Sampling and Analysis Share One Uncertainty Budget

Overview

Short answer: because a laboratory instrument only reports the property of the specific aliquot presented to it; the act of taking that aliquot from a heterogeneous lot introduces its own, often dominant, uncertainty. To understand what a reported measurement actually says about a bulk material you must consider sampling, sample handling, and analysis together as a single measurement process with a combined uncertainty budget .

Why this matters A precise, well-calibrated instrument can give a repeatable number for the piece of material you feed into it. But if that piece is not representative of the bulk—because the material is spatially heterogeneous, segregated by particle size, or unevenly mixed—that precise number can mislead. Sampling uncertainty quantifies the risk that the aliquot differs from the intended population; analytical (instrument) uncertainty quantifies the imprecision and bias of measuring that aliquot. Both uncertainties are part of the same measurement statement and must be brought together when interpreting results for decisions about a batch, lot, or population .

How sampling and analysis combine

Concrete consequences of ignoring sampling uncertainty

Evidence on magnitude and behavior Peer-reviewed work on analytical practice demonstrates that sampling uncertainty is frequently a leading term in total variance for many real-world matrices, particularly heterogeneous and particulate materials. Studies and reviews show how between-increment heterogeneity can dominate the uncertainty budget, and how statistical designs that capture between-increment variability (replicate increments, composite sampling, stratified sampling) reduce the overall uncertainty more effectively than repeated instrument runs on the same aliquot . EURACHEM/CITAC provides both conceptual framing and practical methods to estimate sampling variance and combine it with analytical variance into an overall uncertainty estimate .

A practical approach to reading evidence and reports When you evaluate a reported measurement for its implications about a bulk lot, look for these elements:

When the available evidence is incomplete If reports omit sampling-plan details or between-increment variability, state exactly what remains unresolved: whether the aliquot was representative; how much of the total uncertainty is unknown; and whether conclusions rely on assumptions about homogeneity. Do not infer representativeness from instrument precision alone. Ask for or seek:

Simple table to clarify evidence types

| Evidence item | What it indicates | |---|---| | Between-increment variance | Direct measure of sampling uncertainty | | Instrument repeatability and calibration data | Analytical uncertainty of the measurement method | | Sampling plan and population definition | Whether increments were likely representative | | Handling and preparation records | Possible additional sources of variance |

Bottom line A measurement that aims to describe a bulk material is only as trustworthy as the weakest link in the chain from population to result. Sampling uncertainty can, and often does, rival or exceed instrument uncertainty. Proper interpretation requires treating sampling and analysis as a single combined uncertainty budget and demanding evidence that both components have been assessed. Keep conclusions sample-, method-, matrix-, and date-specific: absence of reported sampling data leaves an unresolved gap in what the measurement can legitimately claim about the whole material .

• The measurement process starts with defining the measurand: what property of what population are you trying to know? Only after that can you design a sampling plan intended to produce representative increments. EURACHEM/CITAC places sampling and analysis in series: planning → taking increments → preparing a laboratory sample → instrument measurement → result reporting. Each step contributes uncertainty that propagates to the final estimate .
• Sampling uncertainty often arises from spatial heterogeneity and the mechanics of taking increments. For particulate or mixed materials, variance between increments can be large compared with the instrument’s repeatability. Analytical uncertainty arises from factors such as calibration, instrument noise, interferences and method reproducibility .
• These contributions are mathematically combined into a total uncertainty (usually by propagation of variances) so that the final measurement statement reflects both the chance that the aliquot was unrepresentative and the chance that the instrument mismeasured the aliquot .
• Overconfidence: A small instrument standard deviation can give the illusion of a tight result band even when subsampling variability is large. Treating the instrument uncertainty alone underestimates total uncertainty and can lead to false confidence in conformity or non-conformity decisions.
• Misclassification: For heterogeneous lots, single-aliquot results can randomly over- or under-estimate true lot content. Decisions based on single precise measurements may therefore be wrong at a known frequency unless sampling variance is accounted for .
• Wasteful confirmatory testing: When sampling is poor, repeating the instrument measurement of the same aliquot will not reduce sampling variance. Resources are better spent on improved sampling plans or on increasing the number of independent increments, not only on repeated analyses of the same prepared sample.
• Definition of population and measurand: Is it clear what material the result is intended to represent and how that population was delimited?
• Sampling plan description: How were increments chosen (random, systematic, stratified)? How many increments were taken? Were increments combined or analysed separately?
• Replicate and between-increment data: Does the report show variability between independent increments (not just repeat measures of the same aliquot)? Estimates of between-increment variance are necessary to quantify sampling uncertainty .
• Uncertainty budget: Does the report present a combined uncertainty that includes both sampling and analytical components? If only instrument uncertainty is reported, the total uncertainty for the population is incomplete.
• Chain-of-custody and handling notes: Sample preparation, subsampling, transport and storage can introduce additional variance. Reports should state handling procedures and any known limitations .
• replicate independent increments, ideally analysed separately and reported,
• descriptions of how increments were selected and whether stratification or composite sampling was used,
• explicit combining of sampling and analytical variances into a single uncertainty estimate .