Building a risk-based sample of expense applications

A sample that only picks the largest claims will miss patterned small abuses. A sample that is purely random may under-represent overseas lodging. For travel expense applications, we usually blend three lenses: monetary materiality, destination or category risk, and behavioural flags such as repeat high claimants or frequent exceptions.

Monetary materiality starts from your own thresholds — often a fixed KRW amount for lodging or a percentage of the population total. Destination risk considers cities where average hotel rates sit near or above policy caps. Behavioural flags come from the extract itself: the same employee appearing with weekend returns, or multiple applications amended after first rejection.

We keep a slice of the sample random so departments with modest average claims still see testing. That balance is explained in the engagement letter so audit committees understand why not every tested item is a headline amount. After testing, exception rates are reported both for the risk-weighted slice and, where useful, for the random slice separately.