What is Discovery Sampling?

Discovery sampling involves the use of a sample to determine whether a percentage error does not exceed a designated percentage of the population. If the sample does not contain errors, then the actual error rate is assumed to be lower than the minimum unacceptable rate. The main variables used in the sampling calculation include the confidence level, minimum unacceptable error rate, and population size.

Example of Discovery Sampling

An auditor is evaluating a company's accounts payable process to identify unauthorized payments to vendors. The auditor suspects that a small number of fraudulent transactions might exist in a population of 10,000 transactions. The auditor follows these steps in the discovery sampling process:

Define the objective. The auditor wants to detect any unauthorized payments.
Set the tolerable exception rate. The auditor sets this rate to 0%, since any unauthorized transaction is unacceptable.
Determine the confidence level. The auditor chooses a 95% confidence level, meaning that she wants to be 95% sure that the sample will include at least one unauthorized transaction if such transactions exist in the population.
Select the sample size. Based on statistical tables or software, the auditor determines that a sample size of 300 transactions is required to meet these parameters.
Perform audit procedures. The auditor examines the 300 sampled transactions for evidence of authorization.
Evaluate the results. If no unauthorized transactions are found, the auditor concludes that it is unlikely (within the set confidence level) that unauthorized transactions exist in the population. If even one unauthorized transaction is found, the auditor investigates further, possibly expanding the sample or conducting additional audit procedures.

When to Use Discovery Sampling

Discovery sampling is applied when you expect the rate of occurrence for a particular data characteristic or attribute to be quite low. This sampling approach is intended to find out whether at least one occurrence of the characteristic or attribute in question exists in the underlying population. An auditor could employ discovery sampling to determine whether a particular transaction error or instance of fraud exists in the data set being examined.