An auditor can scan every recorded invoice and still miss sales omitted from the accounting records entirely. The starting population, matching direction, and follow-up determine which assertion the resulting information can support.
Audit data analytics uses computational methods to organize and examine audit-related data for relationships, patterns, and transaction-level findings. The auditor specifies the evidence needed before selecting a tool or designing an attractive dashboard.
For nonissuer audits, the auditing standards for nonissuers govern the evidence evaluation. Issuer audits follow Public Company Accounting Oversight Board (PCAOB) standards; automation does not create a separate assurance level.
The same software can produce reports for several audit purposes, each requiring a different design and conclusion. A procedure intended to serve multiple purposes must satisfy each objective independently.
A full-population recalculation or document match need not follow the expectation-and-variance model governing substantive analytical procedures. Understanding a control is also distinct from testing its operation; control testing is not mandatory in every financial statement audit.
Common mistakes
- Match-only output. Reporting only joined records suppresses R-481 and R-485, which require investigation precisely because no warehouse match exists.
- Automatic error classification. Treating R-481's $28,000 as overstated ignores the direct-fulfillment evidence available in Example 1.
- Date substitution. Replacing R-483's January 3 dispatch date with its December 31 invoice date conceals the cutoff exception.
Bottom line
- The audit purpose, assertion, population, and procedure determine what an analytic output can support.
- Recorded invoices matched to shipments primarily address occurrence; shipments matched to recorded invoices address completeness.
- Reliable inputs, appropriate transformations, preserved unmatched records, and correct parameters are necessary for useful outputs.
- Descriptive, diagnostic, predictive, and prescriptive labels do not determine evidential strength.
Exam shortcut
Read the requested assertion before the software description. Identify the first file searched, then decide which missing records could appear in its output. For a flagged invoice, follow the identifier through the contract, transfer evidence, amount, and reporting period. Reject an adjustment supported only by a missing match. When two cases differ, isolate the changed evidence. Preserve unchanged amounts and findings rather than rebuilding the entire conclusion.
The full lesson (about 4,563 words, 30 min read) adds 2 worked examples, all 6 common mistakes, a self-check, free in the app.
Learning objectives
- III.A3
Browse all free CPA AUD lessons or jump into free CPA AUD practice questions.