Most of the transactions in a ledger arrive through systems: sales invoices from the billing system, supplier invoices through purchasing, payroll from the payroll system. Those systems apply their own checks — valid customers, approved purchase orders, authorised pay rates. Manual journals arrive differently. Someone types them, and the only checks they pass are the ones the finance function has designed around them.
That is why manual journals, although a small share of the transactions in a ledger, account for a large share of its errors, and why they are a standard focus of audit and fraud risk assessment.
Classify before you review
Not every journal carries the same risk. A recurring depreciation journal prepared from the fixed asset register is low risk; a one-off adjustment to revenue at quarter-end is not. A proportionate process classifies journals and matches the review to the risk:
- Recurring journals from approved schedules: reviewed for accuracy against the schedule.
- Standard adjustments such as accruals and reclassifications: reviewed with their supporting calculation.
- Non-standard and top-side journals, particularly those posted late in the close or directly to the consolidation: reviewed individually by a senior member of the finance team.
Approval that means something
Every manual journal above an agreed threshold should be approved by someone other than the person who prepared it before it posts, with the supporting evidence attached. The approver should be able to answer three questions: why is this entry needed, is the amount supported, and is the account coding right?
Where the accounting system allows it, approval should be enforced by the system rather than by policy alone. A rule that says journals must be approved, in a system that lets anyone post them, is a rule that relies on nobody being in a hurry.
Look for the unusual
Beyond approving journals one by one, a periodic review of the journal population as a whole can surface entries that individual approval would miss. Characteristics that warrant a closer look include:
- entries posted outside normal working hours or after the close deadline;
- round-number amounts, or amounts just below an approval threshold;
- entries to accounts that are rarely used, or combinations of accounts that do not normally go together;
- entries posted by people who do not normally post journals;
- descriptions that are blank, vague or identical across unrelated entries.
None of these proves anything is wrong. Each is a reason to ask.
Close the loop
A journal review is only useful if it changes something. Errors found should be corrected and their cause recorded, and patterns — the same kind of error recurring, a process that routinely needs manual correction — should lead to a fix in the process that produces them. A finance function that posts fewer manual journals over time, because it has fixed the reasons they were needed, has a stronger ledger than one that simply reviews more of them.