The Calibration Schedule Worth Keeping
Why this matters
Most shops that "do calibration" are doing one of two useless things. Either they send everything out on the same annual cycle whether it needs it or not, which spends real money proving that a tape measure is still a tape measure, or they have a drawer of instruments with expired stickers that everyone keeps using because nothing bad has happened yet.
The gap between those and a real schedule is not effort. It is a ranking. Some instruments, when they read wrong, produce a callback. A small number, when they read wrong, produce an injury - because a person acted on the reading and put a hand somewhere. Those two groups do not belong on the same interval, and sorting them is most of the work.
Three words that get used interchangeably and should not be
Get these straight before setting any interval, because a shop that thinks its field checks are calibration has no calibration program at all.
- Calibration is a comparison against a reference standard traceable to a national measurement standard, performed by a lab or with traceable equipment, producing a record with an as-found result (how the instrument read when it arrived) and an as-left result (how it read when it went back). Calibration is a measurement of the instrument, not a repair.
- Adjustment is bringing an out-of-tolerance instrument back into tolerance. A calibration can pass with no adjustment at all, and that outcome is the one that earns you a longer interval.
- Verification, also called a field check or a functional check, is confirming before use that the instrument responds correctly on a known source. It catches gross failure - a dead instrument, a broken lead, a bad probe - and it catches nothing about accuracy. It does not replace calibration and calibration does not replace it. See related: How to Verify a Test Instrument Before You Trust It.
The distinction that matters in the field: verification tells you the instrument is alive. Calibration tells you it is right.
Tier the register by what a wrong reading causes
Sort every measuring instrument into one of three tiers, based on what happens if the reading is wrong and somebody believes it.
| Tier | Test | Examples of the shape | How it is handled |
|---|---|---|---|
| A | A wrong reading can lead to a person contacting energy, entering a space judged safe, or opening something judged empty | Instruments used to establish an absence of voltage, gas and atmospheric detectors, gauges used to declare a system depressurized, torque tools on safety-critical fasteners | Traceable calibration on a hard interval, plus a verification before and after every use |
| B | A wrong reading causes a misdiagnosis, a wrong part, or a wrong pass or fail call, but no injury | Most general measurement instruments and gauges | Traceable calibration on an interval that flexes with as-found history |
| C | The reading is indicative, used for orientation, and no decision rests on it alone | Rough-check and go/no-go devices | Functional check on the inspection cycle, no calibration slot |
Tiering is where the money gets found. In a typical small-shop register, tier A is a handful of instruments, tier B is a modest set, and tier C is most of what people were sending out annually out of habit.
A tier A instrument is never assigned a tier B interval because it is expensive to calibrate. If the cost is a problem, the answer is fewer tier A instruments in circulation, not longer intervals on them.
Set the starting interval from three inputs, in this order
- The manufacturer's stated interval. This is your default. It is not a law of nature, but it is the only number anyone has actually tested the instrument against.
- Whether a standard or a customer contract fixes it. Some work is done under a specification that names the interval. That number wins over the manufacturer's.
- The tier. Tier A gets the shorter of the manufacturer's interval and 12 months, and it does not lengthen past 12 months no matter how good the history gets. Tier B starts at the manufacturer's interval and then flexes. Tier C has no interval.
Do not set the starting interval from your own confidence, from what the last shop did, or from the sticker on the instrument when you bought it used.
The as-found rule: let the instrument's own history set the interval
This is the part almost nobody does, and it is the entire payoff of keeping calibration records rather than just calibration stickers.
Start at the interval from the section above. At every calibration, record the as-found result. If a tier B instrument comes back in tolerance on three consecutive calibrations, lengthen its interval by one step, where a step is 50% of the current interval, capped at twice the original interval, and reset the streak to zero. If it comes back out of tolerance even once, halve the interval immediately and reset the streak to zero. Tier A instruments follow the same streak logic downward only, and never exceed 12 months.
Three parameters in that rule, and each is there for a reason:
- Three consecutive passes, not one. A single pass is noise. Three in a row is a pattern that the instrument is stable at your duty.
- A 50% step, not a jump to the cap. Doubling the interval on the strength of one good streak is how a shop discovers a drifting instrument two years late. One step per pass, then earn the next one.
- Halving on one failure, not on two. The asymmetry is deliberate. Confidence should be slow to build and fast to lose, because the cost of a wrong reading is not symmetrical with the cost of an extra calibration.
Worked example: one instrument over four calibration cycles
A tier B instrument, manufacturer's stated interval 12 months.
- Calibration 1, at 12 months. As-found: in tolerance. Streak: 1. Interval unchanged at 12 months.
- Calibration 2, at 12 months. As-found: in tolerance. Streak: 2. Interval unchanged.
- Calibration 3, at 12 months. As-found: in tolerance. Streak: 3, so the rule fires. One step is 50% of the current 12 months, which is 6 months, so the new interval is 18 months. The cap is twice the original 12, which is 24 months, so 18 is allowed. Streak resets to zero.
The temptation at this point is to jump straight to the 24-month cap, because the instrument has three clean cycles and the cap says 24 is permissible. The rule says one step per pass, and the reason shows up next.
- Calibration 4, at 18 months. As-found: out of tolerance. Interval halves from 18 to 9 months. Streak resets to zero.
Had the shop jumped to the cap at calibration 3, the drift would have been caught at 24 months instead of 18, adding six more months of readings taken on an instrument of unknown accuracy.
The back-trace. Calibration 4 failing as-found does not only change the interval going forward. Every reading taken since calibration 3 was taken on an instrument that is now known to have been out of tolerance at some point in that window. The instrument was used on roughly 130 jobs over those 18 months, but on most of those the reading was informational. The job records show it drove a pass-or-fail call or a component condemnation on 22 of them. Those 22 are the review set, not the 130.
Direction tells you which error you made. If the as-found shows it reading low, you passed things that should have failed, and the review set gets a proactive check. If it reads high, you condemned components that were probably fine, and the review set is a customer-goodwill problem rather than a safety one. Both are worth knowing; only one gets a phone call.
What changes this response. Magnitude matters as much as direction. An as-found that is barely outside tolerance affects only the calls that were close to the line, and the review set shrinks to those. An as-found that is far outside tolerance means every one of the 22 is suspect. Read the actual as-found numbers on the report rather than treating "failed" as a single condition, which is the main reason to read calibration reports at all instead of filing them.
Read the report, not the sticker
The sticker tells you the instrument left the lab in tolerance. The report tells you whether it arrived that way, and that is the only piece of information that can change your schedule.
Keep the reports, one file per instrument ID, with as-found and as-left on every entry. After three or four cycles you can see whether an instrument drifts in one direction consistently, which is a different situation from one that fails randomly. Consistent one-direction drift is often predictable enough to plan around; random failures mean the instrument is unreliable and belongs on the replacement list rather than the calibration list.
Use a lab whose calibrations are traceable to a national measurement standard, and for tier A instruments prefer one accredited to the international laboratory competence standard. Traceability is what makes the as-found number mean anything.
The tier A rules that do not flex
Two of them, and they are the reason tier A exists as a separate category.
Before each use of a test instrument, visually inspect the instrument, its leads, cables, probes and connectors for external defects or damage, which OSHA 29 CFR 1910.334(c)(2) requires, and take it out of service if you find any. A cracked lead or a damaged probe makes the calibration record irrelevant.
When using a test instrument to verify an absence of voltage, confirm the instrument operates on a known live source before the test and again after it, as NFPA 70E requires for absence-of-voltage verification. This is the control that makes tier A workable at all: it catches a failure that occurred between calibrations, which no interval can. A test instrument that reads wrong is a safety device that lies, and the before-and-after check is what stops the lie from being believed.
When a tier A instrument fails an as-found calibration, it comes out of service the moment the report arrives - not at the end of the week, and not held on a truck until a replacement is ordered. Tag it, remove it physically, and pull anything that was in a case with it.
What changes the interval outside the rule
- A drop, an impact, a heat event, or a water exposure. Any of these voids the interval on the spot. The instrument goes for calibration regardless of how much time is left, because the whole basis of an interval is that nothing unusual happened during it.
- A change in the work the instrument does. An instrument moved from bench use into daily field use is now in a different environment than the history was built in, and its streak restarts.
- A new instrument of a type you have not run before. No history means no basis for lengthening. Run at the manufacturer's interval for three cycles before the rule can apply.
- An instrument shared across several techs. Higher handling, more chances of an unreported drop. Treat a shared instrument as one tier more conservative than its consequence tier suggests, or keep the interval fixed and skip the lengthening entirely.
References
- OSHA 29 CFR 1910.334(c)(2), test instruments, leads, cables, probes and connectors shall be visually inspected for external defects and damage before use
- NFPA 70E, verification of the absence of voltage requires confirming the test instrument operates on a known voltage source before and after the test
- ISO/IEC 17025, general requirements for the competence of testing and calibration laboratories, the accreditation to look for when selecting a calibration lab
- Manufacturer documentation for stated calibration intervals and tolerance specifications
- See related: The Tool Calibration and Tracking SOP; How to Verify a Test Instrument Before You Trust It; Calibration: When Your Meter Lies