Setting calibration intervals with data, not habit
Twelve months for everything is easy to administer but rarely optimal. How to use as-found history to set intervals you can defend.
Most calibration programmes start with a single rule: every instrument, every twelve months. It is simple, auditable and easy to explain. It is also almost certainly wrong for a large part of your inventory, calibrating stable instruments more often than necessary and unstable ones not often enough.
After a few calibration cycles, you have the data to do better.
What the guidance says
ISO/IEC 17025 and ISO 9001 do not prescribe intervals. They require that intervals are established and reviewed. The most widely used guidance on how to do this is ILAC-G24 / OIML D 10, Guidelines for the determination of calibration intervals of measuring instruments, published jointly by ILAC and OIML.
It describes several methods, ranging from simple to statistically sophisticated. Three are particularly practical.
Method 1: Automatic adjustment (“staircase”)
Each time an instrument is calibrated, its next interval is adjusted based on the result:
- if it was in tolerance, the interval is extended, for example by 25 %;
- if it was out of tolerance, the interval is reduced, for example by 50 %.
This is simple to implement and responds to each instrument’s behaviour. Its weakness is that a single result drives each change, so intervals can swing.
Method 2: Control charts
Plot the as-found value of a key parameter over successive calibrations. If the instrument drifts steadily, you can project when it will reach its tolerance limit and set the interval accordingly, leaving a margin. This works well for instruments with a dominant, predictable drift, such as voltage references and frequency standards.
Method 3: Reliability across a population
For large populations of similar instruments, such as fifty identical handheld multimeters, you can look at the percentage found in tolerance at calibration. If your target reliability is, say, 95 % in tolerance at the end of the interval, and you are achieving 99 %, the interval can reasonably be extended. If you are achieving 85 %, it should be shortened.
Inputs beyond the data
Data should drive the decision, but it should not be the only input. Also consider:
- the manufacturer’s recommendation and the specification period you rely on,
- the consequences of an instrument being out of tolerance,
- the usage: hours of use, environment, transport,
- intermediate checks that might detect a problem between calibrations.
Rules of thumb
- Do not extend an interval on the basis of a single good result.
- Never extend an interval for an instrument that has had a recent failure.
- Make changes in moderate steps.
- Document the method, the data and the decision for every change.
Where we help
Because we record as-found data on every certificate, after two or three calibrations we can provide a summary for each of your instruments showing how close each parameter came to its limit. That gives you an evidence-based starting point for interval decisions, and the records to show your auditor.
