Toolgeist › Developer

Benford's Law Auditor — Free Chi-Square & MAD Fraud Test

Paste any column of numbers — invoice totals, expense claims, transaction amounts, reported figures — and get the first-digit analysis a forensic accountant would actually run. This computes the real Pearson chi-square goodness-of-fit statistic against P(d) = log₁₀(1 + 1/d) with 8 degrees of freedom, Nigrini's mean absolute deviation with his published conformity bands, and a per-digit Z statistic with continuity correction so you can see exactly which leading digits are over- or under-represented. Nothing is uploaded and no threshold is faked: every number on the page is derived from the data you paste.

How to use Benford's Law Auditor — Chi-Square & MAD Fraud Test

  1. Paste your numbers — one per line, or comma / tab separated. Currency symbols, percent signs, accounting parentheses and thousands separators like 1,234.56 are all handled; check the "read as" line to confirm the parse.
  2. Compare the blue observed bars against the amber Benford markers.
  3. Check the MAD verdict and the chi-square result together.
  4. Look for digits flagged with Z > 1.96 — those are the ones deviating significantly.

Frequently asked questions

What is Benford's Law in one sentence?
In data spanning several orders of magnitude, the leading digit is 1 about 30.1% of the time and 9 only about 4.6% — because each digit's share is log₁₀(1 + 1/d), the width it occupies on a logarithmic scale. It is not a curiosity about numbers; it is what you get whenever a quantity grows multiplicatively across scales.
Does failing this test prove fraud?
No, and treating it that way is the classic misuse. A failed test tells an auditor where to sample next — it is grounds for investigation, never proof on its own. Plenty of honest datasets fail because they are bounded or assigned rather than naturally occurring.
When does Benford’s Law not apply?
Whenever the data has a built-in floor or ceiling, is assigned rather than measured, or spans a narrow range: phone numbers, ZIP codes, human heights, ages, invoice sequence numbers, prices clustered at 9.99. Nonconformity there means the data is bounded, not that anyone cheated. Applying the test to such data is the most common way people reach a wrong conclusion with it.
Why report both chi-square and MAD?
Chi-square scales with sample size, so a very large dataset will reject Benford on a deviation far too small to matter in practice — a well-known trap. Nigrini's MAD does not scale with n, which is why forensic accounting practice reports it alongside, and why both appear here rather than a single reassuring verdict.
How many values do I need?
Roughly 100 as a working minimum, and more is better. Below about 50 the chi-square statistic is simply not trustworthy, and this tool warns you rather than printing a confident-looking verdict on a sample that cannot support one.
Is my data uploaded anywhere?
No. The parsing and all statistics run in JavaScript in your own browser. Nothing is transmitted, which matters given this tool is most useful on financial records.

Related Developer tools

Also searched for: benford's law calculator · first digit law fraud detection · chi square goodness of fit calculator · nigrini MAD test · forensic accounting digit analysis

Browse all 738 free tools on Toolgeist →