Receiving 101 · Stock
A cycle count is a smoke alarm, not a fire door.
It tells you a balance is wrong. It cannot tell you when it went wrong, or why, and it finds most errors weeks after the moment that created them. Counting well is worth doing; knowing what the count is for is worth more.
In short
- Count a slice often rather than everything once. Nothing shuts down, and a small difference can still be traced to a cause.
- Measure accuracy by location or line. By value, a warehouse can score 99 per cent while being 80 per cent wrong.
- Count blind. A sheet showing the expected quantity produces agreement rather than a count.
- Most of what a count finds was created at goods-in, and by then the delivery is long gone.
Why not just count everything once a year
A full physical inventory has two problems and only one of them is the shutdown. It gives you a single accurate day out of three hundred and sixty five, and it tells you nothing about how the errors arrived, because a year’s worth of causes have been mixed together by the time you find the difference.
Cycle counting spreads the same work across the year in small pieces. Nothing stops, and each count covers a short interval since that item was last right, which is the property that actually matters: a difference found across three weeks can be traced to something that happened in those three weeks. A difference found across a year cannot be traced to anything.
That is the argument for it, and it is worth stating plainly because it is not the one usually made. The benefit is the short interval, which is what makes a difference explicable. Doing the work in smaller pieces is incidental.
What to count, and how often
The usual method ranks items by annual value into three groups and counts the top one most often. It is a reasonable default and it has one blind spot worth correcting: value is not the only thing that makes an item likely to be wrong.
Frequency of handling matters at least as much. An inexpensive fastener picked forty times a day has forty chances a day to be miscounted, put in the wrong bin or taken without a record. An expensive casting touched twice a year has two. Ranking by value alone counts the casting monthly and the fastener never, which is precisely backwards for finding process faults.
So rank by value, then weight by movement, and keep a small control group of items counted very frequently regardless of either. The control group exists to tell you quickly whether something in the process has changed, which is a different job from measuring how accurate the warehouse is.
The measurement that flatters, and the one that helps
Inventory accuracy gets reported as a percentage and the percentage depends entirely on what you divide by. Both of these are true of the same warehouse on the same day:
By value
Reassuring, and nearly useless
A handful of expensive items, counted most often because they are expensive, dominate the total. The cheap items where the process actually fails contribute almost nothing to the number.
By location or line
Uncomfortable, and actionable
Every counted location scores the same whatever is in it. The number is lower and it moves when you fix something, which is the only property a metric needs.
Report both if you must, but manage on the second. A team told it is at 99 per cent has no reason to change anything, and a team told it is at 80 per cent by line has somewhere to start.
Count blind, and keep the difference
Two habits do more for the quality of a count than any amount of scheduling.
The first is counting blind: the counter should not be able to see the expected quantity. A sheet showing the system balance does not produce dishonesty, it produces agreement, and the effect is strongest exactly where you least want it, on the tedious high-count items nobody enjoys counting twice.
The second is keeping the difference rather than only the adjustment. An adjustment tells you the balance changed. The difference, with a date on it, tells you that this item has been short four quarters running, which is a different fact and a far more useful one. Under ISO 9001 clause 7.5.3 the count records are documented information to be controlled and retained anyway, so the discipline costs nothing you were not already meant to be doing.
The part this company exists for
Most of what a count finds was born at goods-in
Cycle counting is an inventory practice, and the errors it finds are mostly not inventory errors. In what we see, the common causes are all upstream of the shelf:
- 01
A delivery received against the wrong order
Two orders open with the same supplier, no order number on the paperwork, and the receiver matched by eye. One line closes short and another closes over, and both look normal until something counts them.
- 02
A quantity copied rather than counted
The receipt was made by typing the supplier's packing slip into the system rather than counting what was in the cartons. The result agrees with the supplier by construction, because it was copied from them.
- 03
A unit of measure difference
Ordered in cases, received in eaches, or the case pack changed and nobody said. The classic silent error, because every document involved is internally consistent.
- 04
A part delivery closed as complete
Matched against the order rather than the order line, so the balance still owed disappears from view. The shortage surfaces months later on an item nobody was watching.
Every one of those is created in a few minutes at a loading bay and found months later by somebody with a scanner. By then the pallet is gone, the driver is gone, and the supplier’s records say the goods were signed for.
There is a legal edge to that delay as well. U.C.C. § 2-607(3)(a) bars a buyer from a remedy unless they notify the seller within a reasonable time after they discover or should have discovered the breach. A shortage found by a cycle count six months later is one a buyer should have discovered at the bay, and “our cycle count only reached that item in March” is not an answer to it.
So the cheapest cycle count is the one that finds nothing, because the receipt was right. Matching a delivery against the order rather than against the supplier’s own paperwork removes the first two causes on that list outright, and a unit of measure on every order line removes the third.
Questions
- What is cycle counting?
- Counting a small part of your stock frequently rather than all of it once a year. Each count covers a handful of items or locations, the result is compared to the system balance, and differences are investigated and corrected. Over a period every item gets counted, usually more often for the items that matter most.
- How is cycle counting different from a physical inventory?
- A full physical inventory counts everything at once and usually needs the operation stopped to do it. It produces one accurate day a year and tells you nothing about how the errors got there. Cycle counting spreads the work across the year, so nothing shuts down, and because each count covers a short period since the last one, the difference it finds is small enough to be traced back to a cause.
- How often should I cycle count?
- Frequency should follow consequence rather than volume. A common approach ranks items by annual value, counts the top group monthly or quarterly, the middle group a few times a year and the tail annually. Velocity is worth weighting alongside value: an inexpensive item picked forty times a day has far more chances to go wrong than an expensive one touched twice a year.
- What counts as an accurate count?
- Count accuracy is best measured by location or by item, not by value. A warehouse that is 99 per cent accurate by value can be 80 per cent accurate by line, because the expensive items are counted most often and the cheap ones absorb the errors. Measuring by value flatters the number and hides the thing you want to fix.
- Should I use a tolerance on cycle counts?
- A small one on low-value, high-count items, and none on anything serialised, controlled or expensive. What matters more than the threshold is that a difference inside tolerance is still recorded rather than silently adjusted, because a pattern of small differences on one item is a process fault and a single large one is usually an event.
- Why does the same item keep coming up wrong?
- Because a cycle count corrects a balance, and correcting a balance does nothing about the process that broke it. If an item is short every quarter, the count is not the control that is failing. Look at how it is received, whether its unit of measure is consistent between the order and the shelf, and whether anybody records the movement when it is taken for a job.
- Where do stock errors actually come from?
- In our experience most of them are created at goods-in rather than on the shelf: a delivery received against the wrong order, a quantity taken from the supplier's packing slip rather than counted, a unit of measure difference between cases and eaches, or a part delivery closed as complete. Picking and put-away contribute too, but those errors are usually found quickly. A receiving error is invisible until something counts it.
- Does any standard require cycle counting?
- No standard or regulation prescribes it as a method. What ISO 9001 asks for, in clause 8.5.2, is that outputs are identified and traceable, and in 7.5.3 that the records are controlled and retained. Cycle counting is one way to show your records are accurate. An auditor is interested in the evidence rather than the technique.
Sources
- 01ISO 9001:2015, clause 8.5.2: identification and traceability · International Organization for StandardizationRequires outputs to be identified and, where traceability is a requirement, the unique identification to be controlled and retained. It is what an auditor asks stock records to demonstrate; it does not prescribe cycle counting or any other method of checking them.
- 02ISO 9001:2015, clause 7.5.3: control of documented information · International Organization for StandardizationDocumented information must be available, protected and retained for a defined period. The count sheets and the adjustments made from them are documented information, and an adjustment with no record of who made it and why satisfies none of it.
- 03U.C.C. § 2-607: Effect of Acceptance; Notice of Breach · Legal Information Institute, Cornell Law SchoolSubsection (3)(a) bars a remedy unless the buyer notifies the seller within a reasonable time after they discover OR SHOULD HAVE DISCOVERED the breach. A short delivery found by a cycle count six months later is one a buyer should have discovered at the bay.
Worth saying plainly: no standard or regulation prescribes cycle counting. The clauses above are what a quality system asks stock records to demonstrate, and counting is one way of demonstrating it. Anyone telling you a standard requires a particular counting frequency is describing their own policy.
This describes how the rules generally work in the United States and links to the primary sources so you can read them yourself. It is not legal advice, and your contract with a supplier or carrier can change any of it. For money that matters, ask a lawyer.
A count tells you the balance is wrong. A receiving record tells you when it went wrong and who to ask. DockCount checks the packing slip against the purchase order at the bay, line by line, and keeps the photographs. Free for 25 deliveries a month.