Audit sampling means testing fewer than all the items in a population in a way that gives every item a chance of selection, so the auditor can draw a conclusion about the whole population. ISA (UK) 530 requires the auditor to design the sample around the purpose of the test, choose a size that keeps sampling risk acceptably low, investigate every error found, project errors to the population and judge whether the result is good enough. Picking only the largest or oddest items is not sampling, and its results cannot be projected.
When do auditors sample instead of testing everything?
ISA (UK) 500 paragraph 10 requires the auditor to choose a way of selecting items that is effective for the purpose of the test. Paragraph A63 sets out three options, which can be combined:
- Selecting all items. Paragraph A64 says 100% examination may suit a population made up of a small number of large items, a significant risk where other means do not give enough evidence, or a repetitive automated calculation that can be checked efficiently.
- Selecting specific items. These might be high value items, all items over a set amount, or items that are unusual, suspicious or have a history of error (paragraph A65).
- Audit sampling, covered by ISA (UK) 530.
The distinction is important. Paragraph A66 states that selecting specific items is not audit sampling, and its results cannot be projected to the rest of the population. If an auditor tests every sales invoice over £20,000, they know nothing about the invoices below that amount unless they test those separately or have other evidence.
What counts as audit sampling?
Paragraph 5(a) of ISA (UK) 530 defines audit sampling as applying audit procedures to less than 100% of items within a population of audit relevance such that all sampling units have a chance of selection. The objective in paragraph 4 is to provide a reasonable basis for conclusions about the population from which the sample is drawn.
Sampling can be statistical or non statistical. Under paragraph 5(g), statistical sampling uses random selection and probability theory to evaluate the results, including measuring sampling risk. Anything else is non statistical. Paragraph A9 says the choice is a matter for the auditor’s judgement, and that sample size alone does not tell you which approach was used.
Sampling risk is the risk that the conclusion from the sample differs from the conclusion that testing everything would give (paragraph 5(c)). The auditor is chiefly concerned with concluding that controls work, or that no material misstatement exists, when the opposite is true, because that leads to the wrong opinion. Non sampling risk covers everything else, such as using the wrong procedure or failing to spot an error (paragraph A1).
How is the sample size decided?
Paragraph 7 requires a sample size sufficient to reduce sampling risk to an acceptably low level. Paragraph A10 says the lower the risk the auditor will accept, the larger the sample. Paragraph A11 allows the size to be set by a statistical formula or by professional judgement. Appendices 2 and 3 list the factors.
| Factor | Effect on sample size |
|---|---|
| Higher assessed risk of material misstatement (tests of details) | Increase |
| More reliance on other substantive procedures for the same assertion | Decrease |
| Higher tolerable misstatement | Decrease |
| More misstatement expected in the population | Increase |
| Stratifying the population where appropriate | Decrease |
| More reliance placed on controls (tests of controls) | Increase |
| Higher tolerable rate of deviation (tests of controls) | Decrease |
| Number of items in a large population | Negligible |
Tolerable misstatement is the amount of error the auditor is prepared to accept in the population. Paragraph A3 describes it as the application of performance materiality to a particular sampling procedure, which may be the same amount or lower. That is why the work on materiality in an audit feeds directly into sample sizes. Paragraph A7 adds that if the expected rate of control deviations is unacceptably high, the auditor will normally not test controls at all.
How are the items selected?
Paragraph 8 requires selection so that every sampling unit has a chance of being chosen, and paragraph A12 stresses that the sample should be representative and free from bias. Appendix 4 describes the methods:
- Random selection, using random number generators or tables.
- Systematic selection, picking every nth item from a starting point, while checking that the population does not have a pattern matching the interval.
- Monetary unit sampling, a value weighted method in which the sampling unit is each individual monetary unit, so larger balances have a greater chance of selection.
- Haphazard selection, without a structured technique but avoiding conscious bias. It is not appropriate for statistical sampling.
- Block selection, a run of consecutive items, which can rarely support conclusions about the whole population.
Appendix 1 explains stratification, dividing the population into groups, often by value, so effort goes where the risk is. Results for one stratum can only be projected to that stratum.
What if a selected item cannot be tested?
If a procedure is not applicable to an item, the auditor tests a replacement (paragraph 10). Paragraph A14 gives the example of a properly voided cheque selected when testing payment authorisation. But if the auditor cannot apply the planned procedure or a suitable alternative, paragraph 11 requires the item to be treated as a deviation in a test of controls, or as a misstatement in a test of details. Paragraph A15 gives lost documentation as an example. Paragraph A16 suggests that where a customer does not reply to a confirmation request, subsequent cash receipts can be examined instead.
For a client, the practical point is simple: a missing invoice or contract is not a neutral event. It can count as an error and push the projected misstatement up.
How are errors in the sample evaluated?
Paragraph 12 requires the auditor to investigate the nature and cause of every deviation or misstatement and consider its effect on other areas. Paragraph A17 notes that errors with a common feature, such as one location or one period, may lead the auditor to test every item with that feature, and that errors may be intentional and indicate fraud.
An error may be treated as an anomaly, meaning demonstrably not representative of the population, only in extremely rare circumstances and only with additional procedures that give a high degree of certainty (paragraph 13). For tests of details, paragraph 14 requires misstatements to be projected to the population. Paragraph A22 says the projected misstatement plus any anomalous misstatement is the auditor’s best estimate, and if it exceeds tolerable misstatement the sample does not give a reasonable basis for conclusions. For tests of controls, paragraph A20 explains that the sample deviation rate is itself the projected rate, and ISA (UK) 330 paragraph 17 requires the auditor to decide whether reliance is still justified, more controls testing is needed or substantive procedures must be used instead.
How might the projection work?
Illustration only. Trade debtors total £500,000 across 400 balances. The auditor tests every balance over £25,000, which covers £150,000, and samples £70,000 of the remaining £350,000. The sample contains overstatements of £1,400, which is 2% of the value tested. Using a simple ratio projection, the projected misstatement in the sampled stratum is 2% of £350,000, or £7,000. Any errors in the specifically selected large balances are added in full.
If tolerable misstatement for debtors is £10,000, a projected £7,000 is uncomfortably close. Under paragraph A23 the auditor may ask management to investigate and correct the errors, or extend the testing. Paragraph A18 warns that the projection gives a broad view of the scale of error but may not be enough to determine the adjustment to record, so management usually has to identify the actual errors.
What mistakes do clients make?
- Offering the auditor a “representative” bundle of invoices. The auditor must choose the items, and a sample chosen by the client gives no basis for a conclusion.
- Failing to produce support for selected items, which turns gaps into misstatements.
- Correcting only the errors found in the sample. Projected errors still count when the auditor evaluates uncorrected misstatements under ISA (UK) 450.
- Supplying listings that do not agree to the ledger. Paragraph A5 links sampling to evidence that the population is complete, so a listing that does not reconcile delays the work. Good source documents and an audit trail make sampling faster.
How sampling fits the risk based plan is explained in our page on audit risk assessment under ISA (UK) 315, and the wider process in how a statutory audit works.
Solar Protect provides tax investigation insurance for businesses and fee protection schemes for accountancy practices. Read how our cover works or get a quote.
Related Accountapedia entries
- What counts as sufficient appropriate audit evidence
- Journal entry testing by auditors
- Analytical procedures in an audit
- Fraud and the auditor: ISA (UK) 240
- Group audits and component auditors
Sources
- Financial Reporting Council: ISA UK 500 Updated September 2025
- Financial Reporting Council: ISA UK 530 Updated September 2025
- Financial Reporting Council: ISA UK 330 Revised July 2017 Updated September 2025
- Financial Reporting Council: ISA UK 450 Revised June 2016 Updated September 2025
Reviewed by Christine Anderson, Broker, Solar Protect. Last reviewed 8 October 2026. Part of Accountapedia, the Solar Protect guide to UK tax and accounting. This page is general information, not personalised advice.