The annual stock-take is a costly ritual: two days of shutdown, the whole company mobilised, and at the end a global discrepancy adjusted without knowing where it comes from. Cycle counting produces a more reliable stock without ever closing the shop floor. It just has to be structured properly.
Why the annual count fixes nothing
A discrepancy found in December on an item consumed all year long is useless: it may come from a keying error in March, an undeclared scrap in June or an issue without a note in September. You adjust the quantity, post the accounting entry, and the cause stays intact — it will reproduce the same discrepancy next year. The annual count measures; it does not correct.
Classify before counting
The principle is the same as ABC analysis, applied to counting frequency rather than value alone. A breakdown that works in most shops:
- Class A — the items that carry most of the value, or that stop production if they run out: monthly count.
- Class B — intermediate value, regular turnover: quarterly count.
- Class C — fasteners, consumables, small tooling: annual count, or even by sampling.
The daily volume then becomes trivial: a few dozen items a day, absorbed by the storekeeper at the start of the shift, without interrupting anything.
Count at the right moment
One rule saves the reliability of the exercise: count a location when no movement is in progress on it. In practice this means freezing the item for the duration of the count, or counting before the shop opens. A count taken while a forklift operator is picking from the same bin produces an artificial discrepancy, which lastingly discredits the approach with the teams.
The discrepancy is not the result: it is the starting point
This is where the whole value of a short interval lies. A discrepancy found on an item counted thirty days ago ties back to a short window, where you can re-read the movements one by one. You then identify the real cause: an item issued without a note, a mis-set packing unit, two visually similar items confused at picking, a production return never declared. Fixing the cause removes the discrepancy for good.
Track two indicators, and only two: the reliability rate by class — the percentage of items counted without discrepancy — and the number of causes handled in the month. The first should rise, the second should fall. If the first stalls below 95% in class A, the problem is not the counting but the discipline of recording movements.
What the auditor expects
Perpetual inventory is accepted in place of the annual count, provided you can demonstrate that each item was counted at least once during the financial year, that discrepancies were traced and justified, and that the procedure is written and applied. An ERP that timestamps each counting session, keeps the initial discrepancy and its adjustment, and can print the year's coverage by item, provides that file with no extra work.