What a Calibration Interval Is For

Why this matters

Ask a shop why an instrument is on a twelve-month interval and the answer is almost always that twelve months is what the manufacturer said, or what the last shop did. Ask what shortening it to six would buy and the answers get vague, usually something about the instrument being more reliable.

It would not be. Nothing happens to an instrument on the anniversary of its calibration, and shortening an interval does not make the instrument more likely to be right on any given day. What a shorter interval buys is a smaller pile of past work to re-examine when the instrument turns out to have been wrong. That is the entire product. Once you see the interval as a cap on contamination rather than as maintenance, the right length stops being a calendar question and becomes an arithmetic one, and it frequently turns out that the honest answer is not a shorter interval at all.

The companion card on calibration scheduling owns how to tier instruments by consequence, where a starting interval comes from, and how as-found history flexes it up or down. This one is about the quantity that decides whether any of those intervals is defensible for your shop: how much work a single undetected failure would put in question.

What an interval does not do

An interval does not hold an instrument in tolerance. It does not slow drift, and it has no effect whatsoever on whether the instrument gets dropped in week three.

What it does is set the width of the window of doubt. When a calibration comes back out of tolerance as-found, everything measured since the previous calibration was measured by an instrument now known to have been wrong at some point in that window. The interval is the width of that window. A twelve-month interval means twelve months of readings become questionable in one morning.

That reframing matters because it tells you what a shorter interval is competing against. It is not competing against a better instrument, it is competing against any other control that narrows the window, and a check standard read monthly narrows it far more cheaply than a calibration booked twice as often. The sibling card on drift versus breakage owns how that series is read; what belongs here is that both controls do the same job and only one of them is expensive.

The gate

An interval is defensible when the number of consequential decisions it can contaminate is a number your shop could actually work through. Bound that window by the shorter of the calibration interval and the check-standard cadence.

Three parameters, and each needs stating before anyone can use the rule.

• The unit of analysis is a consequential decision, not a reading. A reading nobody acted on contaminates nothing. The sibling scheduling card establishes that only a fraction of any instrument's readings drive a pass-or-fail call, a condemnation or an acceptance, and that fraction is the number you want. Count it from job records, not from memory.
• The cap, R, is the size of review your shop could genuinely complete inside a week without stopping other work. A common starting point is 25 consequential decisions; tune it to your own capacity, because a shop of two and a shop of twelve do not have the same answer.
• The output is a maximum window, equal to R divided by the decision rate, expressed in whatever period the rate is in, then rounded down to a practical schedule. Weekly, monthly, quarterly, annually. A calculated 1.4 months becomes monthly, not six weeks, because a cadence nobody can remember is a cadence nobody runs.

Run that single rule against two instruments on the same shelf, on the same twelve-month interval, and watch it resolve in opposite directions.

Outcome one: the instrument that decides four times a week

A handheld used on routine service calls, driving roughly 4 consequential calls a week. Over a twelve-month interval that is about 208 consequential decisions.

Against a cap of 25, the twelve-month interval is off by a factor of about eight. If that instrument comes back out of tolerance as-found, the shop is looking at 208 decisions with no way to prioritize them beyond direction and magnitude, and in practice a review that large does not happen. It gets discussed, deferred, and quietly dropped, which means the real response to an out-of-tolerance result is nothing.

The tempting fix is a shorter calibration interval. Run the arithmetic on it: 25 divided by about 17 decisions a month gives roughly 1.4 months, so holding the cap through calibration alone would mean calibrating monthly. No small shop is doing that, and it would be the wrong purchase anyway, because calibration is not what the shop needs here. It already knows what the instrument's tolerance is. What it lacks is a way to place a failure in time.

So the answer is to leave the calibration interval where the manufacturer put it and add a monthly check-standard reading. The window of doubt becomes the shorter of the two, which is one month, and one month at this decision rate is about 17 decisions, inside the cap. The instrument is no more accurate than it was. The consequence of it being wrong went down by a factor of about twelve, at the cost of a few minutes a month against a bench artifact.

Outcome two: the instrument that decides twice a month

A second instrument, same class, same twelve-month interval, used on specialised work that comes up rarely: about 2 consequential decisions a month, so about 24 over the interval.

Against the same cap of 25, the twelve-month interval is already correct. Twenty-four decisions is a review a shop can actually perform, sorted by margin, in an afternoon. Adding a monthly check standard here would produce almost nothing, because the window it would close is already small enough to work through.

Note that the two instruments differ by nearly nine times in exposure while sitting on an identical interval, and that nothing about the instruments themselves distinguishes them. The property that set the answer was the work, not the hardware, which is why an interval copied from another shop's register is a number with no argument behind it.

One condition flips outcome two straight into outcome one, and it is not a property of the instrument either. The specialised work becomes a new service line, the decision rate rises to several a week, and the interval that was defensible last quarter now contaminates eight times the reviewable set. Nothing about the instrument changed and nothing about the calibration changed. This is the reason a decision rate belongs in the instrument register as its own field, reviewed when the work mix changes rather than when the sticker expires.

Where this rule stops

The whole argument above assumes a wrong reading produces a reviewable mistake: a part condemned that was fine, an acceptance issued that should have been a rejection, a trend point that was noise. Those can be found afterwards and put right.

It does not apply to an instrument used to establish that something is de-energized, and no review set exists for that class of error. A person acted on the reading and put a hand somewhere. The control there is not an interval and not a monthly artifact: it is proving the instrument on a known live source immediately before the test and again immediately after it, the live-dead-live sequence at 120.5 of NFPA 70E-2021, in whichever edition your employer's electrical safety program has adopted, inside the electrical lockout duty at 29 CFR 1910.333(b)(2) in general industry or 29 CFR 1926.417 in construction. That per-use proving is what catches a failure that arrived this morning, which no interval of any length can. 29 CFR 1910.334(c)(2) requires the instrument, leads, cables, probes and connectors to be visually inspected for external defects before each use, and the instrument's measurement category and voltage rating under IEC 61010-1, binding through its listing mark, must suit the circuit. Where a reading genuinely has to be taken energized, 29 CFR 1910.333(a)(1) permits it only where de-energizing is infeasible, with the protective equipment 29 CFR 1910.335(a) requires for the exposure.

The same boundary applies to any instrument whose reading is the basis for entering a space, opening something believed empty, or declaring an atmosphere acceptable. Those instruments live under a per-use verification regime, and the exposure arithmetic in this card is not offered as an alternative to it.

What this changes in the register

Three fields per instrument rather than one, and the third is the one nobody keeps.

• Calibration interval, from the manufacturer, a contract, or the tiering the sibling scheduling card owns.
• Check cadence, set from the exposure arithmetic above, and frequently much shorter than the calibration interval.
• Decision rate, counted from job records, dated, and revisited whenever the work mix moves.

The reason the third field earns its place is that it is the only one that changes without anyone touching the instrument. An interval and a cadence both go stale silently when a shop picks up a new kind of work, and the register will look perfectly maintained the whole time.

How to verify you got this right

Pick your most-used instrument and count, from actual job records over the last three months, how many times a reading from it drove a call somebody acted on. Divide your review capacity by that rate. If the answer is shorter than the current window of doubt, you have found a real exposure, and the first thing to price is a check cadence rather than a calibration.

Then run the uncomfortable test. Imagine that instrument's next calibration comes back out of tolerance as-found, in the direction that means you condemned good components. Write down, in one sentence, what your shop would actually do that morning. If the honest answer is "nothing, because it would be too many jobs to look at," the interval is not protecting anything, and the number on the sticker is decoration.

References

• Manufacturer stated calibration intervals and the conditions that void them
• 29 CFR 1910.333(a)(1) energized-work gate; 29 CFR 1910.333(b)(2) and 29 CFR 1926.417 electrical lockout; 29 CFR 1910.334(c)(2) pre-use inspection of instruments, leads, cables, probes and connectors; 29 CFR 1910.335(a) protective equipment
• NFPA 70E-2021, 120.5, absence-of-voltage verification, adopted through an employer electrical safety program or contract; IEC 61010-1 measurement categories, binding through the instrument's listing
• See related: The Calibration Schedule Worth Keeping, which owns tiering and the as-found interval rule; When an Instrument Drifts and When It Just Breaks; What Traceability Means and Why It Matters to You