How to Handle Store Outages Without Breaking Inventory, Ecommerce and Pickup

A product-detail page can tolerate some uncertainty. A pickup reservation cannot. The moment the retailer promises “ready today at Store 214,” availability becomes an operational commitment tied to a specific location. That makes BOPIS one of the first workflows to expose weaknesses in inventory freshness.

Useful policies include reserving inventory as soon as the order is accepted, separating the reservation count from physical stock, and imposing a stale-data cutoff for locations with unstable connectivity. The exact threshold depends on product velocity and store operations. A high-volume fashion SKU may need a tighter rule than a slow-moving home-goods item.

Reconnection is a controlled recovery event

When the connection returns, the wrong move is to push the entire backlog as fast as possible. Hundreds of stores reconnecting after a regional outage can overwhelm the central broker, inventory service or downstream ERP. Recovery needs rate control, ordering rules and visibility into backlog age.

1. Confirm the store identity and last acknowledged event position.

2. Replay queued movements in the order required by the business domain.

3. Reject or quarantine malformed events without stopping the entire queue.

4. Deduplicate every replayed event using its unique identity.

5. Throttle recovery if downstream systems begin to saturate.

6. Run reconciliation after the queue is drained instead of assuming successful delivery equals a correct inventory state.

Conflict resolution has to be a business rule

Two systems can legitimately change the same SKU during the outage window. The store records a sale. The central OMS creates a reservation. A manager performs a count adjustment. Simply choosing the “latest timestamp” can produce the wrong result because clocks differ and the events represent different business meanings.