When making sample selections from a population in Excel, auditors may be inclined to use the “=RAND()” formula without understanding the broader implications of this approach. For instance, what if you are testing approvals on purchase orders. Under this example, you want to validate that the approver(s) have an appropriate authorization. The “=RAND()” approach will not add much value to your testing if the samples amounts are well below the authorization thresholds.
Instead, consider another way to make your sample selections by tailoring your approach to better fit to the nature of the control. Using the purchase orders as our example, consider using a filter on the PO header amounts and look for items that are higher than the minimum authorization thresholds. In fact, look for significant outliers above several authorization thresholds to check if one or more appropriate approvals were obtained. Once you have an initial set of POs amounts, judgmentally select a number from the desired time periods.
Another example involves journal entries. While the “=RAND()” may give an auditor some statistical comfort, it’s entirely agnostic to keywords in the JE descriptions that might raise potential red flags, i.e., error, correction, etc. When dealing with JEs, the auditor may want to consider performing a review on keywords along with other data points, i.e., amount, account, etc., to derive an initial list that can then be reduced judgmentally to a set of sample selections.
The better you understand the nature of the controls, the better you can design your sampling methodology.