An analytical procedure that compares the current year to the prior year and writes "fluctuation due to growth" is not an audit procedure. It is a description. A substantive analytical procedure has a defined structure: an expectation built independently of the recorded number, data that has been checked for reliability, a threshold for how large a difference to accept, and corroborated explanations for what exceeds it. This post walks through one from start to finish. See the survey of phases in analytical procedures planning, substantive and final review and the related flux variance analysis.
The five elements
The GAO Financial Audit Manual describes substantive analytical procedures as evaluations of financial information through analysis of plausible relationships among financial and nonfinancial data, including investigation of fluctuations that are inconsistent with other information or deviate significantly from predicted amounts. It builds the procedure around an expectation of the recorded amount, derived from an understanding of relationships. Five elements follow from that:
- Plausible relationship. Why should the data be related?
- Independent expectation. Built from data not produced by the account being tested.
- Reliable data. Tested or obtained from sources the auditor has comfort with.
- Threshold. The difference the auditor accepts without further work.
- Investigation and corroboration. Differences above the threshold are explained with evidence.
Worked example: interest expense
A manufacturer recorded interest expense of $412,000. The auditor tests it analytically.
Element 1 and 2: relationship and expectation
Interest should equal average debt times average rate. Data for the expectation comes from the debt schedule, tested in the debt program, and from lender confirmations of rates.
| Facility | Average balance | Rate | Expected interest |
|---|---|---|---|
| Term loan | 2,250,000 | 6.50% | 146,250 |
| Revolver (monthly averages) | 1,600,000 | 7.10% | 113,600 |
| Equipment note | 369,000 | 6.25% | 23,063 |
| Shareholder loan | 300,000 | 8.00% | 24,000 |
| Fees (commitment and unused line) | 18,500 | ||
| Expected total | 325,413 | ||
| Recorded | 412,000 | ||
| Difference | 86,587 |
Element 3: data reliability
The debt balances on the schedule were agreed to confirmations and the rollforward in the debt program. The schedule's completeness is a separate question, and this procedure tests it. The rates came from the agreements and confirmations. The revolver's monthly averages were computed from lender statements. The data does not come from the interest expense account. Document this reasoning, since the reliability of the inputs decides the weight of the procedure.
Element 4: threshold
The threshold (the GAO manual calls it the limit) is the difference the auditor will accept without further investigation. It is a matter of judgment, and the manual gives guidelines tied to performance materiality: 20 percent or less when the auditor relies completely on the analytical procedure, and 30 percent or less for partial reliance. If performance materiality is $65,000, a threshold of $65,000 would let a full performance-materiality misstatement through unexamined. Here the analytical procedure is the primary test of a predictable account, so the auditor uses complete reliance and sets the threshold at 20 percent of $65,000, which is $13,000.
The $86,587 difference exceeds the $13,000 threshold, so investigate. The manual asks for an explanation of the whole difference, and allows the auditor to stop once the explanation covers the part above the limit: $86,587 less $13,000 is $73,587.
Element 5: investigate and corroborate
Ask management what explains the difference, then obtain evidence. The auditor finds three causes:
| Cause | Amount | Evidence |
|---|---|---|
| Interest on a $1.1 million bridge loan the debt schedule omitted, December to year end plus earlier draw | 42,000 | Lender statement, loan agreement |
| Late payment penalties paid to the equipment lessor, recorded in interest | 14,500 | Invoices, lessor correspondence |
| Amortization of debt issuance costs recorded in interest | 22,000 | Amortization schedule, closing costs statement |
| Total explained | 78,500 | |
| Unexplained | 8,087 | Below the $13,000 threshold; no further work |
The bridge loan omission is the important finding: debt that did not appear on the schedule used to build the expectation. The auditor now reopens the completeness of debt. The procedure paid for itself because the expectation was independent of the recorded amount. A client's explanation without corroboration would have missed it.
Management's explanations were corroborated by independent evidence for each item. The GAO manual makes the point that the auditor should consider whether a difference is caused by error or fraud and should not accept an explanation unsupported by evidence.
Where analytics work and where they do not
| Works well | Works poorly |
|---|---|
| Interest expense from debt and rates | Accounts driven by judgment, such as impairment |
| Payroll from headcount and rates | Highly volatile items with no stable relationship |
| Rent from lease terms | One-off or unusual transactions |
| Depreciation from asset classes | Accounts affected by management discretion |
| Revenue in subscription businesses from customers and prices | Accounts where the underlying data is the same source as the recorded number |
| Commissions from sales and rates | Small, immaterial accounts where expected effort exceeds value |
For payroll and depreciation, see payroll audit procedures and depreciation recalculation.
Common errors
- Expectation based on the prior year's recorded number plus a percentage. That reuses a number that may itself be wrong.
- Using client-prepared data that comes from the same system as the account being tested.
- A threshold chosen after seeing the difference.
- Accepting the client's explanation as evidence.
- Treating the analytic as sufficient where the risk requires detail testing.
- Not documenting why the relationship is plausible.
Combining with other procedures
If the analytic is not precise enough to be the sole test, combine it with detail testing: use analytics to identify where to select items, or to reduce detail testing for the remainder. Document how the analytic and the detail tests together reach the required level of assurance. Analytics also serve at planning and final review, with different rigor. See flux variance analysis.
Documentation
The relationship and why it is plausible, the expectation and its sources, the reliability of the data, the threshold and how it was set, the comparison, the investigation of each difference above the threshold with corroborating evidence, and the conclusion. Cross-reference to any entries proposed.
CORAA prepares ready-to-review audit files for US CPA firms; the firm keeps the client, the review and the signature. See /us.
Sources
- AICPA, AU-C Section 520, Analytical Procedures; AU-C Section 330, Performing Audit Procedures in Response to Assessed Risks and Evaluating the Audit Evidence Obtained
- U.S. Government Accountability Office, Financial Audit Manual, Volume 1 (GAO-26-108577), June 2026, FAM 475 and 495 A (substantive analytical procedures, expectations, investigation of differences): gao.gov
Part of CORAA University · Evidence and audit procedures: related guides, calculators and templates.
AU-C references are to AICPA Professional Standards, paraphrased here; read the standard itself before relying on a requirement. This page is guidance, not professional advice. Report an error