First, is the difference real?
Before looking for causes, it is advisable to check whether the difference is significant. The difference between the two reported errors is compared with the combined uncertainty of both calibrations, as with a normalized error. If the difference falls within it, there is no contradiction: they are two compatible estimates of the same thing.
Causes within the instrument
- Drift over time: components age and sensors shift.
- Intermediate adjustment: if someone adjusted the instrument between the two, the errors are not comparable.
- Impacts, overloads or transport that modify the response.
- Hysteresis or poor repeatability, which make the reading depend on how the point was approached.
Causes in the calibration
- Different points and ranges: the error varies along the range and the same point is not being compared.
- Different method: direction of approach, waiting time, number of cycles.
- Environmental conditions: temperature, humidity, pressure, mounting position, prior warm-up.
- Different standards with different uncertainties and traceability.
- Resolution and reading criteria of each laboratory.
- Rounding rules or the way the error is reported (error versus correction, which have opposite signs).
How to investigate
- Align the tables: same points, same units, same sign convention.
- Review the uncertainties and conditions of each report.
- Confirm whether there was an adjustment, repair or accident between the two.
- If the difference is significant, repeat with a third standard or compare against an in-house control artifact.
How to interpret the result
A small, stable difference between calibrations is normal and may be the expected drift. A large, sudden difference indicates an event. A variation that changes direction with no pattern suggests a lack of repeatability in the instrument or the process. Trends across several calibrations provide more information than two isolated points.