Cycle Count Sample Size
“Count 10% of locations” is not a method, it is a habit. This works out how many locations you actually need to count to state your inventory accuracy with a given confidence — and what that costs in man-hours.
Your data never leaves your browser — everything is calculated on your device.The population you want to make a statement about — usually active pick and bulk locations.
A ±2% margin at 95% confidence means: if you measure 97% accuracy, the true figure is between 95% and 99%, and you would be right 19 times out of 20.
Your best estimate of current accuracy. Sample size peaks at 50% and shrinks as accuracy rises — a good warehouse needs to count less to prove it.
A statistical sample proves accuracy. An ABC plan keeps it accurate. Most operations need both — this is the control side.
The formula
The first line gives the sample you would need from an infinite population. The second corrects it for the fact that your warehouse is not infinite — and the correction matters enormously. For a 12,000-location warehouse at 95% confidence and ±2%, the uncorrected figure is around 369; corrected it drops to about 358. For a 500-location warehouse it falls from 369 to about 213, which is the difference between a sensible programme and an impossible one.
Two different jobs, often confused
A statistical sample answers “how accurate is my inventory?” — you count a random sample and state a number with confidence bounds. An ABC cycle count programme answers “how do I keep it accurate?” — you count high-value and fast-moving items often and the tail rarely, and you fix errors as you find them. They are not substitutes. Use the sample quarterly to report accuracy; run the ABC plan continuously to maintain it.
Common mistakes
- Counting the easy locations. A sample is only valid if it is random. Counting whatever is convenient at 4 pm on a Friday produces a number that means nothing, however many locations it covers.
- Counting what the system says is there. Blind counts only — the counter should not see the expected quantity. Showing it turns counting into confirming, and error rates collapse to an implausible zero.
- Adjusting without finding the cause. An adjustment closes the variance and loses the information. Every discrepancy above a threshold deserves a root cause: wrong pick, wrong put-away, unrecorded damage, or a receipt never posted.
- Measuring accuracy by value instead of by location. Both are valid, but they answer different questions and are not comparable. Pick one, define it, and keep it.
- Treating the annual wall-to-wall as the real count. A well-run cycle count programme is more accurate than an annual full count, because it is done by people who are not exhausted and not counting against a deadline.
Get the free KPI starter pack
A ready-to-use Excel sheet with the 15 warehouse KPIs worth tracking, the formulas behind each, and a monthly review format. No cost, no catch.
One email with the file. No newsletter spam, unsubscribe any time.