Rule of the week: Receipt Splitting

The error, and why it happens

You're reconciling expense claims and you notice something odd: the same employee has lodged three taxi receipts to the same merchant, on the same day, each one just under the $75 no-receipt-required threshold, or just under whatever approval limit sits on their cost centre. Individually they look unremarkable. A cab to the airport, a cab back, a cab to a client site. It's only when you add them up that you notice they total more than a single trip should ever cost, and more than the approval limit that would have triggered a manager's sign-off.

This isn't usually fraud in the dramatic sense. More often it's an employee who knows their manager glances at anything over $100 and doesn't glance at anything under it, and who has learned — consciously or not — to keep every claim just under the line. Sometimes it's genuinely three separate trips. Sometimes it's one trip split into three receipts because the taxi driver was happy to print them that way. The pattern is identical either way, which is exactly why it's easy to miss: nothing in a single line item looks wrong.

A worked example

Say the approval threshold for un-countersigned expense claims is $100.

• 14 March, Employee: J. Halloran, Merchant: Silver Cabs, $82.00
• 14 March, Employee: J. Halloran, Merchant: Silver Cabs, $76.50
• 14 March, Employee: J. Halloran, Merchant: Silver Cabs, $68.00

Three claims, same day, same merchant, same employee, each safely under $100, none of them flagged by the approval workflow because none of them crossed it. Total: $226.50 — well above the threshold that would normally have required a second signature. If this were one $226.50 claim, someone would have had to approve it. Split three ways, nobody did.

How to find it manually

You'll need your expense claims export with, at minimum, employee name, date, merchant, and amount — most systems (Xero Expenses, Expensify, a plain spreadsheet) will give you this in a CSV.

1. Sort by employee, then by date, then by merchant. This groups anything that could be a split before you've written a single formula.
2. Add a helper column that concatenates employee + date + merchant into one key.
3. Use a pivot table (or SUMIF/COUNTIF if you'd rather stay in formulas) to sum amounts and count line items per key.
4. Filter for keys where the count is more than one and the summed amount exceeds your approval threshold, but no individual line does.
5. Pull the flagged groups and check them against actual receipts or trip logs — some of this will be legitimate (two genuinely separate trips), and you need the source documents to tell the difference, not just the numbers.

The honest cost of this: it's not conceptually hard, but it's tedious to do properly across a few hundred employees and a full quarter of claims, and it's easy to under-filter (missing splits across midnight, or across near-identical merchant names like "Silver Cabs" vs "Silver Cabs Pty Ltd") or over-filter (flagging a genuine run of separate small trips). Doing it well wants someone comfortable with pivot tables and prepared to spend an hour or two per review period, plus the judgement to know a real split from a coincidence.

FAQ

How do I tell if an employee is splitting receipts to stay under the approval limit? Group their claims by date and merchant, sum each group, and compare that sum — not the individual line amounts — against the approval threshold. A pattern of multiple same-day, same-merchant claims that each sit just under the limit is the tell.

Is receipt splitting always deliberate? No. It can be genuine repeat trips, a merchant that splits its own invoices, or a policy quirk (e.g. daily meal limits) that produces multiple legitimate claims. Treat every flagged group as a question to ask, not a conclusion.

How the rule does this automatically

The rule can't see receipt images or trip logs, so it can't tell a deliberate split from three genuine separate trips — it will flag both, and it will miss splits where the merchant name is entered inconsistently across claims. What it does catch: for every employee, date, and merchant combination, it sums the expenses and checks whether individually-small amounts add up to something above the relevant threshold, surfacing it as EXP-001 with a severity that reflects how far over the total sits.

Finanomaly is currently waitlisted.