What Is Attribute Sampling?

Attribute sampling is a statistical process used in audit procedures that aims to analyze the characteristics of a given population. This practice is often used to test whether or not a company's internal controls are being correctly followed. Without the ability to rely on controls, it would be very difficult and exceedingly costly to conduct a meaningful audit.

How Attribute Sampling Works

Let's say an auditor wants to test the effectiveness of a company's rule that purchases of more than $10 must be authorized with a purchase order. Since examining every vendor invoice over $10 is often not feasible, the auditor will take a sample. The size of the sample must be large enough to provide an accurate picture of the entire population of purchase orders over $10, though that accuracy is always a matter of degree and must be tested. In examining the sample, the auditor may discover that 5% of the vendor invoices over $10 were not authorized by a purchase order.

On the other hand, 5% may be deemed acceptable. Because the auditor has taken a sample and was not able to examine the entire population of vendor invoices, she must do an additional analysis because any time a sample is taken, a phenomenon known as “sampling error” occurs.

A sampling error occurs when the values of the sample do not match those of the entire population from which the sample was taken. So if additional analysis shows that the margin of error is 2.5%, then a 5% non-compliance rate would be acceptable because the confidence interval is 5% plus or minus 2.5%, and the tolerable 3% rate falls within that range.

This 5% non-compliance rate may be acceptable or not, depending on the rate the auditor has determined to be a tolerable figure. If the auditor believes a 3% rate is tolerable, 5% would consequently appear to be too high and would indicate that the company's internal controls were not effective. In this scenario, additional investigation by the auditor would be necessary. This data might also suggest that additional controls are necessary in the future.

Consider an election poll, where sample data indicated that 49% of those surveyed say they plan to vote for Candidate A, and 51% of those surveyed say they plan to vote for Candidate B. In this case, the 2.5% sampling margin of error is greater than the 2% differential between the figures, which would throw the veracity of the poll results in question.

Types of Questions Asked in Attribute Sampling

Many items may be studied using attribute sampling. A partial list includes:

  • Does the shipment of goods always come before billing?
  • Are goods only purchased from approved vendors?
  • Are credit sales made to consumers with approved credit?
  • Are all checks signed by an authorized person?

[Important: Attribute sampling is only meaningful if used to audit internal controls that are correctly designed and efficiently executed.]