Audit Sample Size Calculator
Find how many records to test for a statistically sound audit sample.
Your Sample Size
Choosing how many records to examine is one of the most important decisions in any audit. Test too few, and your conclusions may not reflect the whole population. Test too many, and you waste time and money checking records that add no real assurance. The Audit Sample Size Calculator above removes the guesswork by telling you exactly how many items you need to sample, based on your population size, your chosen confidence level, and the margin of error you can accept.
What This Calculator Does
This calculator determines the minimum number of records you need to test so that your audit results are statistically valid. You enter three values: the population size (the total number of records, invoices, transactions, or items you could test), the confidence level (90%, 95%, or 99%), and the margin of error (the range of uncertainty you are willing to accept, in percent). The tool then applies the standard sample size formula, n = Z squared times p times (1 minus p), divided by e squared, and adjusts it with the finite population correction so the answer fits your actual population.
The calculator uses a conservative assumed proportion of 50%, which produces the largest possible sample size and therefore the safest answer. This is standard practice in audit sampling when you do not know the true error rate in advance. Along with the final required sample size, the calculator shows you the Z-score for your chosen confidence level and the initial sample size before the population correction, so you can see exactly how the numbers were derived.
How To Use This Calculator
Using the calculator takes less than a minute. First, enter your population size as a whole number. This is the total count of items in the group you are auditing, for example 5,000 sales invoices, 800 employee expense reports, or 12,000 inventory items. Next, choose your confidence level from the dropdown. The default is 95%, which is the most common choice in audit work. Then enter your margin of error as a percentage, for example 5. A smaller margin of error means more precision but requires a larger sample. Finally, click the Calculate button.
The result box will appear with three figures. The Z-score tells you the statistical value behind your confidence level: 1.645 for 90%, 1.960 for 95%, and 2.576 for 99%. The initial sample size is what the formula gives before adjusting for your population. The required sample size, shown in the green banner, is your final answer: the number of records you should actually test. If you enter an invalid value, such as a population of zero or a negative margin of error, a red error message will explain what to fix. The Reset button clears everything so you can run a new calculation.
The Formula Behind Sample Size
The calculation starts with the classic sample size formula used in statistics: n0 = Z squared times p times (1 minus p), divided by e squared. Here, Z is the Z-score that matches your confidence level, p is the assumed proportion (set to 0.5 for the most conservative result), and e is the margin of error written as a decimal. The assumed proportion of 0.5 matters because p times (1 minus p) reaches its maximum at 0.5, which gives the largest sample size. If you later learn that the true proportion is far from 50%, you could use a smaller sample, but 0.5 keeps you safe when you are unsure.
The second step applies the finite population correction: n = n0 divided by (1 plus (n0 minus 1) divided by N), where N is your population size. This adjustment matters whenever your population is not enormous. Without it, the formula assumes an infinitely large population and can recommend a sample larger than necessary. For very large populations, the correction has almost no effect, and the final answer is close to n0. For small populations, the correction shrinks the required sample noticeably, because testing a large fraction of a small group gives you most of the information anyway. The calculator always rounds the final answer up to the next whole number, because you cannot test a fraction of a record.
Worked Examples With Full Arithmetic
Example 1: A company wants to audit its 5,000 purchase invoices at 95% confidence with a 5% margin of error. The Z-score for 95% confidence is 1.96, so Z squared equals 3.8416. Multiply by 0.5 times 0.5, which is 0.25, giving 0.9604. The margin of error as a decimal is 0.05, and 0.05 squared is 0.0025. Divide 0.9604 by 0.0025 to get the initial sample size n0 = 384.16. Now apply the finite population correction: n = 384.16 divided by (1 plus (384.16 minus 1) divided by 5,000). The bracket becomes 1 plus 383.16 divided by 5,000, which is 1 plus 0.076632, or 1.076632. Dividing 384.16 by 1.076632 gives 356.82, which rounds up to 357 invoices. The calculator shows 357 records, and this matches the textbook answer for these inputs.
Example 2: An internal auditor must review 200 travel expense claims, choosing 90% confidence and a 10% margin of error for a quick review. The Z-score for 90% confidence is 1.645, so Z squared is 2.706025. Multiply by 0.25 to get 0.67650625. The margin of error 0.10 squared is 0.01, so n0 = 0.67650625 divided by 0.01 = 67.65. The correction is 1 plus (67.65 minus 1) divided by 200, which equals 1 plus 0.333253, or 1.333253. Dividing 67.65 by 1.333253 gives 50.74, which rounds up to 51 claims. Notice how the small population of 200 pulled the answer well below the initial 68, saving the auditor from testing extra records for no statistical gain.
Example 3: A bank audits 100,000 loan files at 99% confidence with a tight 3% margin of error. The Z-score for 99% confidence is 2.576, so Z squared is 6.635776. Multiply by 0.25 to get 1.658944. The margin of error 0.03 squared is 0.0009, so n0 = 1.658944 divided by 0.0009 = 1,843.27. The correction is 1 plus (1,843.27 minus 1) divided by 100,000, which equals 1 plus 0.0184227, or 1.0184227. Dividing 1,843.27 by 1.0184227 gives 1,809.93, which rounds up to 1,810 files. Here the population is so large that the correction barely changed the answer, dropping it only from 1,844 to 1,810.
Interpreting Your Results
Your required sample size is a minimum, not a target to hit exactly. If the calculator says 357 records, testing 357 randomly selected records gives you the stated confidence and margin of error. Testing a few extra records is harmless and sometimes practical, for example when you pull records in batches, but testing far fewer than the calculated number means your results no longer carry the confidence level you selected.
Pay attention to the relationship between the three numbers. If the initial sample size and the final answer are almost the same, your population is large enough that its exact size barely matters. If the final answer is much smaller than the initial figure, the finite population correction did heavy lifting, which is normal for small populations. Also remember that the sample must be selected randomly or by a proper systematic method. A statistically perfect sample size means nothing if you hand-pick the easiest records to test, because the math assumes every record had a fair chance of being chosen.
Choosing a Confidence Level and Margin of Error
The 95% confidence level is the default for good reason: it is the standard in most audit guidance and strikes a balance between assurance and effort. Choose 90% when the risk is low and you want a smaller, cheaper sample, such as a routine review of low-value transactions. Choose 99% when the stakes are high, for example when auditing financial statements that investors rely on, but be prepared for a much larger sample. Moving from 95% to 99% confidence can easily increase your sample by 70% or more.
The margin of error works the same way in reverse: smaller is stricter. A 5% margin of error is the common middle ground. Cutting it to 3% roughly triples the sample size, while relaxing it to 10% shrinks the sample to about a quarter. Before you demand a tiny margin of error, ask whether that extra precision would actually change any decision. In many audits, knowing the error rate within 5 percentage points is plenty, and the extra hundreds of records a 2% margin would require add cost without adding insight.
Common Mistakes to Avoid
The most common mistake is confusing the population with the sample. The population is every record that exists, not the number you feel like testing. Entering a guessed population gives you a wrong answer with false precision, so count or estimate the population carefully first. Another frequent error is entering the margin of error as a decimal, typing 0.05 instead of 5. This calculator expects a percentage, so type 5 for a 5% margin. A related mistake is mixing up confidence level and margin of error: confidence is how sure you want to be, margin of error is how close the estimate must be.
People also forget that the formula assumes simple random sampling. If your records are grouped in a way that matters, for example transactions clustered by branch or month, a plain random sample may miss important patterns, and you may need stratified sampling instead. Finally, do not treat the sample size as the whole plan. You still need to define what counts as an error, decide how to handle missing records, and document your selection method. The calculator gives you the number; your audit program gives it meaning.
Frequently Asked Questions
1. What is an audit sample size?
An audit sample size is the number of records an auditor tests out of a larger population to draw conclusions about the whole group. Instead of checking every invoice or transaction, the auditor checks a carefully chosen subset and uses statistics to generalize the findings. The sample size must be large enough that the results are reliable at the chosen confidence level and margin of error.
2. Why does this calculator assume a proportion of 50%?
A 50% assumed proportion produces the largest possible sample size, which is the safest choice when you do not know the true error rate in the population. The term p times (1 minus p) reaches its maximum value of 0.25 when p equals 0.5. If you already have reliable evidence that the error rate is much lower, you could justify a smaller sample, but 0.5 is the standard conservative default in audit sampling.
3. Which confidence level should I choose for an audit?
For most audits, 95% confidence is the accepted standard. Use 90% for low-risk, routine reviews where a smaller sample is acceptable, and 99% for high-risk areas such as financial statement audits where you need maximum assurance. Keep in mind that each step up in confidence increases the required sample size noticeably.
4. What margin of error is normal in audit sampling?
A 5% margin of error is the most common choice and works well for most audit purposes. Auditors sometimes use 3% when they need tighter precision and 10% for quick, low-risk reviews. Margins below 3% usually demand very large samples that are hard to justify, while margins above 10% give results too vague to support strong conclusions.
5. What is the finite population correction?
The finite population correction adjusts the sample size formula for populations that are not infinitely large. The basic formula assumes an endless population, which overstates the needed sample when your population is small. The correction divides the initial result by (1 plus (n0 minus 1) divided by N), shrinking the sample to reflect that testing part of a small group already tells you a lot about the rest.
6. Does the population size really matter that much?
It matters most for small and medium populations. For a population of 200, the correction can cut the sample nearly in half compared with the uncorrected figure. For a population of 100,000, the correction changes the answer by only a few percent. This is why auditors of very large populations often skip the correction entirely, while auditors of small populations should never skip it.
7. Should I round the sample size up or down?
Always round up. A sample size of 356.82 means 356 records are slightly too few to deliver the promised confidence and margin of error, so you test 357. This calculator rounds up automatically. Rounding down to save effort would quietly break the statistical guarantee behind your sample.
8. Can I use this calculator for surveys as well as audits?
Yes. The same math works for customer surveys, employee polls, quality inspections, and any situation where you sample from a known population. Just replace records with respondents: enter the total number of people you could survey, pick your confidence level and margin of error, and the calculator gives you how many responses you need.
9. What happens if I test fewer records than calculated?
Your results lose the statistical backing of your chosen confidence level and margin of error. With a smaller sample, the true margin of error grows wider than the number you entered, so your conclusions become less precise than planned. If practical limits force a smaller sample, document the limitation and acknowledge the reduced assurance in your audit report.
10. How do I select the actual records to test?
Use random selection so every record has an equal chance of being chosen, for example with a random number generator applied to a numbered list of records. Systematic selection, such as every 20th invoice, also works when the list has no hidden pattern. Avoid hand-picking records, because judgmental selection breaks the randomness the sample size math depends on.
11. Why does a 99% confidence level need so many more records?
Because the Z-score grows from 1.96 at 95% to 2.576 at 99%, and the sample size formula squares it. Squaring 2.576 gives 6.64, compared with 3.84 for 1.96, so the sample jumps by roughly 73%. That extra assurance is expensive in audit hours, which is why 99% confidence is reserved for the highest-risk work.
12. What if my population is smaller than the calculated sample?
That cannot happen with the finite population correction, because the corrected sample size is always smaller than the population. If you ever see a sample size larger than your population, the correction was not applied. In extreme cases where the calculated sample approaches the population size, it is simplest to test 100% of the records and skip sampling altogether.
13. Does the calculator work for attribute and variable sampling?
This calculator covers the general sample size formula used for proportions, which fits attribute sampling such as checking whether invoices are approved or not. Variable sampling, which estimates dollar amounts rather than yes-or-no outcomes, needs the population standard deviation as an extra input and a different formula, so this tool is not designed for it.
14. How does margin of error differ from confidence level?
The margin of error describes precision: a 5% margin means your sample result should land within 5 percentage points of the true value. The confidence level describes certainty: 95% confidence means that if you repeated the sampling many times, about 95 out of 100 samples would capture the true value within that margin. You need both numbers to define a sample, because precision without certainty, or certainty without precision, is incomplete.
15. Is testing extra records beyond the sample size a problem?
No. The calculated size is a minimum, so testing more records only improves your assurance. Extra testing is common when records are pulled in fixed batches or when a few selected records turn out to be unusable and need replacing. Just make sure the extra records are selected the same random way, not cherry-picked.
Final Thoughts
A defensible sample size is the foundation of every sampling-based audit. With the Audit Sample Size Calculator, you can set your population, confidence level, and margin of error, and get a statistically sound answer in seconds instead of wrestling with formulas by hand. Use it at the planning stage of every audit, document the inputs you chose and why, and select your records randomly. Do that, and your sample will stand up to review, your conclusions will carry real weight, and your audit hours will go exactly where they add the most value.