Substantive analytical procedures

Decide what the comparison is meant to establish

Substantive analytical procedures seek evidence about an assertion by comparing a recorded amount with a defensible expectation. The reasoning starts with a relationship: if the underlying activity and applicable rates are known, a particular financial result should follow. A comparison becomes useful evidence only when that relationship, the inputs and the acceptable difference are sufficiently strong for the audit purpose.

Analytical procedures have different jobs during an audit. Planning comparisons help identify risks. Substantive analysis responds to an assessed risk. Completion analysis assesses whether the financial statements remain consistent with the auditor's understanding. A spreadsheet used at planning does not automatically satisfy the requirements for a substantive procedure. Completion analysis is required even when detailed work has already been performed; it need not itself be a substantive analytical procedure.

Concept sources: ASA 520, December 2021, paragraphs 1, 5–7 — substantive use and the separate overall conclusion requirement.

Match the method to a predictable relationship

Large, recurring transaction streams often support prediction better than unique negotiations or discretionary spending. A stable contractual service charge can be modelled from service periods and agreed rates. A one-off restructuring provision needs evidence about its particular obligations and assumptions; a prior-year percentage is unlikely to capture those conditions.

Concept sources: ASA 520, A6–A10 and A15 — original comparison of procedure design and its limitations.

A broad trend may identify a question; a precise independent model may answer one. Assurance depends on design and execution, rather than the sophistication of the software or the number of ratios calculated. A detailed model using weak inputs can be less persuasive than a simple calculation using well-supported contractual facts.

Build the expectation before seeing the answer

Start with the assertion and a possible misstatement. Decide what amount should result if that assertion is correct. Develop the expectation and investigation threshold before comparing it with the recorded figure. Otherwise, knowing the result can encourage the auditor to choose assumptions that explain it away.

Useful drivers may be financial, such as verified loan balances and interest rates, or non-financial, such as completed service visits or operating hours. These drivers must describe the same activity and period as the financial amount. Cash collected, invoices raised and services delivered can each produce different totals for legitimate reasons.

Concept sources: ASA 520, paragraph 5 — original learning sequence for designing substantive analysis.

Establish whether the inputs deserve confidence

Ask where each input comes from, how it was prepared and whether it is complete and accurate for this purpose. A management forecast may be an aspirational target. Industry figures may describe businesses with different customer profiles. Even independent information requires evaluation for relevance and reliability.

Non-financial records are not inherently independent of accounting records. An operational report generated from the same incomplete transaction feed as the ledger may reproduce the same omissions. Understand the common source before treating agreement as corroboration.

Concept sources: ASA 500, December 2022, paragraphs 7 and 9 — original questions applying information reliability to analytical inputs.

Make precision visible in a worked calculation

The auditor set a $15,000 investigation threshold before comparison. The difference therefore needs investigation. Management identifies additional collections worth $18,000. Inspecting service logs and contract terms may support those additional charges, but an explanation alone does not. If supported, the revised expectation becomes $751,200 and the residual difference is $10,800.

That residual does not prove correctness. Consider whether the model remains sufficiently precise, whether other evidence conflicts and whether smaller differences could combine with other misstatements. If customer numbers actually changed each month, the half-year shortcut would need refinement. It cannot be retained simply because its result seems convenient.

Set a threshold that reflects the evidence needed

The acceptable difference is the maximum unexplained variance that the auditor plans to accept without further investigation under that procedure. Set it with reference to materiality, aggregation risk, the assessed risk and the desired persuasiveness of evidence. Higher risk normally calls for a smaller acceptable difference or a stronger alternative procedure.

Performance materiality informs the decision but is not an automatic pass mark for every analytical test. A difference below performance materiality may exceed the procedure's planned threshold. It may also reveal fraud, inconsistent accounting or an unreliable model. Consider both amount and nature.

Disaggregation often improves precision. A favourable variance in one operation can cancel an adverse variance elsewhere. Analyse coherent groups before combining conclusions. Regression and other statistical models can incorporate several drivers, but demand competent use, sensible assumptions and an understanding of prediction uncertainty.

Investigate the cause and revise the audit response

Recheck calculations and data definitions, then seek management's explanation and evidence supporting it. Separate a defect in the auditor's model from a genuine accounting difference. Unrecorded service credits, for example, might require an adjustment; an independently verified mid-year price reduction might explain the result without one.

Unresolved differences can require tests of details, additional information or a revised risk assessment. Do not repeatedly widen the threshold until the recorded result passes. Document why an explanation is supported and how remaining uncertainty affects the conclusion.

An effective analytical procedure can reduce other detailed work on the same assertion. It cannot justify omitting required procedures or leaving another assertion unaddressed. If low assessed risk depends on control reliance, support that reliance with control testing. For a significant risk, perform specifically responsive substantive procedures; if the response consists only of substantive procedures, include tests of details.

Concept sources: ASA 330, December 2021, paragraphs 8, 18 and 21 — control reliance, substantive coverage and the significant-risk qualification.

Sources and further reading

• ASA 520, December 2021 — expectation design, investigation and completion analysis.
• ASA 500, December 2022 — information used as audit evidence.
• ASA 330, December 2021 — how substantive procedures respond to risk.