILAC/OIML, G24:2022, section 6.
Calibration also does not guarantee that the instrument suits every purpose. Compare its result with the requirements and uncertainty allowed for the use. BIPM/JCGM, VIM3 2.41. The VIM definition of metrological traceability cautions that a documented chain alone guarantees neither suitable uncertainty nor the absence of errors. BIPM/JCGM, VIM3 2.41.
An instrument can have a certificate from a competent laboratory and still be unsuitable for a task that requires finer resolution or lower uncertainty.
In the records, separate the initial finding, decision to intervene, work performed, parts replaced or parameters changed, and the new calibration or verification. G24 recommends selecting and reviewing intervals based on the equipment and calibration data, including methods based on time in use and checks during service. ILAC/OIML G24:2022, sections 5 and 6.
This timeline distinguishes a repair from a calibration and prevents a later certificate from replacing the history needed to assess the earlier period.
Close the assessment with a local decision and reviewable evidence
An impact report should let another person reconstruct how the conclusion was reached. Include the instrument, “as found” certificate, acceptance criterion, error and test points, known dates and uses, records reviewed, available uncertainty, limitations, and the decision made by the responsible person. If you corrected data for analysis, retain the original, the formula used, and who approved the method.
The conclusion may be that there is no evidence of a relevant impact, that some measurements need further assessment, or that the effect cannot be determined from the available records. The third outcome is not a writing failure; it identifies a real limit to what is known. State what is missing, why it matters, and what action could reduce the uncertainty.
Do not turn “we found no evidence” into “there was no impact” when records do not cover the period or use.
Also review the system that allowed the instrument to remain in use. ILAC-G24 proposes several methods for reviewing recalibration intervals, including the history, control charts, time in use, and intermediate checks. The guide is aimed at laboratories and other conformity assessment bodies, and also mentions manufacturers that use measuring equipment. It does not prescribe the same frequency for every company. ILAC/OIML G24:2022, scope and review methods.
Historical records may support trying a different interval, but operating conditions, criticality, observed stability, and ability to detect change still matter.
Metrological traceability links results to references through documented calibrations that contribute uncertainty at each step. BIPM/JCGM, VIM3 2.41. The GUM explains that combined uncertainty depends on the measurement model and the contributions from its input quantities. JCGM 100:2008, §5. An internal investigation also needs to link each result to the instrument identifier and its use.
A maintenance system can help find overdue calibrations, alerts, and work orders; it cannot replace reading the “as found” data or assessing which decisions relied on them.
The response may lead to different changes: reviewing the method, adding an intermediate check, training users, or changing the interval. Choose an action that addresses the observed weakness and record the reason. Keep the rationale with the investigation closure, not only in a status field. The next person should be able to see what was checked and what remained outside the assessment.
Gather the certificate, applicable criterion, use history, and earlier calibration results. With those records, metrology and process staff can bound the measurements that may be affected and document the local response. If the records cannot reconstruct the earlier condition, state that uncertainty plainly.
Frequently asked questions
What does it mean for an instrument to be out of tolerance?
It means one or more results failed the specified acceptance criterion for the instrument, range, and conditions evaluated. The phrase does not establish which historical reading was wrong; to assess that, relate the “as found” certificate to the actual period and use. The decision rule used to report conformity should be identifiable. ILAC G8:09/2019, §§2.3 and 2.6.
Should I adjust the instrument right away?
First preserve the “as found” result, the full report, and the configuration. Adjustment may alter evidence about the condition on arrival; after adjustment, the instrument is normally recalibrated to establish its later behavior. Keep the before and after data as separate results. BIPM/JCGM, VIM3 3.11.
Do I need to review every measurement made before the finding?
There is no universal time window for reviewing earlier measurements. Bound which readings the instrument may have produced using use history, the last known result, intermediate checks, maintenance, configuration, and working conditions. Prioritize measurements that supported consequential decisions, and document why groups of records were included or excluded.
Does traceable calibration prove that a measurement is suitable for any use?
No. Traceability relates a result to a reference through a documented chain of calibrations; it does not guarantee that uncertainty is suitable for a particular application or that no error occurred. Also check the uncertainty and requirements for the intended use. BIPM/JCGM, VIM3 2.41.
Are calibration and adjustment the same thing?
No. Calibration establishes a relationship between standard values and indications; adjustment changes the system to obtain prescribed indications. The VIM defines them separately and says a measuring system normally needs recalibration after adjustment. An “as left” certificate does not replace the “as found” evidence. BIPM/JCGM, VIM3 2.39; VIM3 3.11.
What should I do if I cannot reconstruct the impact?
Record what information is missing, which periods or uses remain unevaluated, and which decisions may depend on those measurements. Do not turn missing evidence into a conclusion that there was no impact. The competent metrology and process leads should decide what local action is appropriate and document its limits.