In short
- Sampling risk is the risk that the sample conclusion differs from the conclusion on the whole population; non-sampling risk is everything else.
- Define the population before selecting: completeness, cut-off and the sampling unit decide whether the result can be extrapolated.
- Statistical and non-statistical sampling are both permitted; the method must be appropriate and documented, not just 'judgmental'.
- Investigate every exception, decide whether it is an anomaly, and extrapolate the misstatement unless the item is isolated.
- In Cambodia, test the completeness of the ledger and reconcile it to VAT and Tax on Income data before sampling.
Sampling risk and non-sampling risk
Audit sampling is the application of audit procedures to less than 100% of the items in a population so that each sampling unit has a chance of selection, allowing the auditor to draw a conclusion about the whole population. Sampling risk is the risk that the auditor's conclusion based on the sample differs from the conclusion that would be reached if the whole population were tested. Non-sampling risk is the risk that the auditor reaches a wrong conclusion for any other reason, such as a poorly designed procedure, a misunderstanding of the document or a failure to recognise a misstatement.
For a test of controls, sampling risk is the risk that the auditor concludes the control is more effective than it is, or less effective. For a substantive test, sampling risk is the risk that the auditor concludes a balance is not materially misstated when it is, or that it is materially misstated when it is not. The auditor addresses sampling risk through the sample size, the selection method and the evaluation; non-sampling risk is addressed through planning, supervision and review.
The auditor should not describe a test as statistical unless the sample was selected randomly and the evaluation uses probability theory. A judgmental selection can be useful, but it does not support a statistical projection of the misstatement.
Population, sampling unit and selection
The auditor defines the population to match the assertion and the direction of testing. For an existence test of receivables, the population is the customer list in the ledger; for a completeness test, the population may need to come from an independent source, such as shipping documents, bank receipts or tax invoices. A population that is incomplete or includes the wrong period cannot be fixed by a larger sample.
The sampling unit can be an invoice, a customer balance, a journal entry, a payment, a stock item or a line item. The auditor may use random selection, systematic selection with a random start, monetary unit sampling, or a haphazard selection. Monetary unit sampling gives larger balances a higher chance of selection and is useful for substantive tests of overstatement; it is less effective for completeness.
The auditor should test the completeness and accuracy of the population before sampling. A ledger extract that does not reconcile to the trial balance, a sales list that excludes credit notes or a purchase list that excludes imports is not a valid population. The reconciliation should be in the file.
| Method | Use | Main risk |
|---|---|---|
| Random selection | Any population; supports statistical evaluation | Population must be complete and numbered |
| Systematic selection | Large, homogeneous populations | Pattern in the population can bias the sample |
| Monetary unit sampling | Substantive tests of overstatement; large balances | Not suitable for completeness or understatement |
| Haphazard selection | Small, homogeneous populations; judgment | Not statistical; difficult to defend |
| 100% testing | Small population or high-risk items | Not sampling; no extrapolation |
Sample size and evaluation
The sample size depends on the sampling risk the auditor is willing to accept, the tolerable misstatement, the expected misstatement and, for statistical sampling, the confidence level. A higher risk of material misstatement, a lower tolerable misstatement or a higher expected misstatement increases the sample size. The auditor documents the basis for the sample size, even when the sample is non-statistical.
Every exception is investigated. The auditor determines whether the item is a misstatement, whether it is an anomaly and whether it has a broader cause. An anomaly is a misstatement that is not representative of the population, and the auditor should obtain sufficient evidence that it is isolated before excluding it from the projection. A misstatement that is not an anomaly is projected to the population, and the projected misstatement is compared with tolerable misstatement.
If the projected misstatement exceeds tolerable misstatement, the auditor performs additional procedures, increases the sample, tests a different population or concludes that the balance is materially misstated. The auditor does not simply select more items until the result looks acceptable; the additional work must address the risk.
A judgmental sample is not a free pass
Selecting the largest 20 invoices is not audit sampling unless every item had a chance of selection. It can be a valid targeted test, but the conclusion cannot be extrapolated to the population.
Cambodian data and practical populations
Cambodian audits often rely on manual ledgers, spreadsheets and accounting software with limited user controls. The auditor should test the completeness of the ledger, the cut-off around the year-end and the reconciliation to the VAT return and the annual Tax on Income return. A sales population that does not agree to the VAT return turnover is a risk indicator, not a shortcut.
Common populations are sales invoices (VAT output), purchase invoices (VAT input), payroll lines (Tax on Salary and NSSF), inventory items, fixed-asset additions, journal entries and related-party balances. The auditor can use data analytics to select the population and test 100% of a small population, but the same documentation and evaluation requirements apply.
For a small company, the population may be small enough to test 100%. That is not sampling, and the auditor should say so. For a large company with many low-value transactions, sampling is efficient, but the auditor should still test the high-risk items individually.
- Reconcile the ledger to the trial balance and the tax returns.
- Check the cut-off around the year-end and the tax period.
- Identify related-party and unusual items for 100% testing.
- Use data analytics for the population, not as a substitute for the evaluation.
- Document the sample, the exceptions and the projection.
Documentation
The audit file should describe the population, the sampling unit, the selection method, the sample size, the basis for the sample size, the items selected, the procedures performed, the exceptions, the evaluation and the conclusion. The documentation should be sufficient for an experienced auditor to reproduce the sample and understand the judgement.
ISA 230 applies. The auditor should also document the linkage to the risk assessment and the planned response under ISA 330, and any use of the work of an expert or an internal auditor. A file that contains only a list of selected invoices and a tick mark is not enough.
Key facts
- Standard
- ISA 530
- Audit sampling and evaluation.
- Risk
- Sampling risk and non-sampling risk
- Sample size and method address sampling risk.
- Population
- Complete and relevant
- Test completeness before selecting.
- Exceptions
- Investigate, project, document
- Anomalies require evidence of isolation.
How this article is checked
Figures were verified against the sources listed below in November 2026 and are re-checked each January or when the rule changes. Written by Liv Bonito, Co-founder & Managing Director. General information, not advice on your situation: for that, talk to a partner.