Calculate a sample size, right now.

What this does and doesn't do

• Nothing you type leaves your browser — there's no file, no upload, no server call.
• This gives one standard statistical calculation. SA 530 permits a non-statistical approach too, provided the basis is documented just as clearly — this tool doesn't decide which approach is right for a given population.

How to use this

1. 1

   Decide which test this sample is for

   A test of controls (did it operate) uses attribute sampling; a substantive test of detail (is the balance right) uses monetary-unit sampling.

2. 2

   Set tolerable error or deviation rate from materiality or risk tolerance

   Tolerable error for a substantive test is usually performance materiality; tolerable deviation rate for controls reflects how much control failure is acceptable.

3. 3

   Enter an honest expected error/deviation rate

   Zero if you have no reason to expect any — inflating this without evidence produces an oversized, not more rigorous, sample.

4. 4

   Select the sample using the method the size assumes

   Monetary-unit sampling needs a value-weighted or systematic selection, not simple random, to deliver the coverage this number is based on.

This free tool, and where it stops

One run, one file, four or five rules — versus every procedure a real engagement needs, run continuously against the full population, with every exception linked to its source document.