Every unit has a chance
Sample items are selected so that every sampling unit in the population has a chance of selection.
Standards on Auditing on ICAIStock audit glossary · Records and evidence
noun · the subject of Standard on Auditing 530
Audit sampling means checking only part of a population, such as some of the lines in a stock list, and using the result to reach a conclusion about the whole. For that conclusion to hold, every item must have had a chance of being picked. Sampling can be statistical, with random selection so the risk of a wrong conclusion can be measured, or based on the auditor’s judgement of where errors are likely.
Meaning
Checking every line in a large stock is slow, costly and often unnecessary. So the auditor checks a selected part and projects what was found onto the whole. This works only if the items were chosen fairly: a sample made of whatever was easiest to reach tells you about those items and nothing more.
There are two broad approaches. In statistical sampling, items are picked at random and probability is used, so the chance of a wrong conclusion can be measured. In judgemental selection, the auditor picks the items where errors seem most likely; that often finds useful problems, but the result cannot be projected to the rest of the stock. Most stock audits use both, and keep the two results apart.
In the file
The evidence for a sample is the record of how it was chosen.
In practice
Including any lines left out and the reason, since the conclusion covers only what was in scope.
Sampling them saves little and risks a lot.
Pick some lines from every band, so no band is left unchecked.
This avoids choosing only the lines that are easy to reach and neatly labelled.
A judgemental check of anything odd describes only those items; it is never projected.
Worked example
Sample bands as an example; the thresholds are not a rule.
| Band | Lines tested | How picked |
|---|---|---|
| Lines above ₹5,00,000 each | 74 of 74 | Every line checked |
| Lines from ₹50,000 to ₹5,00,000 | 68 of 610 | Weighted by value |
| Lines below ₹50,000 | 25 of 3,900 | Random, for coverage |
| Odd items seen on the walk | Recorded apart | Not projected |
Only a small share of the lines is tested, but most of the value is covered, because every line in the top band is checked. The bottom band has thousands of lines and little value, so it is sampled lightly on purpose.
The result of the top band is not projected at all: nothing in it was left untested. Coverage should always be stated both ways, by value and by number of lines, so a reader can see what the sample really reached.
Framework
The rule comes from Standard on Auditing (SA) 530, Audit Sampling, issued by the Institute of Chartered Accountants of India. Where an auditor samples, the items must be selected so that every sampling unit has a chance of selection, and misstatements found in the sample must be projected to the whole population. So errors found in a sample are never treated as the only errors; they are used to estimate the likely error across all the stock.
Sample items are selected so that every sampling unit in the population has a chance of selection.
Standards on Auditing on ICAIMisstatements found in the sample are projected to the population, not reported as the total error.
Standards on Auditing on ICAILines tested in full and items picked because they looked unusual sit outside the sample. SA 530 applies only when the auditor has decided to use audit sampling; SA 500 treats selecting all items and selecting specific items as separate ways of choosing what to test.
Watch out
Most defects concern what the sample is later claimed to show.
Items that had no chance of selection cannot be represented by the result.
Judgemental picks are dressed up as if they carried measurable confidence.
The two shares can be very different, and each alone misleads.
It cannot be told apart from a method picked to fit the answer.
Questions
Statistical sampling picks items at random and uses probability to judge the result. Because every item had a known chance of selection, the auditor can measure the risk that the conclusion about the whole population is wrong. A judgemental selection cannot do this, so its findings describe only the items that were picked.
For a stock audit the usual sequence is: define the population, check the highest-value lines in full, split the rest into value bands, select within each band at random or at fixed intervals, and record any judgemental picks separately. The selection basis is written down before counting starts.
No single method suits every stock. A common approach is to check every high-value line, split the remainder into value bands and select at random within each band. A separate check of unusual items can be added, but its results are not projected to the whole stock.
Stock samples are usually drawn in three ways: random selection, where every line has an equal chance; systematic selection, where lines are taken at a fixed interval from a random start; and judgemental selection, where the auditor picks lines that look risky. Only the first two support a conclusion about the whole population.
Patron Accounting LLP. Statements on SA 530 on this page signed off by CA Sundram Gupta: 23 October 2026
Sources: Standard on Auditing 530, Audit Sampling (ICAI). General information, not professional advice; check the current position before acting.
Need help with audit sampling? See our stock audit service. A stock audit sets the sample before anyone counts.