Benford's Law states that in data sets, the leading digit is distributed in a non uniform way. The number 1, is not represented 11.1% of the times (as one would expect), but about 30% and the number 9 is represented only about 5%. This theory of leading digits is so consistent that it is used for fraud detection by accountants and in the banking system. It has also proven useful to detect electoral fraud. It is admissible evidence in the United States criminal courts at the federal , state and local levels.
Benford's Law it is not useful when dealing with data sets with less than 500 numbers, when data is generated by formulas and with data restricted by maximum or minimum numbers.
Please read the following article for further informtion