Selecting from a list instead of a number range
The random number generator draws item numbers from a range. This tool draws from the list itself, which suits populations that are not numbered 1 to N: customer names, vendor IDs, journal entry references, or rows exported from the general ledger. It returns the items with their position in the list and the line they came from, so you can go straight to the source document.
Method
Random selection uses a partial Fisher-Yates shuffle, the in-place form that Durstenfeld published in 1964 from the shuffle of Fisher and Yates. For each of the n draws, the next item is swapped with one chosen from the items not yet used, so every item has the same chance and nothing repeats. Systematic selection computes the interval k as the population divided by the sample size, draws a random start between 0 and k, and takes the item at floor(start + j × k) for each j from 0 to n − 1. Seed text is hashed (xmur3) into a mulberry32 generator, so the sequence is fully determined by the seed.
Where this fits in the standard
AU-C 530 lists random selection, systematic selection, monetary unit sampling, haphazard selection and block selection as selection methods, and requires that sampling units be selected so each has a chance of selection. A statistical sample needs a random-based selection method. See the guide to audit sampling methods and the AU-C 530 summary for how selection connects to sample size and evaluation.
When to use it, and limits
- Use random selection when the population has no order that matters, or when you want to avoid any link between position and selection.
- Use systematic selection on an ordered list when you want even coverage across the period, and check that the list does not cycle at the interval.
- The generator is repeatable, not cryptographic. That is what an audit file needs, since a reviewer must be able to re-perform it.
- Random selection removes selection bias. It does not decide the sample size, and it does not test completeness of the list. Reconcile the list to the ledger first.
- For a population dominated by a few large items, test those items outright and sample the rest, or use monetary unit sampling.
Illustrative figures only. Check the result against your firm's methodology before relying on it.
Common questions
How do I pick a random sample from a list for an audit?
Paste the population one item per line, enter the sample size from your sampling plan, keep or change the seed, and read the selected items. Trace each one to its source document. Save the seed, the method and the item count in the workpaper so a reviewer can run the same selection and get the same items.
What is the difference between random and systematic selection?
Random selection gives every item the same chance and picks without regard to position. Systematic selection takes a random start and then every kth item, where k is the population divided by the sample size. It is quick on an ordered list but can miss a pattern that repeats at the same interval, such as a monthly or weekly cycle.
Does a random selection make the sample statistical?
Only with the second feature. AU-C 530 treats a sample as statistical when it uses random selection and probability theory to evaluate the results, including sampling risk. A random draw evaluated by judgment alone is a nonstatistical sample with a random selection method, which is still a legitimate approach.
Why does the tool ask for a seed?
The seed starts the pseudo-random generator. The same list, seed, method and sample size always produce the same items, which lets a reviewer re-perform the selection. A different seed gives a different, equally valid sample. Choose the seed before you look at the list, and do not keep changing it until the sample looks agreeable.
What happens to duplicate lines in the list?
By default every pasted line is an item, so a repeated line can be selected twice because it counts twice. Tick the duplicate option to remove exact repeats first. Decide beforehand whether two identical lines are two transactions or one, because that choice changes the population count.
Runs in your browser. The figures you type are not sent to CORAA or stored. Check the result against your firm's methodology before relying on it.