Audit sampling is choosing a subset of a population to test and drawing a conclusion about the whole population from the results — either statistically, where sample size and selection follow probability theory, or judgmentally, based on experience and risk. On a Tally-based client where every voucher in the population is available electronically, many procedures that firms still sample — matching, footing, three-way matching, recomputation — can be run on the full population instead, which removes the sampling risk entirely for that procedure. Sampling remains necessary where the test itself requires physical or third-party action, such as physical verification of inventory or external confirmations, because those cannot be run on every item regardless of how complete the electronic population is.
Statistical versus judgmental sampling
Both are valid under SA 530. The difference is in how the sample is selected and what can be claimed about the result, not in which one is more rigorous in principle.
| Statistical sampling | Judgmental sampling | |
|---|---|---|
| Selection | Random or systematic, so every item in the population has a known probability of selection | Based on the auditor's judgement of risk — larger balances, unusual items, related parties |
| What the result supports | A mathematically defensible projection of the misstatement rate across the population | A reasoned conclusion that depends on the quality of the judgement applied to select the sample |
| Sample size basis | Calculated from confidence level, tolerable misstatement, and expected misstatement | Set by experience and risk assessment, without a formula behind the number |
| Where it tends to work best | Large, homogeneous populations where a quantified projection is useful — trade receivables confirmations, inventory counts | Populations with a wide range of item sizes or risk, where a few high-value or unusual items matter more than a random cross-section |
| Common weakness in practice | Mechanically applying a formula without checking the population is actually homogeneous enough for the result to generalise | A sample that quietly converges on the same items tested every year, rather than being re-derived from the current year's risk assessment |
Regulatory requirement
SA 530 deals with the auditor's use of audit sampling when designing and selecting the audit sample, performing tests of controls and tests of details, and evaluating the results from the sample. Read the current text of the standard rather than relying on this summary.
When full-population testing removes the need to sample
Sampling exists because testing every item used to be too expensive relative to the assurance gained. That trade-off assumed the population was on paper, or that pulling every voucher meant manually opening every file. On a Tally-based client, the general ledger, voucher-level detail, and often the underlying documents are already in electronic form, which changes the economics for a specific category of procedure: one where the test is mechanical and repeatable rather than a matter of professional judgement applied item by item.
| Procedure | Why sampling becomes unnecessary |
|---|---|
| Matching purchase vouchers to goods receipt notes and purchase orders | A three-way match is a rule, not a judgement. Running it on all 4,000 vouchers costs no more analyst time than running it on 30, once the data is extracted. |
| Footing and casting schedules | Arithmetic has no sampling risk worth discussing once it can be recomputed on every row. |
| Identifying duplicate payments or vouchers | A duplicate that falls outside a sample is still a duplicate. Sampling this test specifically means accepting that some duplicates will never be looked for. |
| Reconciling subsidiary ledgers to the general ledger | A reconciliation is either complete or it isn't; there is no probabilistic version of "these two totals agree." |
| Checking TDS rates applied against the applicable rate for the nature of payment | A rule-based check against a rate table runs the same way whether applied to one voucher or all of them. |
How we think about this
Lekhya's position is that where a procedure can be run on the full population at no material extra cost, sampling that procedure doesn't make the audit more efficient — it makes it incomplete on purpose. Full-population testing is run wherever the underlying check is deterministic; sampling and judgement are reserved for what genuinely requires a person's attention, such as evaluating an unusual item the full-population run surfaces.
When sampling still applies
Full-population testing works for checks a computer can perform against existing electronic data. It does not extend to procedures where the evidence itself has to be generated through physical or third-party action — those remain sampling decisions regardless of how complete the client's books are.
- Physical verification of inventory: someone has to be physically present to count, and counting every location on every date is rarely practical, so locations and items are sampled based on value and risk.
- External confirmations from debtors, creditors, and banks: each confirmation requires a third party to respond, so the population of parties confirmed is a sample selected by balance size and risk, not the full ledger.
- Physical verification of fixed assets: large or dispersed asset registers are tested at a sample of locations or asset classes rather than every asset being physically inspected every year.
- Attendance at inventory counts and observation of physical controls: inherently a sample of the occasions and locations where counts happen, since the auditor cannot be everywhere at once.
- Tests of controls performed through inquiry and observation, such as an approval control evidenced by a signature: sampled, because the full population of instances cannot be meaningfully recomputed the way a numeric check can.
A useful question to ask before designing any test: is the thing being checked already sitting in the client's electronic records in a form that can be checked by rule, or does checking it require a person to go somewhere, ask someone, or physically count something? The first category is a candidate for full-population testing; the second is where a sampling decision genuinely has to be made.
What drives sample size
Where sampling is still the right approach, the size of the sample follows from a small number of factors, whether the sample is statistical or judgmental. Treating sample size as a fixed number — "we always test 25 items" — ignores all of them.
- 01Tolerable misstatementSet in relation to performance materiality. A lower tolerable misstatement requires a larger sample, because the test has to be more sensitive to catch a smaller amount of error.
- 02Expected misstatement in the populationA population with a history of errors or exceptions needs a larger sample than one with a clean track record, because more items are needed to form a reliable view of the error rate.
- 03Assessed risk of material misstatementHigher risk at the assertion level — from the planning-stage risk assessment — pushes toward a larger sample or a different testing strategy altogether, such as full-population testing where available.
- 04Population size and variabilityFor statistical sampling specifically, a more variable population (a wide spread of item values) generally needs a larger sample to achieve the same precision than a homogeneous one.
- 05Desired level of assuranceA higher required confidence level, appropriate for a significant risk area, increases the sample size needed for the same tolerable misstatement.
| Mistake | Effect |
|---|---|
| Using last year's sample size without revisiting this year's risk assessment | The sample no longer reflects current-year risk, and a change in risk goes unaddressed |
| Confusing sample size with sample coverage in value | A small number of high-value items can cover most of the population's value while leaving the bulk of transactions untested for other risks, such as unauthorised small payments |
| Stratifying by value but not by risk | A population split only by size can still miss a cluster of smaller but higher-risk items, such as related party transactions below the high-value threshold |
A sample size copied from last year's file, for a population that has grown, for a client whose risk profile has changed, is not a documented basis for the sample. If asked to justify the number, the file should show the reasoning for this year, not a reference to what was done previously.
This is practical guidance written for audit teams, not an authoritative statement of any standard or statute. Standards, forms and timelines change. Check the current text issued by the ICAI and the relevant authority before relying on anything here in an audit file.
Related resources
- Audit workflow & automationManaging audit evidence across hundreds of client documentsRequesting documents so they arrive usable, classifying what comes in, keeping versions straight, and always knowing what is still outstanding.
- Audit workflow & automationAutomating repetitive audit procedures without automating judgmentWhere the boundary between mechanical work and judgment falls, a test for deciding which side a task is on, and how to keep the boundary from eroding.
- Audit workflow & automationStatutory audit planning checklistUnderstanding the entity, setting materiality, risk assessment at planning, and briefing the team against a realistic timeline — a rounded planning reference.
- Audit workflow & automationAI in audit: what it is good at, and what it must not touchThe three tasks language models genuinely do well in audit, the three they must never be given, and the questions to ask a vendor before any client data moves.
Free tools, no signup
Each one reads a file entirely in your browser — or, for two of them, just answers a few questions — and nothing is ever uploaded.
- Purchase register checkDuplicate payments, round-sum amounts, weekend-dated invoices, incomplete records.
- GSTR-2B reconcilerMatches your purchase register against GSTR-2B — missing invoices either side, amount mismatches.
- Bank reconciliation checkerMatches your cash book against a bank statement — what's outstanding on each side.
- Fixed asset register checkDuplicate asset codes, missing location or useful life, negative net book value.
- Payroll red-flag scannerBank accounts shared across employees, and records missing an ID, PAN or salary.
- CARO applicability checkerAnswer a few questions about the company and see which CARO 2020 clauses likely apply.
- MSME payment delay checkerMSME-flagged invoices paid — or still unpaid — beyond the 45-day Section 22 limit.
- Related-party transaction screenerTransactions above a threshold you set with no approval reference, grouped by counterparty.
- Depreciation useful life checkerCompares a useful life in use against the Schedule II, Part C indicative figure.
- Materiality calculatorOverall and performance materiality against PBT, revenue, assets or equity, with reasoning shown.
- Sample size calculatorStatistical sample size for a test of controls or a substantive test of detail.