Audit Attribute Sampling

Description

Samples records to estimate the deviation rate of a control or quality attribute with a stated confidence and tolerable rate, then accepts or rejects.

Auditors cannot inspect every transaction, so they sample statistically. In attribute sampling, the auditor sets a tolerable deviation rate (e.g. 5%), an expected rate and a confidence level (e.g. 95%), derives the sample size from tables or the binomial distribution, selects randomly, inspects each item against the attribute, and computes an upper deviation limit. If the upper limit exceeds the tolerable rate, the control is not relied upon. Discovery sampling is designed to find at least one instance if the rate exceeds a threshold. These methods suit UX quality audits: accessibility compliance of content, accuracy of AI responses, policy compliance of moderation decisions, data quality of research repositories.

Attributes

Loss type
Unmigrated
Effort
Days

Steps

  1. Define the population, sampling unit and the attribute (what counts as a deviation).
  2. Set tolerable deviation rate, expected rate and confidence level.
  3. Compute sample size (e.g. about 59 items for zero expected deviations, 5% tolerable, 95% confidence).
  4. Select items randomly and inspect them with clear criteria, ideally with two reviewers on a subset.
  5. Compute the sample deviation rate and the upper deviation limit.
  6. Accept or reject the control or quality claim, and investigate the nature of deviations.

Sources

  • AICPA, Audit Guide: Audit Sampling (2019)
  • IAASB, ISA 530: Audit Sampling
  • Dodge & Romig, Sampling Inspection Tables (1944)

Also known as

  • Attribute sampling
  • Statistical audit sampling
  • Discovery sampling
  • Acceptance sampling

Categories

  • Research Operations
  • Research Design & Rigor