A production flow meter is calibrated and found outside its permitted error at part of its range. Several pump batches were released using it during the previous month. Must every result be discarded, or can the team simply recalibrate and move on? Neither response is adequate.
Stop or restrict the affected measurement function, preserve the as-found results and assess the relevant range, period, products and decision margins. Restoring the instrument addresses future use. Earlier release decisions require a separate review.
What an as-found result establishes
Calibration describes the instrument’s performance at the time and under the conditions tested. It does not usually identify the day an error began or prove that every earlier reading had the same error.
The instrument may have drifted gradually, or its condition may have changed after impact, repair, contamination or another event. Preserve the complete pre-adjustment data rather than accepting only a post-adjustment pass certificate.
If the report simply says “calibrated”, request the original condition and measured errors. Missing evidence makes retrospective assessment harder; it should not be replaced by an invented history.
Link the affected function to actual use
Identify instrument number, function, range, station and usage records. Connect them to the tests and batches actually measured. An error in an unused function may have a different impact from an error at the release point.
The presence of other instruments in the laboratory is not a reason to place every order within the same investigation.
| Review question | Evidence | Purpose |
|---|---|---|
| Which function or range is affected? | As-found calibration data | Defines the technical scope |
| Which products used it? | Instrument-linked test records | Identifies relevant results |
| When might performance have changed? | Intermediate checks, history and events | Supports a justified review period |
| Could acceptance change? | Original readings, limits and uncertainty | Prioritises reassessment |
A recent check is useful only within its coverage. A check at another function or range cannot automatically establish that every measurement was satisfactory. Where the history remains uncertain, retain that uncertainty and choose an appropriate conservative disposition.
Error direction matters
Consider a hypothetical minimum acceptance flow of 10.0 m³/h and a recorded result of 10.2 m³/h. Later calibration finds a positive indication error of approximately 0.4 m³/h under relevant conditions.
If that error can legitimately be applied to the original measurement, the corrected value would be approximately 9.8 m³/h, potentially changing acceptance. An error in the opposite direction has a different effect.
The example is not permission to apply one current calibration point to every result from the past month. Range behaviour, installation, measurement uncertainty and the agreed decision rule still matter.
Near-limit results deserve attention, but apparently comfortable margins also need a documented explanation for any conclusion of no identified impact. An informal verbal exclusion is not a reviewable decision.
Decide what to retest and communicate
The responsible quality team should determine whether to remeasure, review records, hold shipments or take another action. Delivered products need treatment consistent with contractual and applicable requirements.
Do not declare every pump defective before assessing the evidence. Equally, do not declare no impact simply to avoid disruption. Where evidence cannot exclude an effect, state that limitation and define the next verification action.
Retesting requires a suitable controlled measurement system and conditions corresponding to the acceptance requirement. A later pass does not justify deleting the original record. Preserve the original result, impact assessment and new disposition together.
Sample identity, intervening product changes and storage can affect what a repeat demonstrates. A new test should not be described as recreating the past unless its limitations are addressed.
Prevent an untraceable review next time
Record critical instrument identities in actual test data, retain intermediate checks and document abnormal events. Calibration intervals should reflect use, stability and consequence, rather than an unexamined fixed period or automatic halving rule.
For high-impact measurements, suitable independent checks or reference monitoring may help identify changes earlier. Define who opens the investigation, approves its scope and closes the product actions.
Keep “instrument restored to service” separate from “previous measurement impact resolved”. Otherwise, a repaired meter can accidentally close an unfinished product-release problem.
Frequently asked questions
Does out-of-tolerance calibration mean every earlier pump failed?
No. Assess range, actual use, error direction and the possibility of a changed acceptance decision.
Does a successful recalibration close the historical issue?
No. Instrument restoration and retrospective impact assessment are separate activities.
Can daily error be interpolated linearly between calibrations?
Not automatically. Without evidence of the drift mechanism, linear change remains an assumption rather than an established history.
