Measurement Uncertainty, Accuracy and Resolution

The three terms

Accuracy is a qualitative description of closeness between a measured value and the quantity value being measured; it is not a complete numerical specification by itself. Resolution is the smallest change that produces a perceptible change in indication. Measurement uncertainty is a non-negative parameter that describes the dispersion of values that could reasonably be attributed to the measurand. A device can display many decimal places and still be biased, unstable or too uncertain for the decision.

Error, correction and uncertainty

Measurement error is the measured value minus a reference quantity value. Calibration may estimate that error and supply a correction, but the corrected result still has uncertainty. Uncertainty is not the same as a mistake and does not mean that the result is useless. It expresses the remaining doubt after known corrections have been applied. Gross mistakes such as using the wrong unit, probe location or sample are controlled by procedure and investigation; they should not be hidden inside an uncertainty allowance.

Sources in cured-meat measurement

Contributors can include reference-standard uncertainty, calibration correction, instrument repeatability, drift, display resolution, temperature, operator technique, probe placement, sample preparation and differences between product units. In pH and water activity, sample heterogeneity may be larger than instrument repeatability. In chamber monitoring, location and load pattern may dominate. In weighing, drafts, vibration, off-centre loading and the mass of a small dose matter. The measurement system includes the instrument, method and sample, not only the display.

Repeatability and reproducibility

Repeatability describes variation when the same method, operator, instrument and conditions are used over a short period. Reproducibility or intermediate precision examines broader changes such as different operators, days, devices or locations. Repeated readings on one homogenised sample can demonstrate instrument repeatability but do not show variability between sausage centres or chamber positions. A useful study separates repeated measurement of the same sample from independent sampling of the process.

Simple evaluation

For routine production, start with the calibration uncertainty or manufacturer performance, observed repeatability, resolution and the largest method or sampling effects known to matter. Express all contributors in compatible units, avoid counting the same effect twice and combine them using an appropriate documented method. A formal accredited laboratory report may require a full uncertainty budget and coverage factor; a small producer can use a simpler conservative evaluation when it is sufficient to show that the measurement is clearly fit for the process decision.

Fitness for purpose

Compare measurement capability with the width of the allowed process range and the consequence of a wrong decision. If a critical limit is narrow, the uncertainty should be small enough that results can be interpreted without routine ambiguity. Selecting a finer display does not solve poor sampling or drift. A measurement system is fit for purpose only when its range, resolution, uncertainty, method and sampling support the intended decision under real operating conditions.

Results near a limit

When a result is close to a specification or safety limit, uncertainty creates a risk of false acceptance or false rejection. The decision rule should be defined before results are seen. Options include a guard band that requires the measured result plus relevant uncertainty to remain on the acceptable side, repeat sampling under a defined rule, or holding product for additional evidence. The rule must match applicable law and the validated process. Rounding a borderline result toward compliance is not a decision rule.

Reporting

State the measured value, unit, method, sample or location, instrument and any correction. When uncertainty is material, report the expanded uncertainty and coverage information or provide the documented decision rule used. Do not report more digits than the measurement can support. A water-activity result of 0.9000 or a temperature of 12.347 °C can imply false precision when the system is only capable of resolving the decision to a much wider interval.

Reducing uncertainty

Improve the largest contributors first. Better sample selection may matter more than buying a higher-resolution meter; better probe placement may matter more than adding decimal places; controlling vibration may improve a balance more than repeated adjustment. Use mapping, representative sampling, operator training, suitable reference standards and trend data to identify the dominant cause. The goal is not the smallest possible uncertainty at any cost, but a measurement system that makes the required product decision reliably.