Choosing a sample size without inventing false precision
A number looks objective when it sits alone. Sampling for compliance testing stays honest only when the rationale travels with it.
Teams often borrow a familiar count — twenty-five, forty, sixty — because last year’s file used it. That habit is not always wrong, but it becomes indefensible when risk signals change and the file still shows the same integer with no explanation.
Separate method from theater
Statistical language impresses nobody if the population definition is fuzzy. Before debating formulas, write what you mean by “complete population,” how you will handle voids or reversals, and which period you cover. Only then does a size discussion mean something.
Document the expansion trigger early
Decide in advance what finding pattern would expand the sample. When the trigger is written after the first exception, it looks opportunistic. One short paragraph above the sample list is enough: the signal, the increment, and who approves the change.
Judgment is allowed — silence is not
Judgmental sampling is legitimate for many compliance tests. What fails review is the missing sentence that explains why these items represent the risk. Name the strata you cared about — new products, high-value approvals, manual overrides — even if you did not run a formal random draw.
Practice material
Module 03 of Korea Compliance Audit Mastery includes worksheet prompts for rationale paragraphs. Contact AlgorithmiTech if your team wants a private sampling clinic.