A laboratory instrument usually measures a small portion of material. The question people want answered is often much larger: what is the composition of an entire batch? Connecting those two scales requires a sampling plan as well as a sound analytical method.
Define what the result is meant to describe
There is a difference between measuring the sample delivered to a laboratory and estimating a property of the batch from which it came. The first question concerns the submitted material. The second adds the challenge of how that material was selected.
Eurachem's sampling-uncertainty guide explicitly treats sampling and physical preparation as part of measurement when the target is the larger sampling object. A precise analysis of an unrepresentative sample can therefore leave substantial uncertainty about the batch.
Variation can occur before the laboratory begins
Consider an imaginary container of mixed granules. The top, middle and bottom might differ if particle sizes separate during handling. Taking one small scoop from the surface could describe that scoop accurately while missing the distribution elsewhere. This example illustrates why sample location matters; it does not prescribe a universal sampling procedure.
A suitable plan depends on the material, batch boundaries, expected variation and purpose of the measurement. It should make clear how increments are selected, combined where appropriate, stored and reduced to the portion used for analysis. There is no single sample count that guarantees representativeness for every batch.
Repeating an analysis is not repeating sampling
Repeated instrument readings of the same prepared solution help examine one part of analytical variation. Preparing several portions of the same laboratory sample addresses another part. Neither automatically captures differences between independently collected samples from the batch.
Nordtest's sampling-uncertainty handbook provides practical guidance on this wider problem. Keeping the stages separate helps reveal what a repeatability claim actually demonstrates and what it leaves unexamined.
Read uncertainty statements with their scope
Measurement uncertainty characterises the dispersion associated with a reported value. A statement such as 10.0 plus or minus 1.0 requires context: which quantity was measured, what units apply and how was the interval calculated? The statement also needs to make clear which stages were included.
NIST's definition of expanded uncertainty describes the use of a coverage factor and the assumptions needed to associate an interval with a confidence level. An interval should not casually be read as a guarantee. Equally, an uncertainty estimate limited to laboratory analysis should not silently be treated as covering batch sampling.
Use the information to frame conclusions
Suppose two independently collected samples produce slightly different results. That difference alone does not establish that one laboratory made an error. Sampling variation, preparation and analytical uncertainty may all contribute. A useful investigation examines the procedures and the size of the difference before assigning a cause.
- Define the batch and the property being estimated.
- Record who sampled, where, when and by which procedure.
- Distinguish independent samples from repeated measurements.
- Check whether uncertainty includes sampling and preparation.
- Keep conclusions within the evidence actually collected.
Sampling quality and analytical quality answer connected questions. Understanding both makes a report more informative and prevents a narrow measurement from being mistaken for a complete description of every part of a batch.
Sources
- Eurachem: Uncertainty from Sampling
- Nordtest: Measurement uncertainty arising from sampling
- NIST: Expanded Uncertainty