What it means
The law describes the frequency of the first significant digit in naturally occurring data such as invoice values, expense claims, populations and river lengths. Roughly 30% of such numbers begin with 1, while under 5% begin with 9, which surprises most people who expect an even spread of about 11% each.
The reason is that many datasets grow multiplicatively rather than by fixed steps. A value has to increase by 100% to move from a leading 1 to a leading 2, but only about 11% to move from a leading 9 to a leading 1, so numbers spend more time in the lower bands.
In finance the practical use is anomaly detection. Fabricated numbers tend to be invented by humans who unconsciously spread digits evenly or cluster them just below approval thresholds, so a digit distribution that is too flat can flag manipulated expenses, invoices or journal entries.
The test is a screen, not proof. A deviation tells you where to look, and the follow-up work is ordinary audit testing: pulling the underlying documents, tracing approvals and interviewing the people involved.
Benford's Law also has real limits. It does not apply to data with assigned or constrained numbers such as invoice sequence codes, telephone numbers, prices clustered around a fixed point, or any dataset with a narrow range or too few records to be meaningful.
Modern accounting software has made the test cheap to run. A first-digit analysis over a full general ledger takes minutes in a spreadsheet or a data tool, so many internal audit teams now run it as a standing quarterly control rather than as a special investigation.
In practice
Real-world examples.
Example
A retailer runs a first-digit test across a year of expense claims and finds a heavy cluster of amounts starting with 4. Investigation shows several managers routinely submitting claims just below a $500 approval threshold.
Example
A forensic accountant reviewing a suspected revenue overstatement applies the test to reported monthly sales figures. The digit distribution is unusually flat, supporting a hypothesis that some figures were typed in rather than accumulated from transactions.
Example
A tax authority screens submitted returns before selecting audit cases. Returns whose deduction figures deviate sharply from the expected digit pattern are moved up the queue for manual review. Most turn out to be innocent, but the screen raises the hit rate enough to justify the analysis.
Think of it
“Benford's Law shows what digit patterns should look like-deviations suggest fraud.
Formula
Calculation
Formula: The expected proportion of numbers beginning with digit d is P(d) = log10(1 + 1/d), for d from 1 to 9. Expected count = P(d) x Total records.
An internal audit team tests 5,000 supplier invoices. For d = 1, P(1) = log10(1 + 1/1) = log10(2) = 0.301, so the expected number of invoices starting with 1 is 0.301 x 5,000 = 1,505.
The team counts only 1,180 invoices starting with 1, a shortfall of 1,505 - 1,180 = 325 records. Meanwhile for d = 9, P(9) = log10(1 + 1/9) = 0.0458, giving an expected count of 0.0458 x 5,000 = 229, but the actual count is far higher. That imbalance points the team towards invoices priced just under an approval limit, which is where they focus their detailed testing.Case study
Seen in the real world.
Ferngate Utilities is a fictional company used to illustrate the technique. Its internal audit team ran a first-digit analysis over 5,000 supplier invoices processed in a single year, expecting roughly 0.301 x 5,000 = 1,505 invoices to begin with the digit 1.
The actual count was 1,180, some 325 fewer than expected, while invoices beginning with 4 and 9 were heavily overrepresented. The pattern was consistent with values being nudged towards, but kept below, a $5,000 single-approval limit.
Detailed testing in this illustrative case found one procurement officer splitting larger orders into pairs of smaller invoices to stay inside his own authority. The digit test did not prove anything by itself; it simply told a small audit team where to spend its limited hours.
Watch out
Common mistakes.
- Treating a failed Benford test as evidence of fraud. It is a screening indicator that directs further work, never a conclusion on its own.
- Applying the test to unsuitable data. Sequential invoice numbers, postcodes and tightly clustered prices do not follow the distribution, so a deviation there means nothing.
- Running it on a handful of records. Small samples produce wild swings, and most practitioners want at least several hundred and preferably thousands of values.
Questions
People also ask.
How many records do I need?
There is no hard rule, but tests on fewer than a few hundred numbers are unreliable, and results become meaningful in the low thousands.
Does it work on second digits too?
Yes, extended versions test the second digit, the first two digits and digit pairs, and those tests are often more revealing for threshold manipulation.
Can honest data fail the test?
Certainly, because business processes create legitimate clustering around round prices, standard fees and contract values, which is why every exception needs human review.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%