In the mining laboratories we work with, balances rarely lose their adjustment for exotic reasons. Someone moves them. Or someone puts a load on them that does not belong there, and on a microbalance that is all it takes.
Neither event leaves a mark on the calibration certificate. The certificate still says the balance was within tolerance on the day the provider tested it, and that remains true. What it cannot say is whether the balance is still within tolerance today, after the move, after the load, after months of dust and use.
That gap is what ISO/IEC 17025 clause 6.4.10 is about: when intermediate checks are needed to maintain confidence in equipment performance, they have to be carried out according to a procedure. In a mining laboratory, where a balance sits behind flux additions, gravimetric results and every weighed reference material, that check is the only evidence that the numbers between calibrations are worth trusting.
Calibration and intermediate checks answer different questions
Calibration tells you how the instrument performed against traceable references, under the provider's conditions, on one date. An intermediate check asks something narrower and more practical:
Since the last calibration, do we still have evidence that this balance is fit for the work it is doing?
One does not replace the other. A passing check does not extend the calibration interval, and a current certificate does not excuse the checks your procedure requires. They are two records of two different moments.
The two events that matter most
The balance was moved. Relocating a balance, even across the same bench, can change its level, its exposure to vibration and drafts, and its behaviour near zero. The certificate was issued for the balance where it stood. After a move, the honest status of the balance is "unverified" until someone checks it.
The balance took a load it should not have. A microbalance is built for milligrams. A container or sample far above its range, placed on the pan for a moment, can shift the mechanism enough that every later reading carries an offset nobody sees. Nothing looks broken. The display still settles. The numbers are simply wrong by a small, consistent amount.
Both events have the same weakness: they happen between calibrations, and nobody writes them down. That is why the checks should not only run on a calendar. A move, a suspected overload, maintenance or an impact are all reasons to check before the next sample, not at the end of the month.
A check must record more than "OK"
A tick in a logbook says someone looked. It does not say what they found. A check that can defend the results that follow it records:
- the balance, by its unique equipment ID;
- the reference weight used, and its nominal value;
- the reading the balance gave;
- the acceptance criterion, set before the check;
- the verdict;
- the date and the person who did it; and
- what was done if the check failed.
The criterion needs a technical basis the laboratory can explain: the method's needs, the balance's class and working range, its calibration history. A default value in a spreadsheet is not a criterion. And a failed check stays in the record. A history with only passing values cannot show when control was lost.
A failed check draws a line in the timeline
When a check fails, the last passing check becomes the most important date in the file. From there, the laboratory has to answer:
- When did the balance last demonstrate acceptable performance?
- What was weighed on it after that point: which samples, worksheets and reported results?
- Do CRMs, blanks or duplicates from those runs tell us anything about the affected work?
- Does the balance need adjustment, service or recalibration before it is used again?
- Is re-weighing, reanalysis or notifying the client required?
- Who authorises putting the balance back into service?
The answers depend on the balance, the method, the size of the deviation and the laboratory's own procedure. What they all need is a record that makes the timeline visible. If the event qualifies as nonconforming work, the failed check is where the investigation starts. The rest of that path is covered in A CAPA Is Not Closed When the Action Is Done.
Where OnQMS fits, and where it does not
OnQMS keeps intermediate checks beside the operational LIMS record, with the instrument taken from the LIMS master list. Each check records the check type, reference standard, nominal and measured values, tolerance and unit, the person responsible, and the expected frequency.
With numeric values, OnQMS calculates the verdict itself: the check passes only if the difference between measured and nominal is within the tolerance, so nobody has to type "OK". Each instrument's latest check sets its next due date, and overdue checks appear among the laboratory's alerts. A failed check with no nonconformity report opened against it stays on the alert list until someone opens one. The Instrument Lookup view shows the balance's calibrations, certificates, intermediate checks and responsible people in one place.
What OnQMS does not do: calibrate the balance, write the check procedure, choose the tolerance, or decide whether past results are still valid. Those remain decisions for qualified laboratory staff. OnQMS makes sure the evidence for those decisions exists, and shows where it is missing.
The practical test: pick one balance
Take the balance your most critical method depends on, and look at the time between its last two calibrations.
Was it moved in that period? Did anyone record a check afterwards? Can you find every scheduled check, with its reference weight, reading, criterion and verdict? If one failed, can you reconstruct what was weighed after the last good check, and who decided what to do about it?
If answering takes a certificate folder, two spreadsheets and someone's memory, the balance may well be calibrated. The evidence between calibrations is what is missing.
Primary reference
International Organization for Standardization. ISO/IEC 17025:2017 — General requirements for the competence of testing and calibration laboratories, clause 6.4.10 on intermediate checks.