Threshold Effects

mechanism

Approval cutoffs do not merely classify behavior; they reshape it. A pile-up just below a limit can therefore reveal the incentive created by the rule rather than the natural distribution of transactions.

In one expense trail, costs were not scattered naturally: they were being squeezed below 5,000, with others squeezed below 10,000. The suspicious feature was not an extravagant outlier but a crowd of transactions that were almost large enough to require scrutiny.

The cutoff becomes a target

A rule creates two zones: below the threshold, a transaction passes with less friction; above it, approval or attention increases. Anyone able to split, resize, or relabel spending now has a reason to land on the convenient side. Repeated across many decisions, that adaptation produces bunching near the boundary. Unlike the threshold-model-of-collective-behavior, the effect does not require people to copy one another: the shared cutoff independently steers them toward the same narrow band. The resulting data is partly a record of spending and partly a record of the control system acting on spending.

Where it shows up

Expenses just under approval limits

Clusters below 5,000 and 10,000 show how two administrative boundaries can leave matching fingerprints in expense data. Looking only for unusually large claims—or mechanically applying outlier-trimming—could discard the wrong signal: here, suspicious regularity sits immediately below ordinary-looking limits.

A fingerprint is not a confession

Bunching near a cutoff is a reason to investigate, not proof of fraud. Legitimate pricing, budgeting, or operational constraints may also concentrate values. And once you inspect only transactions admitted by a rule, selection itself can manufacture patterns that do not exist in the full population.

Audit the boundary, not just the extremes

For every approval limit in your system, compare a narrow band immediately below it with the band immediately above it, then inspect repeated vendors, submitters, dates, or split purchases in the lower band. If the cutoff repeatedly attracts transactions, redesign the control—potentially using fewer, more dependable boundaries in the spirit of rule-certainty-over-rule-proliferation—instead of merely raising the limit.

Episodes that teach this

The Strange Logic Behind Catching Business Frauds - Benford's Law · explained at 6:08 1,703 views
"extra expenses being squeezed into the below 5000... and some being squeezed into the below ten thousand"