Random sampling reduces the role of personal preference in deciding which units are tested. It starts with a defined population and a selection process that gives units a known chance of inclusion. It does not mean choosing whichever vials happen to be nearby, and it does not guarantee that a small sample captures every unusual unit.
Define which units were eligible
NIST’s introduction to acceptance sampling describes random selection from a lot as part of a plan for making a lot-disposition decision. The population from which the units are selected is therefore part of the evidence.NIST — What is acceptance sampling? (opens in a new tab)
In an original example, a lot contains 1,000 vials in ten cartons. Selecting random vial positions only from the first carton does not give the other nine cartons a chance of inclusion. The choice within that carton can be random while the lot-wide claim remains unsupported.
A sampling frame is the list or practical arrangement from which selection is made. Check whether it includes the entire stated lot, whether units can be identified reliably and whether any were excluded before selection.
Separate planned randomisation from convenience
| Approach | What the description reveals |
|---|---|
| Numbered units chosen by a recorded random process | A reproducible account of chance-based selection |
| The nearest unopened vials | Selection depends on access and position |
| Vials chosen for a tidy appearance | Selection depends on an observed product attribute |
| One unit from each defined carton by random position | A plan combines coverage of cartons with random selection within them |
The final example is a stratified design, not the same design as selecting freely from every unit in the lot. Its analysis should respect how the groups were defined and how many units each group represents.
Convenience samples can still reveal problems in the units tested. Their limitation is the inference beyond those units, particularly if the reason they were easy to select is related to the property being assessed.
Record the method before seeing analytical results. Replacing an inconvenient selected vial with a preferred one changes the selection and needs to remain visible.
Consider structure within the lot
NIST’s sampling-scheme guidance discusses stratification and randomisation as tools for reducing confounding and systematic error in process studies. The design should reflect the factors that could otherwise be mixed together.NIST — Choosing a sampling scheme (opens in a new tab)
For peptide vials, relevant grouping information might include filling sequence or storage location if the assessment is concerned with those differences. Naming those factors does not assert that a particular lot has a defect; it defines where variation could be investigated.
If early and late units are deliberately sampled separately, retain those labels. Combining all results immediately into one mean can discard the comparison that motivated the design.
Systematic selection, such as every twentieth unit, also needs consideration of any repeating pattern in the process. The regular spacing alone is not equivalent to a simple random sample.
Keep chance variation and measurement error in view
A properly random sample can miss uncommon defects by chance. That is a sampling limitation, not evidence that randomisation failed. The number of independent units and the decision rule determine how much reassurance the design can provide.
Selection also does not validate the assay. If the analytical method cannot detect the attribute of concern, a well-selected set of vials still cannot answer that analytical question.
A useful sampling statement identifies the lot, eligible units, selection mechanism, sample size and any departures. It should distinguish independent vials from repeat measurements of the same vial.
Sources and further detail
- NIST — What is acceptance sampling? (opens in a new tab)
Official handbook introduction used for random lot selection, not a recommended peptide acceptance plan.
- NIST — Choosing a sampling scheme (opens in a new tab)
Official discussion of stratification, randomisation and confounding. Vial/carton examples are original.
Sources checked 19 September 2026. Worked examples are illustrative unless a supplied report is explicitly identified. This article has not undergone independent scientific peer review.