Two numeric patterns turn up in almost every journal entry test: amounts that are suspiciously round, and amounts that sit just under an approval limit. Both are worth a look. Neither is proof of anything. The skill lies in knowing which ones are routine, which ones need a question and which ones need evidence.
Last reviewed: 17 October 2026. Facts checked. We did not open the full ICAI text of SA 240 or any published study on digit patterns, so no statistic about how often round amounts indicate misstatement is claimed here. Limits and ranges in the examples are invented. Where we describe SA 240 we rely on an ICAI newsletter extract and a web search on ISA 240; paragraph numbers are not cited.
Why these patterns get flagged
Real transactions mostly produce untidy numbers: GST at 18 per cent on an invoice, a salary after deductions, a foreign currency conversion. Estimates and plugs, by contrast, tend to be tidy. A person who is making an entry up, or backing into a figure to reach a target, often picks a number that feels natural. That is the logic of the round-sum test.
The approval limit test follows a different logic. If a manager can approve spend up to a set amount and anything higher needs a second signature, an entry just below the line may be placed there on purpose, or a larger amount may have been split into pieces.
Both tests fit into the journal entry work SA 240 requires; see our overview of journal entry testing procedures and red flags.
Round-sum entries: separate the routine from the odd
Run the test, then sort the results before asking anyone anything.
Usually routine:
- Rent, lease instalments, retainer fees and loan EMIs that are contractually fixed.
- Fixed contractual or statutory payments where the amount is set in advance.
- Bank charges, subscriptions and fixed-price AMC payments.
- Depreciation or amortisation on a fixed schedule.
- Cash deposits and withdrawals in round denominations.
Worth a question:
- Provisions and accruals in round figures with no calculation behind them.
- Round-sum adjustments to revenue, receivables, inventory or reserves.
- Round entries posted in the last days of a period, especially if they reverse in the next.
- Round entries between group companies or related parties.
- Round entries posted manually where similar items are usually system-generated.
The practical step is to define "round" more than once. Test for multiples of 1,000, then 10,000, then 1,00,000 (one lakh), because different amounts point to different behaviours. A whole-lakh entry posted manually to an expense or reserve account near year end tells you more than a Rs 5,000 round entry to a petty cash ledger.
Just-below-limit entries: first learn the limits
You cannot test against a limit you do not know. Ask for the delegation of authority or approval matrix: who can approve what and for which type of transaction. Then:
- For each limit, list the entries within a narrow band under it. Choose the band based on the client's typical amounts rather than a fixed rule; the right width depends on the data.
- Compare the count in that band with the count in neighbouring bands. If the pile-up just under the line is far larger than a smooth decline would suggest, that is informative.
- Check for splitting: the same vendor, same date or consecutive dates, several entries that add up to more than the limit.
- Check who posts the entries. A pattern concentrated in one user or one department is more telling than a pattern spread evenly.
Splitting also appears in tax compliance: the same logic applies to payments and cash limits, though the rules involved are outside the scope of this post.
What the numbers cannot tell you
A round amount is not a misstatement. A just-below-limit entry is not fraud. Treat each as a pointer that narrows where to look. Three habits keep the conclusion honest:
- Compare with prior years. If the same round entries appear every year with support, the flag is likely noise.
- Compare with peers inside the entity. If one branch posts round entries and the other does not, ask why.
- Document what you excluded and why. A reviewer should be able to see that the decision not to follow up was reasoned.
A worked example (invented)
An auditor is testing a company with a Rs 2,00,000 limit for single-approver payments to vendors. Looking at March vendor entries, 31 entries fall between Rs 1,90,000 and Rs 1,99,999. In the same width band above the limit, there are 4. Fourteen of the 31 are posted by one accounts executive, and nine of those relate to one contractor, on consecutive days in the last week of the month.
None of this is a conclusion. The auditor asks to see the contract, the work completion evidence and the invoices. If the invoices are separate, dated and supported by delivery, the pattern may reflect billing cycles. If a single invoice was split, the issue becomes a control failure, and then a question about whether the payee is genuine. The figures are invented to show the method.
Using software without losing judgement
These tests are easy to run on a whole population with a script, a spreadsheet or an audit tool, and a full run is better than a sample. What software cannot do is decide which hits matter. After a run, rank the hits by combination, not by single flag: a round, manual, year-end, high-value, related-party entry posted by someone who rarely posts is a very different thing from a round entry that is rent.
Layering several indicators, as covered in the earlier journal entry guide, helps prioritise. Also keep the link between populations and sampling in mind; if you reduce hits by sampling, see how SA 530 applies when you can test everything.
What to write in the working paper
- The definition of round used and the ranges tested.
- The limits obtained and their source document.
- Counts of hits and how they were sorted.
- The entries followed up, the evidence obtained and the conclusion.
- The entries not followed up and why.
The takeaway
Round and just-below-limit entries are good questions, not good answers. Define them clearly, learn the client's limits first, separate the contractual from the odd, and follow the few that survive with real evidence.
Statutory facts on this page are checked against their sources, and the page says where it relied on secondary reporting. How we verify · Report an error