Stock record accuracy: how to measure it and improve it
What stock record accuracy means, how to calculate it from a count, how to choose a tolerance, and what to do about the differences you find.
What stock record accuracy measures
Stock record accuracy compares what your system says you have with what you actually count. The most useful version works item by item: for each item counted, is the system record right or not? The result is the share of item records that are right.
Stock record accuracy (%) = items within tolerance ÷ items counted × 100
It is deliberately strict. An item that is out by one unit and an item that is out by a hundred both count as one inaccurate record, because either can cause a wrong pick, a missed order or a wrong purchase.
A worked example
| Item | System | Counted | Difference |
|---|---|---|---|
| A-101 | 240 | 240 | 0 |
| A-102 | 180 | 176 | −4 (−2.2%) |
| B-210 | 60 | 61 | +1 (+1.7%) |
| C-330 | 25 | 22 | −3 (−12.0%) |
| D-015 | 120 | 120 | 0 |
With no tolerance, 2 of 5 records are right: 40%. With a 2% tolerance, B-210 is also accepted and the figure is 60%. The overs and shorts in units do not cancel each other out: the record is either right or it is not.
Choosing a tolerance
A tolerance lets small differences count as accurate. There is no universal figure, and it should be a written policy, not something decided during the count. A common approach is no tolerance for high-value, controlled or serial-tracked items, and a small percentage for bulk, low-value items where exact counting is impractical, such as loose fixings or items counted by weight.
Why differences happen
Before adjusting stock, find the cause. In practice many differences are not loss at all:
- Receipts, issues or transfers not yet posted when the count was taken.
- Wrong unit of measure, for example cartons counted but pieces recorded.
- Stock put away in the wrong location, so one bin is short and another is over.
- Mixed batches or wrong item codes on labels.
- Damaged or expired stock removed but not written off in the system.
Recount the largest differences first, check recent transactions, then adjust only what is confirmed, with approval.
Cycle counting
Instead of counting everything once a year, cycle counting checks a small part of the stock regularly. Many sites count fast-moving and high-value items more often and slow, low-value items less often. Tracking accuracy after each cycle shows whether processes are improving, and points to the areas or processes that cause the errors.
Last updated: September 2026. All guides