How to run a stock count: a step-by-step procedure for stores and warehouses
Who it is forGrocery chain branch managers, area managers, warehouse managers, finance and internal audit
In most businesses a stock count is treated as a once-a-year formality done because the accountant asked for it. In fact the count is what makes the income statement true: if closing inventory is wrong, so are cost of goods sold and gross profit. Beyond that, the count is the only moment when process failures become visible. This guide covers the count in three stages and explains why moving to cycle counting produces better results.
Preparation before the count
The quality of a count is largely determined before it starts. A count that begins unprepared is equivalent to one never done, because its results cannot be trusted — and days have been spent.
The most critical preparation is the freeze. If stock keeps moving during the count, the same goods get counted twice or not at all. Goods receipt, dispatch and transfers stop; if selling continues, sales are recorded separately and deducted from the result.
The second critical preparation is zoning. The warehouse and shop floor are divided into zones with physically clear boundaries and each zone is assigned to one team. Where the boundary is ambiguous, the space between two teams is either counted twice or missed entirely.
- Announce the date and form the teamsAt least a week's notice. Each team is two people: one reads, one checks. Teams do not count the zones they are responsible for.
- Divide and label the areaEvery shelf, aisle and storage area is numbered and the count sheet is produced against those numbers. A team does not move on until its zone is complete.
- Make the goods countableMixed cartons are separated, unlabelled items are labelled, and the same product stored in several places is consolidated. Leave this to the day of the count and the count takes twice as long.
- Freeze stock movementsGoods receipt and dispatch stop. Deliveries arriving on the day are not received; they are held in a separate area and excluded from the count.
- Take a system snapshotThe recorded stock at the start of the count is frozen and saved. Reconciliation is done against that snapshot, not against the live record at the end.
- Prepare the devicesHandhelds are charged, the count list is loaded and offline operation is tested. Spare devices are kept on hand.
Refusing deliveries on count day is a difficult decision in most businesses. The alternative is to hold arriving goods in a separate area outside the count and receive them once the count is finished; that preserves the freeze without stopping the operation.
How to count: the double-count rule
One count is not enough. In a zone counted once by one person, the counting error can be larger than the variance the count is trying to correct.
The standard method is the double count: every zone is counted by two different teams, independently. The second team does not see the first team's result. If the two agree the result is accepted; if not, a third count is run and the difference is resolved on the spot.
Where handhelds are used, barcode scanning should be mandatory. Typing a quantity is faster but leaves the wrong-product error completely open. If the handheld does not recognise a scanned barcode it must not halt the count; the item is recorded as "unidentified" and handled in reconciliation.
- The second count is made without sight of the first count's result
- Teams do not count the zones they are responsible for
- Barcode scanning mandatory; manual quantity entry only with a stated reason
- An unidentified barcode does not stop the count; it goes onto a separate list
- When a zone is complete the team locks the result and cannot go back and change it
Handhelds must work offline. Network drops inside a warehouse are common, and a device that stops mid-count puts everything counted so far at risk. Records are held on the device and synchronised when the connection returns.
Reconciliation: before you close the variance
The work is not over when the counting stops; the real value emerges in reconciliation. If the difference between record and count is simply written into the system, nothing has been learned.
Variances are ranked from largest to smallest and the top 20 per cent by value is examined one by one. That portion typically accounts for 80 per cent of the total variance; examining the remaining small differences returns nothing.
A reason is written for every large variance. Variances with no identifiable cause are flagged as "unknown" and that group's share of the total is tracked; growth in that share is a direct indicator of process failure.
- Rank the variance by valueValue matters, not unit count. A thousand-unit variance on carrier bags is smaller than a three-carton variance on cigarettes.
- Verify the top 20 per cent physicallyFor large variances the product is physically re-checked. A significant share of variances are counting errors and are resolved at this step.
- Examine the movement historyFor the remainder, recent goods receipts, transfers and returns are checked. Goods received under the wrong barcode produce equal and opposite variances on two products.
- Code the reasonCounting error, recording error, shrinkage, damage, theft, transfer error, unknown. The code list should be short.
- Post the adjustment in one goApproved variances are posted as a single count document. Piecemeal adjustments lose track of which correction belongs to what.
Full count or cycle counting
The annual full count is a statutory requirement but a weak instrument for stock accuracy. Errors arising during the year go unnoticed for months, and by the time they are corrected the cause can no longer be found.
Under cycle counting, a small group of products is counted every day and the whole catalogue is counted more than once over the year. Products are counted at different frequencies by ABC class: A monthly, B quarterly, C annually.
The biggest advantage of cycle counting is timing: a variance is found within days of occurring, while the cause is still remembered. The second advantage is operational load; it runs without closing the store or working outside hours.
| Full count | Cycle counting | |
|---|---|---|
| Frequency | 1-2 times a year | A small group every day |
| Operational disruption | 1-3 days closed | None |
| Time to detect an error | Up to 6 months | A few days |
| Chance of finding the cause | Low — nobody remembers | High — the movement is recent |
| Staff load | Intense, in one block | Spread, 30-60 minutes a day |
| Statutory standing | Required at period end | Supporting; may not suffice alone |
| Stock accuracy | 92% - 96% | 97% - 99.5% |
The two are not alternatives. Cycle counting keeps accuracy high through the year; the period-end full count satisfies the statutory requirement and validates the cycle counting itself.
Frequently asked questions
- How is a stock count done?
- In three stages: preparation, counting and reconciliation. In preparation stock movements are frozen and the area is divided into zones. In counting each zone is counted independently by two teams. In reconciliation record and count are compared and no variance is closed without a cause.
- When should a stock count be done?
- A full count is required at period end for statutory purposes. For operational accuracy, continuous cycle counting is preferred: A-class products monthly, B quarterly, C annually. Choose the lowest-trading day for the count.
- What are the rules to follow during a stock count?
- Stock movements are frozen, the area is zoned, each zone is counted independently by two teams, teams do not count their own zone, barcode scanning is mandatory and the system snapshot at the start of the count is saved. Reconciliation is done against that snapshot.
- Can selling continue during a count?
- It can, provided the goods sold are recorded separately and deducted from the count result. The cleanest approach is counting outside trading hours. Goods receipt and dispatch, however, must stop; arriving goods are held in a separate area.
- What is cycle counting?
- Counting the whole stock in small daily groups rather than all at once. Products are counted at different frequencies by ABC class. It does not interrupt operations and catches errors within days while the cause can still be found; stock accuracy typically reaches 97-99.5 per cent.
- Who should carry out the count?
- Not the person responsible for the area being counted. The warehouse manager does not count their own warehouse and the department manager does not count their own department. Independence is the basic condition for a count to be credible.