How to set calibration intervals

An interval is a decision about how long you are prepared to trust an instrument between checks. Guidance from ILAC and OIML describes methods, from fixed intervals adjusted on results to statistical control charts, but all of them rest on the same evidence: how the instrument behaved last time.

Inputs

• Manufacturer's stability specification.
• Required tolerance versus the instrument's typical drift.
• Severity of use, environment and handling.
• Consequence of an out-of-tolerance result being discovered late.
• As-found data from previous calibrations: the strongest input.

Adjusting

A run of in-tolerance as-found results with small drift supports lengthening the interval; an out-of-tolerance result shortens it and triggers an impact assessment. Document the rule you use so that intervals are defensible rather than habitual.

A rule that can be written down

The weakness of most interval policies is that they cannot be stated. Everything is annual, with exceptions nobody can explain. A rule based on as-found history is both defensible and simple enough to put in an SOP: compare the as-found error against the tolerance, and move the interval accordingly.

The percentages are the part to argue about; the structure is not. What makes any version of this defensible is that it is written down before the results arrive, applied consistently, and evidenced from certificates rather than from convenience.

The inputs, weighted

Not all inputs deserve equal weight. As-found history beats everything else, because it is the only evidence about this instrument rather than about its model. The manufacturer's stability specification is a starting point for an instrument with no history and is usually conservative. Severity of use matters more than people allow: a reference kept on a bench in a temperature-controlled laboratory and one carried between sites in a van are not the same instrument after two years.

The consequence of a late discovery is the input that should set the floor. If finding an out-of-tolerance reference means reassessing twelve months of certificates for external customers, that is an argument for a shorter interval or for interim checks, independent of how well the instrument has behaved.

Interim checks change the arithmetic

• A monthly ice-point check, trended, means a drift is located to within a month instead of within the whole interval. That turns a twelve-month look-back into a one-month one.
• Interim checks do not extend an interval on their own — they are checks, not calibrations, and create no traceability link — but they are strong evidence in the case for extending it.
• A control chart of check readings is the most informative single record a temperature laboratory keeps, and it costs a few minutes a month.
• For thermocouples, interim checks matter more than the calendar: their change depends on thermal history, so a couple used hard for a fortnight may need checking before one used lightly for a year.

Whatever rule you adopt, document it, and record the decision each time an interval changes. An interval with no rationale is itself a finding, and identical intervals across instruments of very different risk is the version assessors notice.