Natural vs Fabricated Data Signatures

mental-model

Real-world processes and fabricated data often leave different statistical fingerprints. People inventing numbers tend to avoid extremes and crowd the middle, so a distribution’s shape can identify records worth investigating.

Ask someone to fabricate a long column of numbers and the result may look random row by row—yet betray its author in aggregate. The edges thin out and the middle swells because people are reluctant to choose extreme digits.

The fingerprint survives the disguise

A natural process generates values under its own constraints; an inventor generates them through human intuitions about what looks plausible. Those intuitions introduce repeated biases—especially avoidance of extremes—that individual entries conceal but a frequency curve exposes. Benford’s law is a related case: naturally produced leading digits can have a characteristic non-uniform shape. The reusable move is statistical-anomaly-detection: compare the observed distribution with the signature expected from the process that supposedly produced it.

Where it shows up

The hand-typed ledger

A person typing invented business figures may vary every entry and still produce too few values at the sides and too many in the middle. The fabrication lives not in one suspicious number but in the collective shape.

A fingerprint is not a conviction

An unusual curve does not establish fraud; it only shows that the claimed generating process deserves scrutiny. Small samples, bounded choices, or legitimate human input can also produce non-natural shapes. Treat the pattern as a lead, and avoid the McNamara Fallacy of letting a count replace context.

Inspect the distribution before the entries

Tomorrow, take one dataset whose numbers are supposed to arise naturally, plot the frequency of its first or final digits, and mark the largest deviations from the expected shape. Investigate those records individually; do not label them fraudulent from the curve alone.

Episodes that teach this

The Strange Logic Behind Catching Business Frauds - Benford's Law · explained at 8:47 1,703 views
"if somebody sat and typed in... a bunch of numbers then usually... the curve is low on the sides and high in the middle because humans don't want to do anything extreme"