Stock mismatches: cycle counts, recounts and adjustments
By the OrbitPi product team · Updated · 5 min read
A stock mismatch is a difference between the quantity counted and the book quantity for the same item, lot and location. Freeze the book figure before counting, so receipts during the count are not mixed into the snapshot. A blind count hides that figure from the counter. Recount a large difference before you change the books, and post an adjustment only after the count is approved. The reason on the adjustment is what later explains the difference.
Key takeaways
- Freeze book quantity, including cost, before the first count is entered.
- A blind count does not show the book figure, the movements after the freeze, or the variance until the count is submitted.
- Variance is counted quantity minus the frozen book quantity, adjusted for stock that moved at that location after the snapshot.
- A line inside tolerance can post with the count. A line outside tolerance waits for approval.
- A recount keeps the original line and adds a new attempt. It does not overwrite the first count.
- The posted difference uses a count-variance reason, or COUNT_OVER / COUNT_SHORT when no reason is chosen.
What is a stock mismatch?
A mismatch is not a missing pallet by itself. It is a counted quantity that does not equal the book quantity for one stock identity: the item, its lot or serial when you track one, the location, the owner and the stock status. Mixing two lots on one line, or counting a bin that was received after you started, produces a difference that is not a real loss.
Write the difference down before anyone “corrects” the system from memory. The book figure you compare against has to be the figure at a named time, not whatever the screen shows an hour later.
Why freeze the book figure, and when to count blind?
Freeze means storing the book quantity and its unit cost on the count line at the start. Later receipts, picks and transfers are then visible as movements after that snapshot, instead of silently changing the number the counter is judged against.
A blind count is the same freeze, with the book quantity hidden from the person counting until the count is submitted. Counters who can see “the system says 40” tend to find 40. Hiding the figure does not hide it from the record: it is stored, and it is used when the variance is calculated.
How should a variance be classified?
Calculate variance as counted quantity minus the frozen book quantity, then adjust for stock that actually moved at that line after the snapshot and before the count was entered. Do not subtract the live on-hand balance, because that balance already includes those movements and the count’s own activity.
| Step | Quantity | What it means |
|---|---|---|
| Book quantity at freeze | 100 | Stored on the count line |
| Receipt after the freeze | 10 | Movement in the window, not part of the frozen book |
| Quantity the counter should tie to | 110 | 100 + 10 |
| Quantity counted | 104 | What was on the shelf |
| Variance | −6 | 104 − 110, a shortage |
Give the shortage a reason from the count-variance list, such as damage, a picking error or an unidentified loss, and add a note when the reason requires one. If nobody chooses a reason, record the sign: short or over. A reason is what an auditor reads later. “Adjusted” is not a reason.
When do you recount, and when do you post?
- Set a tolerance in quantity and in value before the count starts. A line inside both can be posted with the count.
- A line outside either tolerance holds the whole count for approval. Nothing is written to on-hand yet.
- Recount lines that miss the recount threshold. Keep the first count line and store the new attempt beside it.
- Approve, then post. Posting creates the stock adjustment. Cancelling the count posts nothing and releases the frozen locations.
- The person who counted should not be the only person who approves a difference outside tolerance.
Companies covered by CARO 2020 also have to be able to say whether inventory was physically verified at reasonable intervals and whether a discrepancy of 10% or more, in aggregate for a class of inventory, was dealt with in the books. A count that was never posted, or a posted adjustment with no reason, does not answer that question. Confirm the current clause against the Ministry of Corporate Affairs text before you rely on a percentage.
How often should you cycle count?
A cycle count is a repeating programme, not a once-a-year wall-to-wall stocktake. High-value or fast-moving items are usually counted more often than slow bulk. The programme should name its scope (ABC class, zone, item, or a location that just went empty), how often it runs, whether counts are blind, and the tolerance that forces approval.
An empty-bin check is a useful extra: when the last pick takes a location to zero, count that location while it is empty. A quantity found there is a mismatch you would otherwise carry until the next programme date.
General information, not professional advice. This guide summarises the rules as we understand them on the date shown. Laws, notifications and thresholds change, and their application depends on your facts. Check the official sources below and consult a chartered accountant or tax adviser before acting on it.
How OrbitPi Inventory helps
OrbitPi Inventory freezes book quantity and unit cost on every count line, including a blind count, and withholds the book figure until submit while the count is blind. A line outside the quantity or value tolerance sends the whole count for approval. A recount supersedes the original line instead of replacing it. Posting writes the variance through a stock adjustment. The line uses an active count-variance reason when one is given, otherwise COUNT_OVER or COUNT_SHORT by sign.
Explore OrbitPi Inventory