Which stock are you trying to see?
At opening time, a store manager may need three different answers: how much stock is physically present, how much is sellable, and how much is actually on the shelf. These are not interchangeable. A reserved unit may be on site but unavailable to a walk-in customer. A damaged unit may still be physically present while correctly excluded from sale. A carton in the stockroom may not help a customer facing an empty display.
Start by defining the location, condition and commitment rules for each balance. Automated stock monitoring is useful when it gathers relevant events and highlights exceptions. It is less useful when one apparently precise number hides several incompatible definitions.
Follow the event trail
A stock ledger normally starts from an opening balance and applies receipts, sales, returns, transfers and adjustments. Each movement needs an item identifier, quantity, unit, location and time. The person or system responsible should also be traceable. A current balance without its event trail makes investigation unnecessarily difficult.
At receiving, compare what arrived with what was ordered, not simply what the advance shipping notice predicted. A sealed carton can contain a short quantity. A supplier substitution can carry a different identifier. Recording the expected delivery as an actual receipt can turn a purchasing document into phantom stock.
Use identifiers without overclaiming
GS1 defines the Global Trade Item Number, or GTIN, for trade-item identification. A Global Location Number, or GLN, can identify parties and locations. A Serial Shipping Container Code, or SSCC, identifies a logistics unit. These identifiers solve different problems; a pallet identifier is not automatically a product identifier.
GS1 EPCIS provides a standard for sharing visibility event data. It can support a more consistent account of what happened, where and when, but adopting a standard does not guarantee accurate capture. Barcode readers and radio-frequency identification, or RFID, still depend on correct labels, appropriate equipment and disciplined handling. Validate read coverage in the actual environment before treating an automated observation as a complete physical count.
Investigate a discrepancy before correcting it
Worked example A store record shows 18 units, but a shelf count finds 7. A stockroom check finds 6 more, and 2 units are held for collection. That explains 15 physical units, not 18. The remaining three-unit difference needs investigation; it is not evidence by itself of theft, a scanning error or a supplier shortage.
| Observation | Units | Next check |
|---|---|---|
| Recorded physical stock | 18 | Confirm balance definition |
| Shelf and stockroom | 13 | Inspect other locations |
| Collection holding area | 2 | Confirm reservation status |
| Unexplained difference | 3 | Review events and recount |
Inspect recent receipts, transfers, voided sales and returns. Check whether a return was recorded before inspection, whether a transfer was dispatched but never received, and whether a case was confused with an individual unit. If a correction is necessary, preserve the original evidence and apply an explicit reason code. Avoid silently rewriting the history.
Count where the risk is concentrated
Cycle counting means checking selected inventory regularly rather than relying only on an occasional complete count. ABC analysis can help prioritise economically significant items, but value alone is not enough. A low-value item with repeated availability failures may deserve more attention than an expensive item with stable handling.
Schedule counts around movement controls. A count performed while replenishment and customer picking continue can compare different moments rather than different realities. Record the cutoff time, pending transactions and any temporary handling restrictions. For repeated discrepancies, fix the receiving, picking or returns process that creates them.
Make alerts actionable
Start with a small exception queue: negative sellable stock, a stale transfer, an unusual adjustment or a shelf gap despite recorded stock. Assign an owner and a response window appropriate to the trading context. Too many alerts without clear ownership become background noise. Measure how many exceptions are resolved with a documented cause, not merely how many notifications the system produces.
Reference points: GS1 identification keys and GS1 EPCIS. Standards describe identification and event structures; the worked quantities here are not benchmark data.