Audit Sampling

What Is Audit Sampling?

Examining every item in a large population is neither affordable nor necessary, so procedures are applied to a subset chosen so that what is found can be projected onto the whole. The essential condition is that every item had a chance of selection, because a subset assembled from whatever was convenient supports conclusions about those items and nothing more.

Two approaches are used and they differ in what they can claim. A statistical approach draws items at random and applies probability theory, which lets the chance of landing on the wrong answer be measured and disclosed. A judgemental approach rests instead on the auditor's own reading of where error is most probable, which often yields the most useful single findings and supports no statement about the population at all. Most inventory work combines the two, with a value-weighted structured sample carrying the conclusion and a judgemental layer directed at whatever looks unusual. Reporting the two together as a single coverage figure overstates the statistical basis of the result, which is why the method section separates them.

Audit Sampling Under Indian Law

Position: where sampling is used, the sample must be designed and selected so that every item in the population has a chance of selection and the results provide a reasonable basis for a conclusion about the whole.

• Misstatements found in the sample must be projected to the population, and the auditor must consider the risk that actual misstatement exceeds that projection.
• Source: Standard on Auditing 530, Audit Sampling, covering sample design, size and selection, and the projection of misstatements.
• Note: the standard governs sampling only. It does not apply where an auditor tests a population in full, nor to items selected judgementally because something about them invites attention. Findings from that second group stand on their own and cannot be projected onto anything.

What an Auditor Accepts as Evidence of Audit Sampling

The evidence is the record of how items were chosen, and it has to predate the results.

• Accepted: the population definition, stating what was in scope and what was excluded, since a conclusion reaches only as far as the population it was drawn from.
• Accepted: the selection basis written down before counting, covering the value threshold for complete examination, how the remainder was stratified and how items were drawn within bands.
• Accepted: the list of items actually selected, retained in full so the sample can be reperformed.
• Rejected: a rationale written up once the outcome is visible, since nothing separates it from a method picked to fit that outcome.
• Rejected: a coverage percentage quoted without saying whether it measures lines or value, because those two figures diverge sharply and one of them always reads better.

How Audit Sampling Works in Practice

1. The population is defined precisely, including which items are excluded and why, since a conclusion can only ever cover what was actually in scope.

1. Items above a value threshold are separated out and examined completely, because sampling within them saves little and risks a great deal.

1. The remainder is stratified into bands and a number of items is drawn from each, so no band goes entirely unexamined and the effort follows the value.

1. Selection within a band is made randomly or at systematic intervals, which removes the pull toward items that are accessible and well labelled and therefore least likely to be wrong.

1. A judgemental layer is added for anything that looks unusual, and it is recorded separately. The structured sample supports the conclusion about the balance; the judgemental items support findings about themselves and nothing more.

Audit Sampling: A Worked Example

A Nagpur warehouse holding 4,584 stock lines is tested with 167 of them.

The coverage that matters is by value, not by line count. Those 167 lines are under 4% of the population but reach roughly 72% of the value, because the top stratum is examined in full rather than sampled. The bottom stratum is sampled thinly and deliberately so: 3,900 lines carrying Rs 40 lakh cannot hold an error large enough to matter on its own, and testing them heavily would consume the time the high-value stratum needs. What the auditor cannot do is project the top stratum's result, since nothing there was left untested.

Common Mistakes With Audit Sampling

Most defects here concern what the sample is later claimed to support.

• Selecting whatever is accessible and projecting the result onto the population, when items that had no chance of selection cannot be represented by it.
• Reporting the structured and judgemental layers as a single coverage figure, which dresses a judgemental layer up as though it carried measurable confidence.
• Quoting only the proportion of lines examined, or only the proportion of value, when the two diverge sharply and each alone misleads.
• Deciding the basis after the results are known, which cannot be distinguished from choosing a basis that fits the answer.
• Assuming a larger population needs a proportionally larger sample, when the number required rises very slowly once the population is reasonably large. Sizing the sample from the line count alone therefore wastes effort on breadth that adds nothing.

Related Stock Audit Terms

Where the evidence falls short, the opinion says so in a defined way, so qualified opinion is the entry to read next. Every other term in this cluster is listed in the the glossary, grouped by the part of the work it belongs to.