Conventional DBEF methods are based on two general observations. First, many natural processes produce quantities in which the numerals are distributed according to Benford’s Law. Hill (1995) proved that in numbers drawn from a random mixture of distributions their numerals converge to a logarithmic distribution. Under Benford’s Law, in the decimal number system the distribution of numerals approaches uniformity with digit order, and is practically uniform already for the fourth digit. The second observation is that human subjects of varied backgrounds do not generate numbers that follow Benford’s Law when they are asked to generate “random” numbers in their heads (see Nickerson (2002) and Beber and Scacco (2012)). Combining these two observations, we expect numbers generated by accurate elections to conform with Benford’s Law, while numbers invented by fraudsters should not. This is the strong distributional assumption of digit-based election forensics.