Benford's law
About this app
Number forensics on the first digit: data points land on a glass log drum whose sector width is log₁₀(1+1/d). Rescaling only turns the drum; body height and .99 prices fail, manipulated expenses give themselves away (Nigrini MAD, χ²).
The first-digit law
In many naturally occurring data sets, first digits are not evenly distributed. Benford’s law says a number starts with digit d with probability P(d) = log₁₀(1 + 1/d): about 30.1 % start with 1 and only about 4.6 % with 9. It holds approximately for data that span several orders of magnitude and arise multiplicatively, such as growth processes, invoice amounts or powers of two. Benford’s distribution is also the only first-digit distribution that stays unchanged when you convert to another unit.
What you see
Each data point lands on a glass drum, a logarithmic scale wound up like a slide rule: the angle is the mantissa of log₁₀ x, and one turn is one power of ten. The sector for digit 1 covers 30.1 % of the circumference. You can choose exact sequences (2ⁿ, Fibonacci, n!), simulated processes such as growth and invoices, counterexamples such as body heights and prices ending in .99, and expense claims with an adjustable share of fakes. A test panel shows bars, Nigrini’s MAD, a χ² test and a z-test per digit.
Things to try
Multiply all values by a factor k: the drum only rotates, and Benford data keep their distribution. Body heights, by contrast, move as a clump through the sectors. In the growth data set every number starts with 5 at first; increase the number of periods. For the expenses, raise the share of fakes until the MAD flags it, or type in your own made-up amounts and test them.
Frequently asked questions
- What is Benford’s law?
- Benford’s law describes how often each first digit appears in many real-world data sets: P(d) = log₁₀(1 + 1/d). Almost one number in three starts with 1, only about one in twenty-two with 9. It does not apply to bounded or assigned numbers such as heights, postcodes or prices ending in .99.
- How is Benford’s law used to detect fraud?
- Invented numbers tend to be too evenly spread, and amounts just below approval limits pile up on certain digits. Auditors compare observed digit shares with Benford, for example using Nigrini’s mean absolute deviation (MAD). A deviation is a reason to look closer, not proof of fraud.
- Why does Benford’s law work?
- When data span several orders of magnitude and arise by multiplication, the fractional parts of their logarithms spread almost evenly. On a logarithmic scale, digit 1 owns the largest stretch, from log₁₀ 1 to log₁₀ 2. That is why most values land there.
- Can I test my own numbers in the app?
- Yes. In the forensics panel you enter your own amounts; the app counts the first non-zero digit and rates the distribution with MAD, a χ² test and z-tests. A reliable verdict needs a few hundred values.
Subject: Data & decisions
More from Data & decisions
- Anchoring effect
- Berkson's paradox
- Cherry-picking: base period
- Cherry-picking: base period 2.0
- The three-cowboy duel (truel)
- The three-cowboy duel 2.0
- The three means
- The three means 2.0
A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps