How this instrument works
Benford's law describes a pattern found across a surprising range of naturally occurring numerical data — populations, river lengths, stock prices, physical constants, accounting figures — where the leading (first) digit is not uniformly distributed. Instead of each digit 1 through 9 appearing about 11.1% of the time, as intuition might suggest, the digit 1 leads roughly 30.1% of numbers, the digit 2 about 17.6%, and the frequency keeps falling as the digit gets larger, down to about 4.6% for a leading 9.
The pattern was first noted by astronomer Simon Newcomb in 1881, who observed that the early pages of logarithm tables were far more worn than the later ones — implying people looked up numbers starting with 1 much more often than numbers starting with 9. Physicist Frank Benford rediscovered and documented it far more thoroughly in 1938, testing it against thousands of values from wildly different sources, and the law now carries his name.
The underlying reason is scale invariance: many real quantities span several orders of magnitude, and a number has to 'pass through' the range starting with 1 (from 1 to just under 2) before it can reach the range starting with 2, and so on — and on a logarithmic scale, that first range is proportionally the largest. That's exactly what the formula P(d) = log₁₀(1 + 1/d) captures, and it's also why Benford's law shows up as a screening tool in forensic accounting and fraud detection: numbers that are made up by people, rather than arising naturally, tend to have leading digits closer to uniformly distributed, which stands out against the expected Benford pattern.
- Enter a whole number from 1 to 9 into Leading digit (1-9) — the first digit you want the expected frequency for.
- Read Probability of this leading digit for the fraction of naturally occurring values expected to start with that digit.
- Digit 0 isn't a valid leading digit — a number can't start with 0 — so the field only accepts 1 through 9.
- Check each digit from 1 to 9 in turn to see the full distribution: it decreases steadily from digit 1 to digit 9.
Worked example — the probability of a leading digit 1
Enter 1 into Leading digit (1-9). The instrument computes P(1) = log₁₀(1 + 1/1) = log₁₀(2). Probability of this leading digit reads 0.301030, or about 30.1%.
That figure is the headline result of Benford's law: in a data set that follows the law, nearly a third of all values are expected to begin with the digit 1 — almost three times the roughly 11.1% you'd expect if every leading digit were equally likely. It's the largest single-digit probability in the whole distribution, and every other digit's probability is smaller, tapering down to about 4.6% by the time you reach 9.
Questions
Why isn't every leading digit equally likely?
Because most real quantities span several orders of magnitude, and on a logarithmic scale the interval from 1 to 2 is proportionally wider than the interval from 8 to 9 — a growing quantity spends more of its 'journey' with a leading 1 than with a leading 9 before it rolls over to the next order of magnitude. That geometric fact, not any property of the digits themselves, is what produces the uneven distribution.
What's the probability for each digit 1 through 9?
Applying P(d) = log₁₀(1 + 1/d): digit 1 is about 30.1%, 2 is about 17.6%, 3 about 12.5%, 4 about 9.7%, 5 about 7.9%, 6 about 6.7%, 7 about 5.8%, 8 about 5.1%, and 9 about 4.6%. The nine probabilities sum to exactly 100%, and each one is smaller than the last as the digit increases.
Does every data set follow Benford's law?
No. It applies best to data that spans multiple orders of magnitude and arises from a mix of natural growth or multiplicative processes — populations, financial statements, scientific constants. Data confined to a narrow range (like adult heights in centimetres, or two-digit exam scores), or data that's fundamentally uniform by design (like lottery numbers or assigned ID numbers), doesn't reliably follow it.
How is Benford's law used in fraud detection?
Auditors and forensic accountants compare the leading-digit distribution of a real financial data set — expense reports, tax filings, transaction logs — against the expected Benford proportions. Numbers invented by a person tend to have leading digits closer to uniformly spread across 1 through 9, because people are poor at mimicking the natural skew toward small leading digits, so a significant deviation from the expected pattern can flag data worth a closer look. It's a screening signal, not proof of wrongdoing on its own.
Why does the digit 9 come out so much rarer than 1?
P(9) = log₁₀(1 + 1/9) = log₁₀(10/9) ≈ 0.0458, versus P(1) = log₁₀(2) ≈ 0.3010 — roughly a 6.6-to-1 ratio. The gap comes directly from how the formula shrinks as d grows: 1 + 1/d falls fastest between small values of d and flattens out as d approaches 9, so the difference between consecutive digit probabilities keeps narrowing even as the absolute values keep dropping.