How to Use Data Analytics in Internal Audit

Analytics efforts often stall because they start with software rather than an audit question, or because poor data creates false confidence. This guide shows how to turn audit questions into useful, repeatable data tests.

Use it when planning analytics for an engagement. The six steps are: start with the audit question, not the software; define the population and required fields; validate data completeness, dates, duplicates and key joins; design tests for outliers, threshold behavior, duplicates, timing and unusual relationships; investigate exceptions with business context before concluding; and save logic and thresholds so successful tests can be repeated or monitored.

The worked example asks whether employees are splitting purchases to avoid approval limits. It uses requester, vendor, date, amount, PO and approval level, groups purchases by the same requester and vendor within short windows, flags combined amounts above approval thresholds, and then inspects flagged transactions and their business justification. The quick reference reminds you that analytics finds signals, not guilt, that full-population testing can improve coverage but poor data can mislead, and that logic should be documented so another auditor can reproduce the result. An auditor prompt, quality check and mini self-check help you confirm the work answers the audit question.