Audit Sampling Explained: Attribute Sampling & MUS
Attribute sampling for controls, monetary unit sampling for balances, sample size tables and reproducible random selection under ISA 530.
Editorial disclosure: Pyalm publishes and maintains the products and free tools discussed on this site. Regulatory statements are linked to primary sources where applicable. No independent professional review is claimed unless a reviewer is explicitly named.
You can't test every transaction, so auditors test a sample and draw a conclusion about the population. ISA 530 requires the sample to be designed so that it gives a reasonable basis for that conclusion. This guide covers the two statistical methods you will meet most often and the sample sizes they produce.
Statistical vs non-statistical
Both are acceptable. Statistical sampling uses random selection and probability theory to quantify sampling risk. Non-statistical sampling relies on judgement. Statistical methods make the sample size defensible and repeatable; they don't remove the need for judgement about the population, the tolerable rate and what to do with errors found.
Attribute sampling: testing controls
Use attribute sampling when you ask "did the control operate?" — was the invoice approved, was the three-way match done. Three inputs drive the size:
- Confidence level (commonly 90% or 95%).
- Tolerable deviation rate — the highest failure rate you could accept and still rely on the control.
- Expected deviation rate — what you anticipate from prior years or a pilot.
| Confidence | Tolerable rate | Expected 0% | Expected 1% |
|---|---|---|---|
| 95% | 5% | 59 | 93 |
| 95% | 10% | 29 | 46 |
| 95% | 3% | 99 | 257 |
| 95% | 2% | 149 | 590 |
| 90% | 5% | 45 | 77 |
The "59 items" figure for a 5% tolerable rate at 95% confidence with no expected deviations comes from the binomial formula: 0.95^59 is just under 0.05. If you find a deviation in a zero-expected plan, you either expand the sample or conclude the control isn't reliable at that tolerable rate.
Monetary unit sampling (MUS): testing balances
MUS suits substantive testing of balances where you worry about overstatement — receivables, inventory, fixed assets. Every currency unit in the population is a sampling unit, so larger items are more likely to be picked automatically.
Sample size = book value × reliability factor ÷ (tolerable misstatement − expected misstatement × expansion factor)
Reliability factors for zero errors are about 2.31 at 90% confidence, 3.00 at 95% and 4.61 at 99%. The sampling interval is book value divided by sample size. Choose a random start within the first interval and select the item containing every n-th unit.
Worked example: book value AED 5,000,000, tolerable misstatement AED 150,000, no expected errors, 95% confidence → 5,000,000 × 3.00 ÷ 150,000 = 100 items, interval AED 50,000. Any item larger than AED 50,000 is tested individually.
Selecting items you can reproduce
Whichever method you use, document how the items were chosen. A seeded random number generator lets a reviewer regenerate the identical selection from the population size, sample size and seed. Record all three in the working paper.
Try it
The free Audit Sample Size Calculator computes attribute sample sizes, the MUS size and interval, and draws a seeded random selection — all in your browser. Set the tolerable misstatement from performance materiality using the Audit Materiality Calculator and read audit materiality explained for the logic.
Evaluating what you find
Projecting errors is the step people skip. If you find two deviations in a plan that allowed one, you can't conclude the control is effective at the tolerable rate. For MUS, project the misstatement found over the population using the sampling interval and add an allowance for sampling risk before comparing with tolerable misstatement.
Clean source data makes sampling easier: if your population is a ledger export, Pyalm Books keeps invoices, bills and journals together so the listing you sample from is complete and consistent.
Standards, rates and deadlines change. Confirm the current position with the relevant authority or a qualified adviser before you rely on it.