audit sampling

If you want to know whether a huge pot of soup is seasoned right, you do not drink the whole pot — you stir it and taste one spoonful. Audit sampling is that idea applied to accounting: instead of examining every one of thousands of transactions, the auditor tests a carefully chosen subset and uses the result to draw a conclusion about the whole group. It is what makes auditing a giant company possible within a budget and a deadline.

Auditors sample because checking 100 percent of millions of transactions would be impossibly slow and rarely worth it. There are two broad styles. Statistical sampling uses probability and random selection so the auditor can express the result and its uncertainty in mathematical terms — for example, testing 60 of 5,000 invoices, finding 1 error, and projecting an estimated error rate across the whole population with a stated confidence level. Non-statistical sampling relies on the auditor's judgement to pick the items and interpret the findings. Either way, the sample must be representative, and an error found in the sample is usually projected onto the whole population: if 2 percent of the tested invoices were wrong, the auditor expects roughly 2 percent of all invoices to be wrong, and judges whether that is material.

Sampling matters because it is the practical engine behind 'reasonable assurance' — but it is also the built-in reason audits can miss things. This is sampling risk: the tested sample may, by bad luck, look fine while problems hide in the items not chosen, or look bad while the rest is fine. A spoonful can mislead you if the pot was not stirred. That is why auditors target risky areas, sometimes test 100 percent of unusually large or odd items, and never claim a sample-based audit proves the statements are flawless.