Discovery sampling is a statistical auditing technique used when the occurrence rate of a specific characteristic or attribute is extremely low. The technique is designed to test whether at least one instance of that characteristic or attribute exists in the population. Auditors use this technique to identify whether or not a particular error, fraud, or irregularity (like a transaction error) exists in the data set.
The main idea behind discovery sampling is not to determine the full extent or frequency of a problem, but rather to discover if the problem exists at all. For example, an auditor might use discovery sampling to search for instances of fraud in a company’s financial records. Even a single instance found would be considered significant.
The sample size in discovery sampling is usually larger than in other forms of statistical sampling, due to the low occurrence rate of the attribute of interest. The auditor chooses the sample size based on the size of the population, the tolerable error rate, and the confidence level they wish to achieve.
While discovery sampling can be a useful tool for auditors, it’s important to note that like all sampling methods, it cannot guarantee with 100% certainty the absence of the characteristic or attribute being tested. The results are dependent on the quality of the sample, which must be truly random and representative of the overall population.
Example of Discovery Sampling
Suppose an auditor is hired to check for fraudulent activities in the company’s expense claims. The company has tens of thousands of expense claims per year, but fraudulent claims are believed to be very rare. The auditor’s goal is not to estimate how many fraudulent claims there are, but simply to find out if any exist.
To do this, the auditor decides to use discovery sampling. They determine the sample size based on the size of the total population (the number of expense claims), the confidence level they want to achieve, and the maximum error rate they are willing to tolerate.
The auditor then randomly selects expense claims to form their sample and carefully checks each one. If they find even a single fraudulent claim in the sample, they will have achieved their goal: they’ve discovered that fraudulent claims do indeed exist. The auditor can then report this finding and recommend further investigation or actions to address the issue.
This is a simplified example, but it should give you an idea of how discovery sampling can be used in practice. It’s a powerful tool when the occurrence rate of a specific attribute is expected to be very low, but finding even a single instance is important.