Making an inventory promise the store can keep

A stock count is only the starting point

Consider an item with two units showing in store. One may be in a shopper’s basket and the other in a damaged-goods bin. An online order that treats both as available creates work for the store and disappointment for the customer. Faster replication alone cannot fix a rule that counts the wrong stock.

This composite starts with the points where inventory changes meaning: receipt, reservation, pick, cancellation, return, and adjustment. Store operations, commerce, and supply chain need to agree which system owns each event and when an item becomes eligible for another order.

Separate the reporting view from the reservation

A shared inventory model would combine stock movements, reservations, and sellable status. Every update needs a source timestamp and an identifier that allows duplicate events to be ignored. Late events must be reconciled; arrival order is not necessarily the order in which stock changed.

BigQuery could support reconciliation and analysis of recurring exceptions. The checkout decision belongs in a transactional availability service with explicit reservation rules. An analytical warehouse should not be responsible for preventing two customers from reserving the same last unit.

The pilot would use a small store group and a category with frequent collection orders. When a feed becomes too old, the service should apply an agreed fallback, such as withholding collection availability pending confirmation. That choice has a commercial cost, so the acceptable freshness and stock buffer need business owners.

Measure the promise through to collection

The useful measures are cancellations caused by missing stock, successful first picks, orders ready within the promised window, and time spent resolving exceptions. Conversion also belongs in the review: an overly cautious availability rule can reduce cancellations simply by suppressing too many valid sales.

Before adding more stores, the team would review stock adjustments and false unavailability alongside customer outcomes. The intended result is a promise that reflects both the stock position and the reliability of the information behind it, with a defined response when that information is incomplete.