Inventory Sample Size Calculator
Counting every item in a warehouse takes days, and most audits do not need it. Enter your total stock, the confidence level you need and the margin of error you can accept, and this calculator tells you exactly how many items to count, with the finite population correction applied and every arithmetic step shown.
Assumes simple random sampling. Results are rounded up: you cannot count a fraction of an item, and rounding up keeps the promised confidence.
Statistical guidance for planning inventory audits. Valid results require genuinely random sampling; hand-picked or convenience samples do not carry the stated confidence. This is not professional audit advice.
- Quick Answer
- How to use the inventory sample size calculator
- The formula: Cochran’s proportion formula with a finite population correction
- Worked example: a 5,000-item warehouse at 95% confidence
- Choosing confidence, margin of error and p without guessing
- When sampling beats a full count, and when it does not
- What competing calculators miss
- Frequently asked questions
- Add This Calculator to Your Website
- References
Quick Answer
For a population of 5,000 items at 95% confidence, a 5% margin of error and maximum variability (p = 0.5), count 357 items (7.14%). The unadjusted formula gives 384.16; the finite population correction pulls it down to 356.82, rounded up to 357. The calculator above is pre-filled with this exact audit.
Key takeaways
- Population size matters less than people think. Past a few thousand items, the margin of error and confidence level drive the sample, not N.
- Use p = 0.5 when you have no prior audit. It is the conservative choice and keeps the audit valid whatever the true error rate turns out to be.
- The sample only carries its confidence if it is genuinely random. Audit sampling standards require every item to have a chance of selection.
- When the answer passes half the population, stop sampling and count everything. The saving no longer justifies the risk.
How to use the inventory sample size calculator
Four inputs set the whole calculation: population, confidence, margin of error and expected proportion. Results stay hidden until you click Calculate, and the working is printed under the result so an auditor can re-perform it.
- Population size (N): the total items in the group you are auditing. Count lines, units or locations, but stay consistent: the sample is drawn from exactly this population.
- Confidence level: 95% is the default for a reason. It is the level most audit plans and textbooks treat as standard evidence; 99% is for high-value or disputed stock, 90% for a quick internal check.
- Margin of error: entered as a percent, so 5 means plus or minus 5 percentage points on the error rate you observe. The tool accepts anything above 0 up to 50.
- Expected proportion (p): your estimate of the share of items that will disagree with the records. A decimal entry such as 5000.9 in the population field is read as 5,000, and p must sit strictly between 0 and 1, because a proportion of exactly 0 or 1 would claim there is nothing to measure.
- Read the result as a sentence: the tool states the count, its share of the population and what the confidence promises. After the physical count, turn the discrepancy you find into a loss figure with the inventory shrinkage percent calculator.
The formula: Cochran’s proportion formula with a finite population correction
The tool uses the standard two-step method from Cochran’s sampling framework: size the sample as if the population were infinite, then shrink it because your population is not. Both steps are printed in the result.
Step 2: n = n₀ ÷ (1 + (n₀ − 1) ÷ N), rounded up
Z = 1.645 (90%), 1.96 (95%) or 2.576 (99%) · p = expected proportion · e = margin of error as a decimal (5% = 0.05)
The Z values are the normal-distribution critical values the NIST/SEMATECH e-Handbook’s sample size guidance uses for these confidence levels, and the same values drive the interval you would report afterwards with a confidence interval around the observed rate. The correction in step 2 is what keeps small warehouses honest: ignore it, and a 200-item stockroom at 99% confidence and a 3% margin appears to need 1,844 counts from a population of 200. With the correction the tool asks for 181. At the 5% margin used in the worked example below, the same stockroom needs 154 of its 200 items, which is the tool quietly telling you to count everything instead.
One assumption deserves plain statement: this is simple random sampling of a proportion. It sizes a count aimed at an error or discrepancy rate. It is not monetary-unit sampling, stratified sampling or acceptance sampling for defects per batch, and no setting in the tool turns it into those.
Worked example: a 5,000-item warehouse at 95% confidence
The pre-filled run is a mid-size warehouse: 5,000 items, 95% confidence, 5% margin of error, p = 0.5. The tool returns 357 items, and every line below is a real run of the calculator above.
| Run (N = 5,000, p = 0.5 unless noted) | Unadjusted n₀ | Count | Share |
|---|---|---|---|
| Worked example: 95% confidence, 5% margin | 384.16 | 357 | 7.14% |
| Relaxed: 90% confidence, 5% margin | 270.60 | 257 | 5.14% |
| Strict: 99% confidence, 5% margin | 663.58 | 586 | 11.72% |
| Tighter margin: 95% confidence, 3% margin | 1,067.11 | 880 | 17.60% |
| Known error rate: 95%, 5% margin, p = 0.02 from last year’s audit | 30.12 | 30 | 0.60% |
The margin of error is where the count explodes. Holding the worked example at 95% confidence and p = 0.5, here is what each margin demands from the same 5,000 items:
Halving the margin roughly quadruples the unadjusted sample, because the margin sits squared in the denominator. That is the single most useful fact in audit planning: precision is bought at a steep, predictable price.
Choosing confidence, margin of error and p without guessing
Pick the loosest settings your decision can tolerate, and let prior audits set p. Each input trades accuracy against counting labour in a specific, knowable way.
Confidence. The Z-score is the price of certainty. Going from 95% to 99% multiplies the unadjusted sample by about 1.73 (663.58 against 384.16 at the same margin), because Z enters the formula squared. Reserve 99% for high-value stock, insurance disputes or a count someone may challenge.
Margin of error. Three to five percent suits most cycle-count programs. Go tighter only when a small error rate changes a real decision, such as whether a location passes an audit threshold expressed as an audit percentage of items checked.
Expected proportion. The p × (1 − p) term peaks at p = 0.5, which is why 0.5 is the safe default. A credible prior estimate shrinks the job dramatically: at p = 0.02 the worked example drops from 357 items to 30. The risk is symmetric: if the true discrepancy rate is far from your p, the achieved precision is worse than promised, so only use a small p when a recent, comparable audit supports it.
Stratify when the stock is mixed. One blended sample treats a pallet of phone chargers and a cage of laptops as the same population. Splitting high-value and low-value groups and sizing each stratum with this tool gives tighter control where the money is, a point audit sampling standards make when they discuss stratification reducing overall sample sizes.
When sampling beats a full count, and when it does not
Sampling wins on large, uniform populations; a full count wins when the required sample is a large share of a small population, or when every item is individually valuable. The tool itself signals the crossover: watch the sampling percentage in the result.
Sample when
- The population runs into the thousands and items are broadly similar.
- You need a defensible error rate for an audit plan, not a recount of record.
- Counting everything would shut the warehouse or blow the labour budget.
- Prior audits give you a p you can stand behind.
Count everything when
- The required sample passes roughly half the population, as in the 200-item stockroom needing 154.
- Items are high-value or serialised, so each one needs its own verification.
- You suspect fraud in a specific area; a targeted count beats a random one there.
- The count feeds a legal, insurance or sale-of-business figure.
Whichever route you choose, compare the physical count with the book figure afterwards. A count that disagrees with the records splits into record error and real loss, and the inventory variance calculator separates those two stories.
What competing calculators miss
The formula itself is standard, so rival tools differ on what they hide, skip or leave you to work out:
- CalculatorsHub’s inventory tool: the closest inventory-specific rival, and it adds a sampling interval (count every nth item), which is genuinely useful. But its worked example for 12,000 items lands on ≈372, rounding 372.27 down instead of up: run the same inputs here and the answer is 373, because a sample size is a minimum, not an approximation. Its reference table is rounded the same loose way.
- BraveCalculator: states the adjustment in the equivalent form (n × N) ÷ (n + N − 1), then works an example that stops after the unadjusted step, so a reader cannot see the correction applied. It also omits the expected proportion from its steps even though its own formula needs it.
- SavvyCalculator: its worked example at p = 0.1 for 10,000 items returns 139 units. Run the same inputs here and the answer is 137: the unadjusted figure is 138.30 and the population correction pulls it to 136.42, rounded up. Their 139 looks like the unadjusted number rounded up with the correction never applied.
- Generic survey tools (RoughTools, SuperSurvey, Macaron): correct proportion math aimed at survey responses, with response-rate and subgroup machinery an inventory auditor does not need, and no acknowledgement that the population here is physical stock with a hard counting cost.
- The gap this page fills: the correction applied and printed step by step for your exact inputs, the sampling percentage stated as a share of stock, an explicit warning when the sample passes half the population, and honest limits: random sampling only, no monetary-unit or stratified sizing built in.
Frequently asked questions
It depends on your population size, confidence level, margin of error, and expected proportion. For 5,000 items at 95% confidence, 5% margin of error and p = 0.5, the answer is 357 items (7.14% of the population). Enter your own numbers in the calculator above.
First compute the unadjusted size n₀ = (Z² × p × (1 − p)) ÷ e², then apply the finite population correction n = n₀ ÷ (1 + (n₀ − 1) ÷ N) and round up. The correction matters when the sample is a noticeable share of the population.
Use 0.5 unless you have a solid prior estimate from a previous audit. Variability peaks at p = 0.5, so it produces the largest and safest sample. If last year’s count found discrepancies in about 2% of items, p = 0.02 gives a much smaller sample that is still valid for that error rate.
Because you cannot count a fraction of an item, and rounding down would leave the sample slightly too small to support the confidence level you asked for. Rounding up is the standard, conservative choice in sampling practice.
Yes. The confidence statement only holds when every item has a known, non-zero chance of selection, as audit sampling standards require. Counting the easiest shelf to reach, or only items you already suspect, breaks the math no matter how large the sample is.
Three to five percent is typical for inventory audits. Tightening the margin gets expensive fast: at 95% confidence with p = 0.5 in a 5,000-item warehouse, a 5% margin needs 357 items, a 3% margin needs 880, and a 1% margin needs 3,289.
More certainty requires more evidence. Moving from 95% to 99% confidence raises the Z-score from 1.96 to 2.576, which lifts the unadjusted sample by about 73% before the population correction. In the 5,000-item example the final count rises from 357 to 586.
When the population is small or the required sample is a large share of it. In a 200-item stockroom at 99% confidence and a 5% margin, the tool asks for 154 of 200 items (77%), at which point a full count is simpler and removes sampling risk entirely.
Add This Calculator to Your Website
Embed this calculator on your own site with the code below. Pick a size to preview it live, then copy the code.
Powered by Calculator Zilo
References
- National Institute of Standards and Technology, NIST/SEMATECH e-Handbook of Statistical Methods. Sample sizes required. Accessed October 22, 2026.
- International Auditing and Assurance Standards Board (IAASB). ISA 530: Audit Sampling. Accessed October 22, 2026.
- Cochran, W. G. Sampling Techniques, 3rd edition (Wiley). Accessed October 22, 2026.
- Krejcie, R. V., and Morgan, D. W. Determining Sample Size for Research Activities, Educational and Psychological Measurement 30(3). Accessed October 22, 2026.
- U.S. Government Accountability Office. Using Statistical Sampling (AIMD-10.1.8). Accessed October 22, 2026.