Journal entries: test overrides beyond round numbers
A round amount posted at year-end can be legitimate. An irregular amount generated automatically can still require investigation. Journal-entry testing becomes useful when the selection follows how management could override the reporting process, rather than relying on a single numerical pattern.
This guide concerns a PCAOB issuer audit and AS 2401 as displayed on October 25, 2026. The ledger and eight adjustments below are fictional. A public investor is not assumed to possess private journal exports, user logs or consolidation workbooks.
Define a population wider than the ledger
AS 2401 .58–.62 addresses journal entries and other adjustments, including changes made in preparing statements that are not formal ledger entries. It considers both period-end activity and whether work throughout the year is needed.
A ledger-only export can omit consolidation adjustments, report combinations, reclassifications or direct edits to financial-statement drafts. Define which systems and reporting layers can change the final numbers. Record their owners and reconcile each layer to the published statements.
The population file should retain entry identifier, accounts, debit and credit amounts, effective date, posting timestamp, user, approval trail and description. Add related reversal or replacement identifiers. A posting date and an accounting date can differ; one must not silently replace the other.
Verify that the extraction includes the intended entities and periods, and understand excluded, deleted, rejected or post-closing records as applicable. A balanced debit-credit total is useful but does not prove that every relevant adjustment was included.
Work an eight-item selection
All amounts are hypothetical thousands, measured once as a stated one-sided entry or adjustment amount. They are not summed debit-plus-credit values and are not automatically profit effects.
| ID | Item | Amount | Illustrative selection reason |
|---|---|---|---|
| J1 | Routine depreciation | 5 | Round and year-end, but normal recurring process |
| J2 | Automated sales posting | 9.43 | Non-round; not selected by the stated flags |
| J3 | Payroll posting | 7 | Routine workflow |
| J4 | Acquisition adjustment | 12 | Non-routine accounts |
| J5 | Prepayment amortization | 2.70 | Routine workflow |
| J6 | Manual revenue adjustment | 4 | Unusual privileged user and later reversal |
| J7 | Tax calculation posting | 6.20 | Separate relevant testing still possible |
| J8 | Direct draft-statement adjustment | 2 | Outside formal ledger, weak explanation |
Assume the mechanism-based selection identifies J4, J6 and J8: 12 + 4 + 2 = 18. This is a targeted set, not a representative random sample of all journal entries.
Suppose complete acquisition records, appropriate accounting analysis and authorization support J4 of 12. J6 and J8, totaling 6, remain unexplained after the stated initial work. The selected status bridge is supported 12 plus unresolved 6 equals selected 18.
Do not call 18 fraudulent, or 6 a proved misstatement. A selection flag identifies a question. Supporting a legitimate acquisition also does not establish that every other acquisition entry was correct. The conclusion stays with the items and assertions tested.
Investigate five linked dimensions
For each selected item, inspect the account relationship, user authority, approval, business basis and subsequent behavior. A formally authorized user may still make an inappropriate adjustment. Technical permission to post is not evidence that the accounting reflects the actual transaction.
Approval needs its own context. Was the reviewer independent of preparation, did the review precede final reporting, and what evidence shows the review addressed the issue? A generic approval stamp without the underlying analysis may not explain the entry.
Business support must match the amount, entity and date. For J4, a press release announcing a purchase would be weaker than the detailed support assumed in the example. Confirm that the record explains the specific adjustment rather than merely proving some transaction occurred.
For J6, follow the reversal to its counterpart. An early next-period reversal can be a normal accrual process, a correction or another mechanism. Compare it with the original support and whether the underlying event actually occurred. The reversal alone does not establish fraud.
For J8, identify how the draft edit entered the final report and whether it was reconciled back to the relevant reporting records. A spreadsheet adjustment without formal account numbers still changes reporting and cannot be ignored because the ledger contains no matching journal ID.
A single round-number rule fails in both directions
Assume J1 is supported by the depreciation schedule and the stated recurring controls. Its round 5 and year-end timestamp are not a reason to classify it as false. A filter based only on those features would produce an avoidable false positive.
J2 of 9.43 illustrates the other side. Its non-round amount and automated origin do not prove it is safe. A faulty configuration, duplicate import or unauthorized change to the underlying process could affect automatic entries.
AS 2401 .61 describes multiple characteristics and the reporting process, rather than a rule that round numbers alone identify inappropriate entries. Selection criteria should reflect the entity’s actual risks and evidence.
An override may occur inside valid system permissions. It need not involve an unauthorized login or a visible technical security breach. Conversely, an access anomaly does not itself establish the amount of a financial-statement error.
Reconcile unresolved items as evidence develops
Suppose later evidence supports J6 of 4 but J8 of 2 remains unresolved. The selected bridge becomes supported 16 and unresolved 2, still totaling 18. Preserve the original flag and the evidence that changed its status.
If J8 is subsequently established to be an incorrect revenue increase, determine the affected accounts, period and correction under the actual facts. Do not assume every one-sided amount directly equals a profit overstatement. A reclassification or balance-sheet adjustment can have a different effect.
Do not extrapolate unresolved 6/18 to the whole ledger. Targeted selection is designed to investigate risk mechanisms, not to estimate a population error rate. A monetary subtotal of flagged items is also not the amount of all management overrides.
Keep investigative status separate from posting status. “Posted,” “reversed” and “approved” describe workflow events; “supported,” “contradicted” and “unresolved” describe evidence. Combining them into a single pass/fail column hides important differences.
Execute the entry-to-report checklist
| Check | Evidence | Unresolved result |
|---|---|---|
| Population | Ledger, consolidation and report-adjustment bridge | Completeness not established |
| Identity | Entry ID, entity, dates and accounts | Item cannot be traced consistently |
| User | Role, permission and actual initiator | Authority or unusual involvement unexplained |
| Review | Reviewer, timing and analysis | Approval does not establish substance |
| Business basis | Contract, calculation and appropriate accounting | Flag remains unresolved |
| Reversal | Linked later entry and underlying reason | Timing pattern not yet interpreted |
| Final effect | Statement mapping and correction record | Amount cannot be called a profit error |
Public sources may disclose a control weakness, a restatement, an investigation or specific management conduct. Those disclosures can guide questions, but usually do not provide the complete private population. Do not invent entry-level findings for an issuer from a generic control-risk disclosure.
When logs and documents disagree, preserve both and examine account mapping, batch IDs, timestamps, user attribution and whether a later file overwrote an earlier adjustment. A fresh export is not automatically the original period-close population.
Update the file when support arrives, user involvement is clarified, a report-only adjustment is identified or a confirmed error changes the risk assessment. Retain the version of the selection rules used at each stage, so later evidence cannot silently turn a targeted investigation into an apparently clean retrospective filter.
Preserve the route through a reused user account
Suppose J6 carries a service-account name also used for ordinary imports. That label alone does not establish that the entry was automatic. Follow the batch source, configuration change, job log and initiating person. If those links are missing, record the attribution gap rather than assigning the transaction to an employee by guesswork.
A useful reconciliation has opening trial balance, ordinary postings, consolidation changes and report-only adjustments as separate layers. Map J8 into the layer that actually changed the final statements. Otherwise a later ledger export can appear clean because the disputed change never entered that export. Preserve the original close package before comparing the amended version; two individually balanced versions can still contain a consequential difference.
FAQ
Is every round year-end entry suspicious enough to be an error?
No. The pattern is a possible selection cue, not a conclusion.
Are automated entries outside override testing?
No. Their process and associated adjustments can still require investigation.