Journal entry testing in audit: criteria, a worked example and documentation
Journal entry testing is required on every audit: ISA 240 (AU-C 240 in the US) makes the auditor test whether journal entries and closing adjustments are appropriate, because the risk of management override of controls is present in every entity. It takes three steps: get the complete population of entries, filter it with risk criteria (manual entries, weekends, round amounts, period end, amounts just below an approval limit…), and examine the selected entries against their support.
Updated 10 November 2026
Why auditing standards require journal entry testing
Management is in a unique position to manipulate the accounts by overriding controls that otherwise work well. ISA 240 starts from the premise that this risk of management override exists in every entity, treats it as a significant risk, and requires a response whatever the auditor's own assessment. One of the required responses is to test the appropriateness of journal entries recorded in the general ledger and of adjustments made in preparing the financial statements. The other two are reviewing accounting estimates for bias and understanding the business rationale of significant unusual transactions.
For journal entries, the standard asks you, in short, to:
- ask the people involved in financial reporting about inappropriate or unusual activity in entries and adjustments;
- select entries and adjustments made at the end of the reporting period;
- consider whether entries from throughout the period also need testing.
Before that, you need to understand the controls over journal entries (who can post, which entries need approval, how manual and consolidation adjustments are recorded), because that decides which entries carry risk. In the US the AICPA's requirements are in AU-C 240, which carries forward what SAS 99 introduced, and the PCAOB's in AS 2401 for issuers. In Spain the standard is NIA-ES 240. The IAASB issued ISA 240 (Revised) in 2025, effective for periods beginning on or after 15 December 2026, which strengthens the requirements and guidance on testing journal entries.
The population and the completeness check
A perfect selection from an incomplete listing is worthless: the entry that matters may be the one that is missing. Before you filter anything:
- Get the full journal from the system with every field: entry number, effective date, posting date, user, accounts, description, amount and source (manual or automated).
- Check that it balances: total debits equal total credits, and for each account, opening balance plus journal movements equals the closing balance in the trial balance.
- Check coverage: first and last dates, adjustment periods or "period 13", gaps in the numbering, and entries posted after the period end.
- Add what is not in the journal: consolidation and top-side adjustments recorded outside the ledger, often in spreadsheets.
Document the reconciliation to the ledger. Without it, a reviewer cannot tell whether the population was complete.
Criteria for journal entry testing
The criteria come from your fraud risk assessment: which accounts management might want to move, who could do it, and how. The usual ones, with what they find, their typical false positives, and how many lines each flagged in the example below:
| Criterion | What it finds | Typical false positives | Example |
|---|---|---|---|
| Weekend | Entries posted when nobody reviews | Month-end closes on Saturdays, overnight batch jobs | 35 |
| Public holidays | Same as weekends | 31 December, a working day in many companies | 57 |
| Round amounts | Estimates and made-up figures | Loans, capital injections, flat fees, bank-to-bank transfers | 59 |
| Period end | Cut-off manipulation, accruals and provisions | Every routine monthly accrual, depreciation, payroll | 1,293 |
| Keywords | "Adjustment", "correction", "reclass", "per CFO" | Standard descriptions on recurring entries | 80 |
| Just below an approval limit | Invoices split to avoid a second signature | Items whose normal price falls in that band | 264 |
| Unusual users | Entries by executives, IT staff or generic users | Migration or interface users | 1 user |
| Duplicates | Invoices booked or paid twice | Equal instalments posted on the same day | 12 |
| Unusual account combinations | Revenue against accounts other than receivables or cash, suspense accounts | Legitimate year-end reclassifications | — |
| Post-closing entries | Entries posted after the closing date but dated in the period | Audit adjustments and routine closing | — |
| Very large amounts | Whatever moves the accounts materially | Treasury and financing transactions | — |
"Example": lines flagged by journal/analysis.py on the demo journal with AuditSampler Pro's default settings. The last three rows are not automatic criteria in the app: apply them as filters on the export.
Journal entry testing example
A fictional distributor's general journal for 2025: 6,000 lines totalling 94,119,889.84. The demo file has cases planted on purpose (215 lines) and an answer key, so you can measure what each criterion finds. Settings: round amounts as multiples of 1,000, an approval limit of 10,000 with a 10% margin (9,000 to 9,999.99), 13 keywords, and 1 January and 25, 26 and 31 December as holidays.
Each criterion adds points: weekend 2, holiday 3, round amount 1, month end 1 (plus 1 at quarter or year end), 2 per keyword, just below the limit 2, duplicate 3. The result:
| Score | Lines |
|---|---|
| 0 (no criterion met) | 4,367 |
| 1 to 2 | 1,475 |
| 3 to 4 | 92 |
| 5 to 6 | 41 |
| 7 or more | 25 |
| Total | 6,000 |
- 1,633 lines (27.2%) meet at least one criterion. No audit tests 1,633 entries: a single criterion is not enough to select on.
- The 25 lines scoring 7 or more are almost all real cases: 24 of the 25 are planted. The top ones are manual adjustments on 30 and 31 December by two users; the highest, at 14 points, is a "manual adjustment" for a bonus accrual of 200,000 on 31 December.
- Period end is the noisiest criterion: it flags 1,293 lines (the first three and last three days of every month) and only 40 are planted. Use it in combination, not on its own.
- The approval-limit criterion finds all 45 planted cases, but among 264 flagged lines. The next step is grouping by user: one person has 50 lines between 9,500 and 9,999.99. Benford's law reaches the same group another way, as an excess of amounts starting with 9.
- Holidays flag 57 lines, but 51 are on 31 December. If that is a working day for the client, drop it from the list or the criterion just duplicates period end.
- The generic user
sysadminhas 7 lines, 71.4% at weekends and 100% round amounts: test all seven. - Two cases are invisible to every criterion: lines of 1,318,455.90 and 1,642,880.75 on ordinary days with ordinary descriptions, score 0. That is why large amounts need their own filter.
Keywords are only as good as the list. 13 entries say "per CFO" and none of the default words catches them: add keywords from what you see in the client's journal and hear in interviews.
How to select and document the entries to test
- Start from the risks, not the tool. Write down the manipulation you fear (revenue pulled forward at year end, provisions released, expenses capitalised) and which criterion would find it.
- Combine criteria. Manual entry + year end + revenue account says far more than any of the three alone.
- Select every entry that meets the high-risk combinations, plus amounts above a threshold and a few picked unpredictably so management cannot anticipate the test.
- Examine each entry: support, approval under the entity's authorisation rules, business rationale, who posted it and whether they were allowed to, and whether it was reversed after the period end.
- Document: the population reconciliation, the criteria and why, each filter and its result, the selected entries, the work on each, and the conclusion.
Journal entry testing is not statistical sampling
The selection is targeted: you pick the entries that meet risk criteria and examine all of them. There is no sample size to compute and no result to project, and a clean entry says nothing about the ones you did not look at. To draw a conclusion about a whole population you need sampling under ISA 530 (see ISA 530 and statistical vs non-statistical sampling). The difference between selecting specific items and sampling is covered in audit sample selection methods.
Journal entry testing in AuditSampler Pro
The journal entry testing module imports the journal as CSV or Excel, scores every line on weekends, holidays, round amounts, period end, keywords (your own included), approval limit and duplicates, flags users with unusual patterns, and sorts the lines by score. Account combinations, post-closing entries and large amounts are filtered separately.
Frequently asked questions
Is journal entry testing required on every audit?
Yes. ISA 240 and AU-C 240 treat the risk of management override of controls as present in every entity and require journal entries and closing adjustments to be tested on every audit, whatever the assessed risk.
What are the criteria for journal entry testing?
Manual entries, entries at weekends or on holidays, at period end, with round amounts or just below an approval limit, by unusual users, with keywords such as adjustment or reclass, duplicates, unusual account combinations, post-closing entries and very large amounts.
How do you do journal entry testing in an audit?
Get the complete journal and reconcile it to the trial balance, define criteria from your fraud risk assessment, filter and combine them, select the high-risk entries plus large and unpredictable ones, examine each against its support, and document the whole chain.
What does the AICPA require for journal entry testing?
AU-C 240 requires the auditor to test the appropriateness of journal entries and other adjustments: inquire of people involved in financial reporting, select entries made at the end of the period, and consider testing entries throughout the period.
How many journal entries should be tested?
There is no set number. You test every entry that meets the high-risk criteria you defined, plus a few chosen unpredictably. It is not sampling, so there is no sample size to compute.
- IAASB, ISA 240 The Auditor's Responsibilities Relating to Fraud in an Audit of Financial Statements, and ISA 240 (Revised), 2025
- AICPA, AU-C 240 Consideration of Fraud in a Financial Statement Audit
- PCAOB, AS 2401 Consideration of Fraud in a Financial Statement Audit
- ICAC, NIA-ES 240 (Spain)
- IAASB, ISA 530 Audit Sampling
From the journal to the list of entries to test
AuditSampler Pro imports the journal, scores every line on the risk criteria you set and sorts it so you start with what matters. Client data stays on your computer.