· 5 min read
Calibration gets filed mentally alongside inspection, and it is a different problem. An inspection asks whether a thing is safe to use. A calibration asks whether the numbers it produced were true — and unlike safety, that question reaches backwards.
That is the whole difficulty. When a torque wrench fails its inspection you stop using it. When a torque wrench fails its calibration you stop using it *and* you have to work out what it has been used on since the last time it passed.
What a certificate actually has to give you
A calibration certificate that cannot be traced is decoration. The things that make it worth keeping:
- The specific instrument, by serial number — not the model. A certificate that says "Torque wrench, 20–100 Nm" belongs to no particular wrench.
- Traceability to national standards, usually via a UKAS-accredited laboratory or an in-house standard that is itself traceable. Without the chain, the reading is an opinion.
- The measured values, not just a pass. "As found" and "as left" are the difference between a record and a receipt.
- The uncertainty of the measurement, which is what tells you whether a borderline result is actually borderline.
- The date, and the interval it was issued against.
"As found" is the field that matters. A certificate showing only "as left" tells you the instrument is right now, and nothing about the six months you were using it.
Setting the interval
Twelve months is the default everywhere, and it is a convention rather than a finding. The honest way to set an interval is to start with the manufacturer’s recommendation and then let the instrument’s own history move it.
| What the last few results showed | Do this |
|---|---|
| In tolerance, comfortably, every time | Consider extending — with a documented reason |
| In tolerance but drifting one way | Hold the interval and watch; drift predicts failure |
| Found out of tolerance once | Shorten it, and look at how it is stored and handled |
| Dropped, overloaded, or in an argument about a reading | Calibrate now, whatever the schedule says |
That last row is the one people skip because it is inconvenient. An instrument that has been dropped is out of calibration until proven otherwise, and the cost of finding out later is every measurement in between.
The recall, which is the part nobody plans
An instrument comes back marked out of tolerance. Now what?
You need to know what it was used on, and that is only answerable if custody was recorded. If the register says the micrometer lives in the metrology cupboard, you have nothing. If it says who had it, when, and on which job, you have a bounded list of work to review — usually much shorter than the panic suggests.
- Quarantine the instrument immediately, in a state the system understands. Not a note on the case.
- List every issue of it back to the last passing certificate. This is the recall window, and it is why custody history matters more here than anywhere else in the store.
- Assess the impact against the actual tolerance, not the worst case. An instrument reading 2% high does not invalidate a job whose tolerance was 10%.
- Record the conclusion, including the jobs you decided were unaffected and why. That reasoning is the thing an auditor asks about, and it evaporates within a fortnight if it is not written down.
- Then decide whether to adjust, repair or retire — which is the easy part.
Where calibration control actually fails
- The due date is tracked but the certificate is not attached, so proving compliance means finding an email from a year ago.
- The instrument comes back from the lab and goes on the shelf before anyone records the new date, so it looks overdue and gets sent again.
- Personal instruments. Somebody’s own micrometer, used on real work, on nobody’s register. This is more common than anyone admits and it is the single hardest one to fix, because the fix is a conversation rather than a system.
- The interval is measured from when it was booked rather than when it was done, which drifts by a fortnight each cycle until something is quietly out of date.
- Nothing warns you until it is overdue, so the instrument is unusable on the morning somebody needs it.
That last one is worth designing around. A calibration that expires without warning takes an instrument out of service at random. Thirty days of notice turns it into a booking.
Nilstock holds each instrument as its own asset with its own history, records calibration dates alongside custody so a recall has a list to work from rather than a guess, and carries the states an instrument actually occupies — including quarantined and away for calibration. The laboratories page covers this shape specifically, and inspection records covers the safety side, which is a different duty with different rules. What belongs on the record at all is in what belongs in an asset register.
Nilstock does this bit for you.
Scan a badge, scan the item, confirm. Fourteen days free, no card.
Read next
The inspection records you have to keep — and where they go missing
Almost nobody fails an inspection audit because the kit was unsafe. They fail because the paperwork proving it was safe is in a van.
What belongs in an asset register, and what quietly ruins one
The register that gets maintained is the one somebody can fill in with cold hands in a stores doorway.