There are many definitions of sampling out there, but in general sampling involves selecting individual items from a population, to enable you to draw a conclusion on the population as a whole.
In an audit, sampling procedures are used because it is not practical to examine every single item in a population. For example, the auditor may select an audit sample of non-current assets, and verify their existence, condition and value. It would not be practical for the auditor to track down every single asset on the books. But, if all the items in the audit sample are verified then it may be appropriate to draw the conclusion that all the assets are correctly recorded in the books (assuming the audit sample has been selected correctly and is of sufficient size).
Before you can begin a sampling procedure you must define the population that you wish to test. The population is all of the items that contain the characteristic that you wish to understand. Thankfully, in the case of audit sampling the population is usually easy to define, e.g. all the assets on the asset register, all the sales during the period under review, etc. However, there may be occasions where we need to sample over time or space, and the population may be more difficult to define, for example an investigation into call centre waiting times at different times of the day. In these cases the focus may be on discrete observations or periods.
The basis for selecting an audit sample is the list of items from which the audit sample is being chosen, which is called the sampling frame. The sampling frame may be identical to the population, but this is not always the case. For example, if an internal audit department were investigating customer satisfaction levels in a shop, it would be impossible to identify all the customers who have purchased from that shop. Therefore, an appropriate sampling frame may be all the customers on a specific date.
It is vital that the sampling frame is representative of the population. In the above example, the shop may be considerably busier at weekends than on a Monday. Therefore if the date chosen to sample customers is a Monday, when staff have more time to spend with each customer, the results of the testing may indicate that customers are very satisfied. However, in reality the majority of customers (who shop on weekends) may have a much lower satisfaction level.
Selecting an appropriate sampling frame that is representative of the population is a matter of auditor judgement.
The audit sample must also be representative of the population for the results to be valid, and this is often dependant on the sampling method chosen. For example, in the above scenario, if the auditor is left to select customers at random (haphazard sampling) they may be inclined to approach more attractive customers, who may also receive preferential treatment from staff based on their physical appearance. In this case a different method of selecting the audit sample would be more appropriate, for example approaching every 10th customer (systematic sampling).