Rule of the week: Excessive Frequency

The moment this shows up

You're reconciling the expense claims for the month and one name keeps appearing. Not for a large amount — nothing over $120 — just... a lot. Fourteen separate reimbursement requests from the same person, when everyone else on the team submitted three or four. Each one looks fine in isolation: a taxi, a client lunch, a parking receipt. Nobody would query any single line. But the pattern is the thing, and pattern is exactly what you don't see when you're approving claims one at a time in a queue.

This is how genuine problems hide in expense claims: not in the size of any one transaction, but in how often one person submits relative to everyone doing a similar job. It can be entirely innocent — a field sales role that genuinely incurs more costs than a desk role. It can also be someone who has worked out that small, frequent claims attract less scrutiny than one large one.

A worked example

Say you're reviewing a 14-person team's expense claims for the quarter.

Across the team, the average is 9 claims per person for the quarter, with a standard deviation of 3.4. Two standard deviations above that average is roughly 15.8 — so anyone submitting 16 or more claims sits outside the normal spread for this group.

One employee, a business development coordinator, submitted 22 claims: taxis ranging $18–$45, meals between $22 and $65, and parking around $12–$28 each. Total value: $1,140 — not large, and every receipt attached is legitimate. But 22 claims against a team average of 9 is a frequency outlier, not a value one. It's worth asking why: new client-facing responsibilities, a genuine change in role, or claims being split into smaller amounts than one combined claim would be.

How to find it manually

You don't need a tool for this — it's a spreadsheet exercise, and it's honest work.

1. Export all expense claims for a fixed period — a month or a quarter, whichever matches your normal reporting cycle. Include employee name, claim date, and amount.
2. Pivot to count claims per employee for that period. A simple COUNTIF or pivot table on employee name is enough.
3. Calculate the average claim count across the team: AVERAGE() on the per-employee counts.
4. Calculate the standard deviation: STDEV.S() on the same range.
5. Set the threshold at average + (2 × standard deviation).
6. Flag anyone above that threshold and pull their individual claims to review the pattern — same vendor repeated, amounts just under an approval limit, timing clustered around month-end.

The honest cost: this method needs a reasonable sample size to mean anything. With a team of four, one person having a busy month will blow out your standard deviation and produce a meaningless threshold. It also needs recalculating every period — last quarter's average doesn't tell you about this quarter's behaviour, and a one-off review misses drift that builds slowly over several months. Budget 30–45 minutes if your export is clean; longer if you're pulling from a system that doesn't let you filter by employee and date range in one go.

FAQ

Why is excessive claim frequency treated as a red flag if every individual receipt is valid? Because fraud and process abuse in expense claims usually avoid single large amounts, which get scrutinised, in favour of many small ones, which typically don't. A statistical outlier in frequency — someone claiming far more often than peers in a similar role — is worth a conversation even when every receipt attached is genuine, because it's the pattern that indicates deliberate structuring, not the amount.

What counts as "excessive" — is there a fixed number of claims that triggers review? No fixed number works across all teams, because normal frequency depends on role. The standard approach is relative, not absolute: calculate the average and standard deviation of claim counts across a comparable group for a fixed period, then flag anyone more than two standard deviations above that average. That adjusts automatically for teams where frequent small claims are normal (field sales, travel-heavy roles) versus teams where they aren't.

How the rule does this

The rule (EXP-004, Excessive Frequency) runs the same average-and-standard-deviation calculation continuously rather than as a periodic manual pull, which means it needs a stable comparison group to mean anything — a small team, a newly hired cohort, or a business with genuinely uneven roles will produce false positives or miss real ones until enough history accumulates. It doesn't judge whether the claims themselves are legitimate; it flags the pattern for a human to look at, at varying severity depending on how far above the threshold the count sits. What it catches that a monthly manual pull won't is drift — someone's frequency creeping up gradually over several periods, past the point where any single month would have triggered a review.

Finanomaly is currently waitlisted.