Revisiting Benford’s Law with added AI horsepower

If you’ve passed the CFE Exam or the CPA exam, you’re familiar with Benford’s Law. This principle observes the unexpected regularity that in any large, randomly produced set of natural numbers, such as streamflow statistics, town and city populations, or corporate sales or payment amounts, around 30% of the numbers will begin with the digit 1, 18% with 2, and so on; with the smallest percentage beginning with the digit 9.

Accountant and fraud investigator Mark Nigrini, Ph.D., popularized it in his book “Digital Analysis Using Benford’s Law,” first published in 2001. You’ve probably used Benford’s Law to analyze accounts payable amounts, purchasing card data and journal entries in your search for irregularities or risk areas. By searching for cases where the expected proportion of the first (as well as the first two, or even first three) digits in a payment or transaction stream don’t conform, you’ll find indications that someone might be overriding a control or manipulating the numbers — disrupting the digit patterns.