What you'll learn
- How to determine sample size for MUS using the factors in ISA 530 .A10-A13 (confidence level, tolerable misstatement, expected misstatement)
- How to select items using systematic selection with a random start, and how to handle high-value items that exceed the sampling interval
- How to evaluate results when misstatements are found, including the tainting factor calculation and the upper misstatement limit
- When MUS is appropriate and when a different sampling method is better
Running monetary unit sampling (MUS) on a 12-line revenue test is the classic PIOOMA moment. You pick a sample size that looks reasonable, work backward to the risk, and hope no one in the review team asks where the reliability factor came from. It's also why inspection files keep flagging “insufficient coverage of the population tail” on receivables and inventory engagements.
MUS under ISA 530 treats every individual currency unit in a population as a sampling unit. Each euro has an equal probability of selection. Results get projected back to the population using a tainting factor. The sampling calculator on ciferi.com handles the sample size determination, selection interval, and results evaluation for a standard MUS engagement, so the file should tell a story rather than read like PIOOMA with extra decimals.
What MUS is and why auditors use it
MUS treats every individual monetary unit (every euro, every dollar) in the population as a sampling unit. A customer balance of €500,000 contains 500,000 sampling units. A balance of €200 contains 200. Larger balances get a proportionally higher probability of selection, which aligns with what the auditor actually cares about: larger balances carry more risk of material misstatement because they are larger.
ISA 530.5 defines audit sampling as "the application of audit procedures to less than 100% of items within a population of audit relevance such that all sampling units have a chance of selection." MUS satisfies this definition while building in a natural stratification. High-value items are almost certain to be selected (or pulled out and tested individually). Low-value items have a smaller but non-zero probability.
The method is efficient for testing account balances where overstatement is the primary risk (receivables, inventory, fixed assets). It requires a defined population with a recorded book value for each item. It does not work well for testing understatement (unrecorded items have zero book value and therefore zero probability of selection) or for populations with a large number of expected misstatements.
ISA 530 sampling concepts
ISA 530.6 -8 defines the concepts that underpin any sampling application, MUS included.
Tolerable misstatement (TE) is the maximum amount of misstatement in the population that the auditor is willing to accept and still conclude that the audit objective is met. In practice, TE for a substantive test is typically set equal to performance materiality (PM). ISA 530 .A3 confirms this relationship.
Expected misstatement is the auditor's estimate of the likely misstatement in the population before testing. If zero misstatements are expected, the sample size is smaller. If some misstatements are expected (based on prior-year (PY) results, control weaknesses, or known issues), the sample size increases. ISA 530 .A13 notes that higher expected misstatement requires a larger sample to conclude that the actual misstatement does not exceed TE.
The confidence level (or reliability factor, RF) reflects the level of assurance the auditor needs from the substantive test. In a combined approach (tests of controls plus substantive procedures), the auditor may accept a lower confidence level from the substantive test because controls provide partial assurance. In a fully substantive approach, the confidence level is higher. Common levels are 90% (RF 2.31 for zero expected errors) and 95% (RF 3.00).
Sample size determination
The sample size formula for MUS is Sample size = (Population book value x RF) / TE.
The RF comes from a Poisson probability table and depends on two inputs: the desired confidence level and the number of expected misstatements. For 95% confidence with zero expected errors, the RF is 3.00. For 95% confidence with one expected error, it is 4.75. For 90% confidence with zero expected errors, it is 2.31.
The sampling calculator computes the sample size and the sampling interval from your inputs. Here is the calculation done manually.
Suppose the population is €6,200,000. TE (set equal to PM) is €310,000. The auditor expects zero misstatements based on PY results showing no errors and effective controls over receivables. The confidence level is 95%.
Sample size = (€6,200,000 x 3.00) / €310,000 = 60 items.
The sampling interval is the population divided by the sample size: €6,200,000 / 60 = €103,333.
ISA 530 .A10-A13 identifies four factors that affect sample size. Higher confidence level increases sample size. Higher TE decreases sample size. Higher expected misstatement increases sample size. Larger population increases sample size (in MUS, this effect is embedded in the formula because the population value sits in the numerator).
If one misstatement is expected instead of zero, the RF rises to 4.75: sample size = (€6,200,000 x 4.75) / €310,000 = 95 items. The expected misstatement assumption has a large effect on sample size. That is why it has to be assessed carefully rather than defaulted to zero (the PIOOMA short-cut that regulators keep catching).
Selecting the sample
MUS uses systematic selection with a random start. The process works as follows.
Calculate the sampling interval (done above: €103,333). Generate a random start between 1 and the sampling interval. Suppose the random start is €47,218. The selection points are €47,218, €150,551 (47,218 + 103,333), €253,884, €357,217, and so on through the population.
Sort the population by a logical order (customer number, account code, transaction date). Assign a cumulative monetary amount to each item. The first item contributes its book value to the running total. The second item adds its book value. When the cumulative total crosses a selection point, the item containing that selection point is selected for testing.
High-value items
Any item with a book value equal to or greater than the sampling interval is certain to be selected (it will contain at least one selection point). These are the "individually significant items" and they are tested at 100%. In the example, any receivable balance of €103,333 or more is individually significant. Pull these out of the MUS population and test them separately. The MUS sample size then applies to whatever remains.
Negative balances
Credit balances in a receivables population (customer prepayments, credit notes) are awkward for MUS because they reduce the cumulative total and can distort the selection. Standard practice is to strip out negative balances, test them separately using a different method, and apply MUS only to the positive-balance population.
Testing the selected items
For each selected item, the auditor performs the planned substantive procedure. On a receivables test, that usually means sending a confirmation (conf) to the customer ( ISA 505 ), or, if the conf is not returned, running alternative procedures (inspecting subsequent cash receipts, checking shipping documentation).
The auditor records the book value of each selected item, the audited value (the amount the auditor concludes is correct), and the difference (if any). The difference is the misstatement for that item. Direction matters: overstatement versus understatement.
If the selected item is a logical unit (a complete customer balance) rather than an individual euro, the auditor tests the entire balance, not just the euro that was selected. This is the standard approach in MUS. The monetary unit is the selection unit. The logical unit (customer balance, invoice, inventory line) is the testing unit.
Evaluating results with the tainting factor
When no misstatements are found, evaluation is straightforward. The upper misstatement limit (UML) equals the basic precision: sampling interval x RF. With an interval of €103,333 and an RF of 3.00, the basic precision is €310,000. This equals TE, confirming that the population is not materially misstated at the desired confidence level.
When misstatements are found, the evaluation uses the tainting factor. The tainting factor for each misstatement is the ratio of the misstatement to the book value of the item in which it was found.
Suppose item A has a book value of €85,000 and an audited value of €72,000. The misstatement is €13,000. The tainting factor is €13,000 / €85,000 = 15.3%.
The projected misstatement for this error is tainting factor x sampling interval = 15.3% x €103,333 = €15,810.
The UML is calculated as basic precision + sum of (incremental RF x projected misstatement for each error). The incremental RF for the first error at 95% confidence is 1.75 (the difference between the RF for one expected error, 4.75, and the RF for zero expected errors, 3.00).
UML = (€103,333 x 3.00) + (1.75 x €15,810) = €310,000 + €27,668 = €337,668.
Compare this to TE (€310,000). The UML exceeds TE. The auditor has to consider extending the sample, performing additional procedures, or requesting management to investigate and correct the identified misstatement.
If management corrects the misstatement and the auditor is satisfied no similar errors remain, the evaluation can be reconsidered. Alternatively, the auditor increases the sample size and re-evaluates. The sampling calculator recalculates the UML automatically when you enter the misstatements found.
When MUS does not work well
MUS is not always the right method. ISA 530 .A8 notes that the auditor selects a sampling method based on the circumstances. Four situations make MUS a poor choice.
Testing for understatement
Unrecorded liabilities, omitted revenue, or understated expenses have a book value of zero (or are absent from the population entirely). MUS cannot select items that do not appear in the recorded population. For understatement testing, the auditor needs a different approach. Test from the source (subsequent payments, purchase orders, delivery notes) rather than from the recorded balance.
Populations with many expected errors
MUS works best when few or no misstatements are expected. When the auditor expects a high error rate (weak controls, PY history of multiple errors), the required MUS sample size gets very large. Classical variable sampling or stratified sampling is often more efficient.
Populations with many small balances and few large ones
MUS naturally selects larger items. If the population is dominated by thousands of small balances and the auditor is concerned about errors in those small balances, MUS can under-represent them in the sample. A stratified approach that ensures adequate coverage of both large and small items is often a better fit.
Populations with significant credit balances or negative values
Negative balances distort the cumulative monetary total. If credit balances are a substantial portion of the population, removing them can shrink the MUS population to a point where the remaining sample is not representative of the whole.
The ISAE 3402 template pack's Testing Protocol tab includes an embedded sample size reference for controls testing, which applies ISA 530 by analogy for ISAE 3402 engagements. The sample sizes differ from substantive MUS because controls testing uses attribute sampling (deviation rates) rather than variable sampling (monetary amounts), but the underlying ISA 530 framework governs both.
Worked example: Van Leeuwen Electronics N.V.
Scenario. Van Leeuwen Electronics N.V. distributes electronic components from a warehouse near Breda. Revenue €52M. Trade receivables (TR) at year-end €7.4M across 680 customer balances. Overall materiality (mat) €260,000. PM €169,000. No misstatements were found in PY testing. Controls over revenue and receivables are effective. The audit plan calls for MUS on the TR balance, testing for overstatement.
Remove individually significant items
Calculate the sampling interval first. With a preliminary sample size estimate of approximately 44 items (95% confidence, zero expected errors, RF 3.00), the interval is €7,400,000 / 44 = €168,182. The TE (set to PM) of €169,000 does not drive the interval directly. The correct formula for the interval is Population / Sample size. Sample size = (€7,400,000 x 3.00) / €169,000 = 131 items. Interval = €7,400,000 / 131 = €56,489. Any customer balance of €56,489 or more is individually significant. Remove these items (suppose 8 balances totalling €1,200,000). Remaining population: €6,200,000 across 672 balances. Recalculate. Sample size for remaining population = (€6,200,000 x 3.00) / €169,000 = 110 items. Interval = €6,200,000 / 110 = €56,364. Documentation note. Record the full population (€7.4M, 680 items), the individually significant items removed (8 items, €1.2M, tested 100%), and the remaining MUS population (€6.2M, 672 items). Record the confidence level (95%), expected misstatements (zero), RF (3.00), TE (€169,000), sample size (110), and sampling interval (€56,364).
Select the sample
Generate a random start between €1 and €56,364. Suppose the random start is €23,741. The selection points are €23,741, €80,105, €136,469, €192,833, and so on. Sort the 672 remaining balances by customer number. Assign cumulative monetary amounts. Select the customer balance that contains each selection point. Documentation note. Record the random start, the selection points, and the method for generating the random number (random number generator, firm methodology tool). Retain the sorted population with cumulative amounts as a WP.
Test the selected items
Send positive confs to all 110 selected customers plus the 8 individually significant customers. For non-replies, perform alternative procedures (inspect subsequent cash receipts through the testing date, examine underlying invoices and shipping documentation, then verify ageing). Documentation note. For each item, record the book value, the audited value, the response type (confirmed or alternative), and any difference. Document alternative procedures performed for non-replies.
Evaluate results with no misstatements
Suppose all 110 MUS items and all 8 individually significant items confirm without misstatement. The UML equals basic precision: €56,364 x 3.00 = €169,092. That sits within TE (€169,000, allowing for rounding). Conclude that the receivables population is not materially misstated at 95% confidence. Documentation note. Record the evaluation (zero misstatements found, UML €169,092, TE €169,000). Conclude that the results support the assertion that receivables are not materially misstated.
Evaluate results with one misstatement (alternative scenario)
Suppose one of the 110 MUS items has a book value of €42,000 and an audited value of €36,500. The misstatement is €5,500. The tainting factor is €5,500 / €42,000 = 13.1%. The projected misstatement is 13.1% x €56,364 = €7,384. The UML is basic precision (€169,092) + incremental allowance (1.75 x €7,384 = €12,922) = €182,014. That exceeds TE of €169,000. The auditor has to consider whether to extend the sample, request management to investigate and correct the misstatement, or expand substantive procedures on the receivables balance. Documentation note. Record the misstatement details (item, book value, audited value, misstatement, tainting factor). Record the projected misstatement and the UML calculation. Document the auditor's response to the exceedance (investigation requested from management, additional sample items selected, or the misstatement corrected and evaluation reconsidered).
Practical checklist
Common mistakes
- Setting expected misstatements to zero by default without considering PY results, control weaknesses, or known issues. The PCAOB has flagged this as an insufficient basis for the sample size determination. It understates the required sample when the auditor has information suggesting errors may exist.
- Failing to remove individually significant items before applying MUS, which means large balances are tested as part of the MUS sample rather than 100%, reducing the effectiveness of the test for those high-value items.
- Evaluating results by just extrapolating the raw misstatement amount (misstatement found x population / sample) rather than using the tainting factor approach. Raw extrapolation does not account for the relationship between the misstatement and the book value of the item in which it was found, and it does not incorporate the RF into the UML.
Frequently asked questions
What is monetary unit sampling and when should it be used?
Monetary unit sampling is a statistical method where each individual currency unit in a population has an equal chance of selection, meaning larger balances are naturally more likely to be selected. It is most effective for testing overstatement in populations like receivables, inventory, and revenue where few errors are expected (ISA 530.5, ISA 530.A8). MUS is not suitable for testing understatement because unrecorded or understated items have zero or reduced book value and therefore near-zero selection probability.
How is the MUS sample size determined under ISA 530?
The formula is Sample size = (Population book value × Reliability factor) / Tolerable misstatement. The reliability factor is drawn from a Poisson probability table based on the desired confidence level and expected number of errors (for example, 3.00 at 95% confidence with zero expected errors, per ISA 530.A10–A13). Tolerable misstatement is typically set equal to performance materiality for the assertion being tested.
How should individually significant items be handled in a MUS engagement?
Items with a recorded value equal to or exceeding the sampling interval should be removed from the MUS population and tested individually at 100%. Including them in the MUS population produces unreliable evaluations because the tainting factor formula does not correctly account for items tested with certainty (ISA 530.A13). The proper approach is to stratify: test high-value items separately, apply MUS to the remainder, and combine the results.
How does the tainting factor evaluation work when misstatements are found?
For each misstatement, the tainting factor is calculated as the misstatement divided by the book value of that item. The projected misstatement is the tainting factor multiplied by the sampling interval. The upper misstatement limit equals the basic precision plus the sum of incremental reliability factors multiplied by each projected misstatement (ISA 530.A20–A21). If the upper misstatement limit exceeds tolerable misstatement, the auditor must extend the sample or expand procedures.
Why is it incorrect to set expected misstatements to zero without justification?
Setting expected misstatements to zero reduces the reliability factor, which directly shrinks the sample size. The auditor must consider prior-year results, control weaknesses, and known issues before concluding that zero errors are expected (ISA 530.A13). Defaulting to zero without this assessment understates the required sample size and has been flagged by regulators as an insufficient basis for the determination.
Related content
- Audit sampling glossary entry. Defines the key ISA 530 concepts (sampling risk, non-sampling risk, tolerable misstatement, expected misstatement) with paragraph references and a concise explanation of when statistical versus non-statistical sampling is appropriate.
- ISA 530 sampling calculator. The free tool that automates sample size calculation and sampling interval determination for both MUS and attribute sampling, with built-in results evaluation, referenced throughout this post.
- ISA 450 misstatement accumulation. Covers how projected misstatements from sampling (one of the three ISA 450.5 types) are accumulated alongside factual and judgmental misstatements for the aggregate evaluation against materiality.