Audit materiality and sampling

What audit materiality and sampling are

Audit materiality is the size of misstatement, alone or in aggregate, that could reasonably be expected to influence the economic decisions of users of the financial statements. Audit sampling is testing less than 100% of a population so that every item has a chance of selection and a conclusion can be drawn about the whole. Together they decide how much audit work is needed and which errors matter.

Four standards govern the area. ISA 320 sets materiality for planning and performance, ISA 450 governs how identified misstatements are accumulated and evaluated, ISA 530 covers audit sampling, and ISA 500 decides when a technique other than sampling is the better answer. The percentages quoted below are typical of firm methodologies; the standards themselves leave the numbers to professional judgement.

Overall materiality: choosing the benchmark

ISA 320 para 10 requires the auditor to determine materiality for the financial statements as a whole when setting the overall audit strategy. The usual method is a percentage of a benchmark that users focus on. For a profit-oriented entity that is normally profit before tax from continuing operations; the standard's own application material gives 5% of that figure for a manufacturer, and 1% of total revenue or total expenses for a not-for-profit entity, as examples.

The benchmark must be stable enough to be meaningful. When profit is volatile, close to break-even or distorted by one-off items, auditors normalise it (for example, averaging three years or excluding a disposal gain) or switch to revenue, gross profit or total expenses. Asset-based benchmarks suit investment entities and property companies, where users care about net asset value.

Para 10 also requires lower, specific materiality for particular items where smaller misstatements would influence users, such as directors' remuneration, related party transactions or a regulatory capital figure.

• Profit before tax: commonly 5%, with a range of about 3% to 10% depending on risk and public interest
• Revenue: commonly 0.5% to 1%, sometimes up to 2% for low-risk private entities
• Total expenses: commonly 0.5% to 2%, typical for not-for-profit and public sector bodies
• Gross profit: commonly 1% to 3%, where operating costs are volatile
• Total assets or net assets: commonly 1% to 2%, for asset-holding or investment entities

Performance materiality and the clearly trivial threshold

Performance materiality under ISA 320 para 11 is set below overall materiality so that the aggregate of uncorrected and undetected misstatements is unlikely to exceed it. Firms typically set it between 50% and 75% of overall materiality: towards 75% for a continuing audit with good controls and few past errors, towards 50% for a first-year audit, weak controls, a history of adjustments or a high-risk area. It drives sample sizes and the threshold for investigating analytical differences.

ISA 450 para 5 requires the auditor to accumulate every misstatement found except those that are clearly trivial, which the application material describes as clearly inconsequential by any criterion of size, nature or circumstance. Firms usually set the clearly trivial threshold at 3% to 5% of overall materiality, occasionally up to 10%. A small item is still not trivial if it indicates fraud. ISA 320 para 12 requires materiality to be revised if information emerges during the audit, such as final results well below the forecast used at planning.

Worked example: setting the three thresholds

A manufacturing company reports revenue of EUR 60 million and profit before tax of EUR 3.0 million, which includes a one-off gain of EUR 600,000 on selling a warehouse. Users focus on recurring profit, so the auditor normalises the benchmark to EUR 2.4 million.

At 5%, overall materiality is EUR 120,000. As a sense check, that is 0.2% of revenue, well inside the revenue range, which confirms the profit benchmark is not producing an unusually high figure. The engagement is a continuing audit with a clean history, so performance materiality is set at 75%, EUR 90,000. Had it been a first-year audit, 60% would give EUR 72,000. Clearly trivial at 5% of overall materiality is EUR 6,000.

The file should record why each percentage was chosen. A percentage at the top of a range with no documented reason is exactly what a file reviewer or audit regulator will challenge.

• Normalised profit before tax: 3,000,000 minus 600,000 = EUR 2,400,000
• Overall materiality: 5% x 2,400,000 = EUR 120,000
• Performance materiality: 75% x 120,000 = EUR 90,000
• Clearly trivial: 5% x 120,000 = EUR 6,000

When sampling is the right tool, and when it is not

ISA 530 applies only when the auditor chooses audit sampling. Often a better route exists. Selecting all items above a threshold, or all items with a risk characteristic, is targeted testing, not sampling: it gives evidence about the items tested but supports no conclusion about the untested remainder. Testing 100% of a population with data analytics, such as recomputing every invoice's VAT, removes sampling risk entirely. Our guide on journal entry testing and fraud red flags applies the same whole-population approach to journals.

Sampling risk is the risk that the conclusion from the sample differs from the conclusion from testing everything. Non-sampling risk is the risk of reaching a wrong conclusion for any other reason, such as applying the wrong procedure or missing an error in an item that was tested. A larger sample reduces the first but not the second. ISA 530 also requires the auditor to consider whether an item is an anomaly, meaning demonstrably not representative, but the bar is high and the reasoning must be documented.

Sampling methods compared

Statistical sampling uses random selection and probability theory to measure sampling risk. Non-statistical sampling relies on judgement for sample size and evaluation, but ISA 530 still requires selection that gives every sampling unit a chance of being picked. Tests of controls use attribute sampling (is the control performed or not); tests of details use variables sampling or monetary unit sampling.

For controls, many firms use tables tied to how often the control operates. Typical ranges are one item for an annual control, two for quarterly, two to five for monthly, five to fifteen for weekly, twenty to forty for daily and twenty-five to sixty for controls performed many times a day. These sizes assume no expected deviations, so finding even one usually means the control cannot be relied upon at the planned level without extending the test.

• Simple random selection: every item has an equal chance; suits homogeneous populations
• Systematic selection: every nth item from a random start; beware populations with a pattern
• Monetary unit sampling: each currency unit is a sampling unit, so larger balances are more likely to be selected; efficient for overstatement
• Stratified sampling: split the population by size or risk and sample each stratum separately
• Haphazard selection: acceptable only for non-statistical samples and only without conscious bias
• Block selection: a run of consecutive items; rarely appropriate because one block says little about the rest

What drives sample size

Sample size moves with a handful of factors. Understanding them lets an engagement leader explain why this year's sample of receivables is 205 items, not last year's 150, instead of hiding behind a software output.

• Assessed risk of material misstatement: higher risk means more assurance needed from the sample, so a larger sample
• Assurance from other procedures: reliance on tested controls or substantive analytics reduces the sample
• Tolerable misstatement: usually performance materiality or lower; halving it roughly doubles the sample
• Expected misstatement: the more errors expected, the larger the sample needed to still conclude
• Population value and variability: stratification reduces the size needed for highly variable balances
• Population size: matters little once a population runs to thousands of items

Worked example: sizing and evaluating a monetary unit sample

Continue the example. Trade receivables have a book value of EUR 4,500,000 and tolerable misstatement is set at performance materiality, EUR 90,000. For 95% confidence with no expected misstatement, the Poisson reliability factor is 3.0. The sample size is 4,500,000 x 3.0 divided by 90,000, which is 150, and the sampling interval is 4,500,000 divided by 150, EUR 30,000. Every balance of EUR 30,000 or more is certain to be selected and is tested in full.

The team finds two errors. A balance of EUR 12,000 is supported only to EUR 9,000, a tainting of 25%, which projects to 25% of the EUR 30,000 interval, EUR 7,500. A balance of EUR 40,000, larger than the interval, is overstated by EUR 2,000; because it was certain to be selected, that error is added as found, not projected.

The upper misstatement limit is basic precision (3.0 x 30,000 = 90,000), plus the projected error times the incremental factor for the first error (4.75 minus 3.0 = 1.75, so 7,500 x 1.75 = 13,125), plus the EUR 2,000 actual error: EUR 105,125. That exceeds tolerable misstatement of EUR 90,000, so the auditor cannot conclude the balance is fairly stated. The options are to extend testing, ask management to investigate and correct the population, perform alternative procedures, or treat the excess as a misstatement under ISA 450.

The lesson is in the planning. A sample sized for zero errors has no room for any error. Had the team expected EUR 15,000 of misstatement and used the 95% expansion factor of 1.6 from standard audit sampling tables, the sample would be 4,500,000 x 3.0 divided by (90,000 minus 15,000 x 1.6), which is 13,500,000 divided by 66,000, rounded up to 205 items with an interval of about EUR 21,950.

Evaluating misstatements under ISA 450

ISA 450 classifies misstatements as factual (no doubt about them), judgemental (differences in estimates or policy choices the auditor considers unreasonable) and projected (the auditor's best estimate of misstatement in a population, extrapolated from a sample). All three go on the summary of unadjusted differences, along with the effect of uncorrected misstatements from prior periods that still affect this year's statements.

The aggregate is compared with overall materiality and with specific materiality for particular items, and the auditor asks management to correct everything above clearly trivial. Size is only half the test. A misstatement below materiality can still be material because of its nature: it may change a profit into a loss, cause a covenant breach, affect a regulatory ratio, mask a change in trend, or involve related parties or fraud. Uncorrected misstatements are listed in the written representations, and ISA 450 requires them to be communicated to those charged with governance. Our guide on how a financial statement audit works shows where this evaluation sits in completion.

Working with sample populations in Skyline Nexus ERP

Every sample starts with a complete population, and that is where Skyline Nexus ERP helps. The Audit Pack (Excel) in Fiscal Authority > Reports produces one workbook with separate sheets for Journal Lines, General Ledger, Sales, Purchases, Customer Dues (AR), Supplier Dues (AP) and the Trial Balance, so the auditor can foot each population, agree it to the trial balance and load it into a sampling tool without asking for a dozen separate reports. The year selector works on calendar years, so non-calendar year ends need the individual reports run by date range.

For receivables and payables work, the AR Aging and AP Aging reports show current, 1-30, 31-60, 61-90 and over 90 day buckets per contact, which supports stratifying a sample by age as well as by value. The Trial Balance offers an Opening / Movement / Closing view and an Excel export, which together give the lead schedule its starting point. See our guide on preparing for an external audit for how these extracts fit into the PBC list.