We start with the common mistake: treating the invoice as the centre of the review. It is not. An invoice records a billing event, while the accounting conclusion depends on what the organisation promised, when that obligation was satisfied, and whether the amount was measured appropriately. We compare the written contract with operational reality, including delivery conditions, customer acceptance, cancellation rights, amendments, service periods, and continuing support. This comparison matters because commercial language often moves faster than policy wording. A transaction may look complete to one team while another obligation remains outstanding. We document the evidence available at the reporting date and separate confirmed facts from unresolved judgment. That approach is slower than accepting the billing record, but it gives future reviewers a defensible path through the decision and reduces dependence on informal explanations.

Cut-off testing is often reduced to invoices issued near the reporting date. That shortcut misses the evidence that explains whether the transaction crossed the boundary correctly. We trace selected items through order records, dispatch or completion documents, customer acceptance, service logs, billing, ledger posting, credit notes, and later amendments. Each source answers a different question, so one matching date is not enough. We pay attention to unusual timing, manual entries, large contract changes, and patterns that differ from normal processing. None of those features proves an error, but each needs a clear explanation connected to the contract. We also examine activity immediately before and after close, then compare the findings with earlier adjustments. Repeated small exceptions can indicate a process weakness that deserves more attention than a single isolated correction.

Revenue estimates require a different kind of discipline because uncertainty cannot simply be removed from the file. We examine returns, rebates, concessions, milestones, and other variable amounts by tracing assumptions to source data, approvals, calculations, and later developments. Later outcomes are useful evidence, but they do not automatically prove that the earlier estimate was unreasonable. The relevant question is whether the original judgment reflected the information available at the reporting date and whether the same approach was applied to comparable arrangements. We record who approved the assumption, when it should be revisited, and what evidence would change the conclusion. Unsupported precision is not stronger analysis; it is uncertainty disguised as a precise number. A clear record allows management to challenge the judgment without losing sight of the facts that supported it at the time.

A written policy can be sensible and still fail in operation. We test the handoffs connecting contract review, delivery evidence, billing, amendments, credit notes, and period-end close. The practical question is whether the assigned person performed the review, retained evidence, and resolved exceptions before reporting. Manual overrides receive particular attention because they can bypass a sound workflow without producing an obvious system warning. We compare system fields with policy requirements and identify where information is re-entered, delayed, or left to informal messages. If a compensating review exists, we document its owner, frequency, evidence, and limits. This is where a three-to-five-year view becomes useful: a process dependent on one experienced employee remains fragile, even when current files appear orderly. Durable controls make decision points explicit and transferable.