Cycle counting: how to stop shutting the store for stocktake

The annual stocktake is one of those rituals that survives because it has always happened. The store closes, people who do not normally count things count things, a large number of discrepancies appear, and the balances are overwritten to match. Everyone goes home.

The problem is not the effort. It is that the output is unusable. You now know that eleven months of unexplained differences exist and no way to attribute any of them, so nothing gets fixed and the same thing happens next year.

Count little and often instead

Cycle counting means counting a small number of lines continuously — say twenty a week — so that everything is counted at least once a year, and important things far more often. It takes about twenty minutes, nothing closes, and crucially the errors you find are recent enough to explain.

A discrepancy found within a fortnight has a cause someone can still remember. The same discrepancy found eleven months later has only a number.

How often to count what

Not everything deserves the same attention. The usual approach is an ABC split by annual value — usage multiplied by unit cost — because that is where the money and the disruption concentrate.

Two sensible amendments in a maintenance or engineering store. Anything statutory — lifting gear, calibrated instruments, PPE — goes in A regardless of value, because being wrong about it has consequences money does not measure. And anything with a history of discrepancies goes up a class until it behaves, since past drift is the best predictor of future drift.

Doing it properly

1. Count blind. Do not show the expected figure. If the counter can see it says 46, some of them will find 46. This single rule changes more than any other.
2. Count at a quiet moment, or freeze the bin while counting. A count taken while someone is drawing from the same shelf is not a discrepancy, it is a race condition.
3. Recount before adjusting. Most first-count differences on small items are counting errors, and adjusting on a single count teaches people the record is soft.
4. Record the difference as a correction, not an overwrite. You want the history to show that on 14 August the figure moved from 46 to 43, and who counted it.
5. Ask why, on anything material. Not to assign blame — to find the cause. It is almost always one of a short list.

What the differences usually mean

• Someone took stock without recording it — the most common by far, and a signal that recording costs too much.
• The unit of issue is ambiguous. A box of 100 issued as "1" by some people and "100" by others will produce spectacular drift.
• Goods in were never booked, or were booked twice.
• Two similar items are being confused at the shelf, which shows up as a mirrored pair: one short, one long.
• Breakage or waste that nobody wrote off, because there was no obvious way to.

That last one is worth designing for. If there is no easy way to record "I dropped it", the difference appears as a mystery instead of a fact.

Measuring whether it is working

Track the percentage of counted lines that were exactly right. Not the total value of adjustments — a single expensive item swamps that and hides a store full of small chronic errors.

A store that has never done this typically starts somewhere between 60% and 80%. Above 95% is good. Above 98% usually means recording happens at the point of issue rather than afterwards, which is the actual cause rather than the counting.

Starting next week

Pick your top twenty lines by value. Count them blind on Monday morning. Record the differences as corrections. Look at the five biggest and ask what happened.

That is the whole method, and it will tell you more in an hour than the last three annual stocktakes combined — because for the first time the answers are recent enough that somebody still remembers them.