What is non-statistical sampling?
Non-statistical sampling is any audit sampling approach where the auditor does not use probability theory to evaluate results or quantify sampling risk. ISA 530.7 permits it explicitly — the decision between statistical and non-statistical sampling is a matter of professional judgement, not a hierarchy of quality.
In practice, most audit firms use non-statistical sampling as their default. Selection methods include haphazard selection (picking items without a deliberate pattern but also without a random number generator) and systematic selection with a random start. Firm methodology tables typically provide pre-calculated sample sizes based on assessed risk levels, acting as a proxy for the statistical inputs the auditor would otherwise specify.
ISA 530.A6 requires the auditor to consider population characteristics when designing any sample, and ISA 530.A13 warns against drawing statistical conclusions from a non-statistical sample. The evaluation must remain consistent with the design: if the sample was not designed statistically, the conclusion must be based on professional judgement, not on implied confidence levels or extrapolated error limits.
Key Points
- Equally valid per ISA 530.A10. The standard treats statistical and non-statistical sampling as equivalent when properly applied. Neither is superior by default.
- Must still project errors to the population. ISA 530.14 requires the auditor to project misstatements found in the sample to the population, regardless of the sampling approach used.
- Sample must still be representative. Using non-statistical sampling does not relax the requirement that the sample should be expected to be representative of the population (ISA 530.A6).
- Documentation burden is heavier. Without recalculable statistical inputs, the auditor must clearly document the rationale for sample size, selection method, and the basis for the conclusion.
Why it matters in practice
A recurring finding from the FRC and other regulators is teams using firm methodology table sizes without adjusting for the specific engagement's circumstances. A table might specify 25 items for a "low risk" control, but if the population has unusual characteristics — such as heavy concentration in a few accounts or seasonal clustering — the standard size may be insufficient. ISA 530.A6 expects the auditor to consider these factors, not simply default to the table.
The second common issue is omitting the projection step required by ISA 530.14. When errors are found in a non-statistical sample, teams sometimes conclude "the errors are immaterial" based on the absolute amount found in the sample, without projecting that rate or amount to the full population. This is both a documentation failure and a logical gap — the sample exists to draw inferences about the population, and skipping the projection undermines the purpose of sampling entirely.
Key standard references
- ISA 530.5(g): Definition distinguishing statistical from non-statistical sampling.
- ISA 530.7: The choice between statistical and non-statistical sampling is a matter of professional judgement.
- ISA 530.14: Requirement to project misstatements found in the sample to the population.
- ISA 530.A6: Consideration of population characteristics when designing the sample.
- ISA 530.A10: Both approaches can provide sufficient appropriate audit evidence.
- ISA 530.A13: Warning against drawing statistical conclusions from non-statistical samples.
Related terms
Related reading
Frequently asked questions
Is non-statistical sampling less valid than statistical sampling?
No. ISA 530.A10 treats both as equally valid when applied properly. The standard does not rank one above the other. The difference is in the evaluation: statistical gives quantified confidence, non-statistical uses professional judgment.
Must the auditor still project errors in non-statistical sampling?
Yes. ISA 530.14 requires projection regardless of the sampling approach. Finding errors in a sample and concluding 'errors are immaterial' without projecting to the population is both a documentation failure and a logic failure.
What is the main risk with non-statistical sampling?
The documentation burden is heavier because the reviewer cannot recalculate the basis for your conclusion. A thin narrative ('no material misstatement exists') without explaining the projected misstatement and why it supports the conclusion is incomplete.