Operational recovery guide
An exception should explain the next safe action.
A red badge is not exception management. Operators need the reason, affected inventory and orders, current authority, available actions, owner, urgency, evidence, and downstream promise in one recovery context.
Exception anatomy
Make the stop understandable and actionable.
What was observed?
Scan mismatch, shortage, damage, quality hold, no stock, wrong location, device timeout, carrier rejection, or policy guard.
What is affected?
Client, product, lot, handling unit, task, order, package, shipment, station, promise, charge, and related work.
Who may decide?
Role, tenant, warehouse, team, lifecycle, segregation-of-duties, step-up, approval, machine, and safety constraints.
What can happen next?
Retry, re-scan, substitute, short, backorder, reallocate, repack, hold, inspect, reroute, void, escalate, or cancel.
Exception families
Route different problems to the correct owner.
Inventory and identity
Unknown SKU, barcode conflict, lot/serial/expiry mismatch, wrong owner, insufficient available stock, location divergence, or count variance.
Picking and consolidation
Short pick, wrong item, blocked location, missing tote, put-wall conflict, sequence failure, or unresolvable allocation.
Packing and label
Carton mismatch, weight/dimension change, validation failure, printer incompatibility, quote expiry, payment action, or carrier rejection.
Automation and device
Capability mismatch, no acknowledgement, bad tag quality, blocked permissive, robot fault, lane divergence, or lost device.
Transport and proof
Missed cutoff, tender rejection, no capacity, late milestone, geofence mismatch, failed delivery, missing POD, or claim trigger.
Recovery tests
Prove that retries do not create new damage.
One result, two attempts.
Idempotency returns or explains the existing outcome without duplicating movement.
Stop stale resolution.
Expected version or lock prevents one operator from overwriting a newer recovery.
Guide, then deny.
Show why the action is unavailable and who can approve it without exposing forbidden controls.
Allow safe actions only.
Queue allowlisted scans; machine commands and unsafe mutations require reconnection.
Propagate the impact.
Update order, shipment, client visibility, notification, billing, and SLA state from one resolution.
Keep the alert open.
Do not close because software transitioned if occupancy or machine observation still disagrees.
Packing control
Put queue and station state beside the recovery action.
Supervisors need workload and exception context; the station operator needs a constrained next step, validation, local devices, and acknowledgement.
- Task-first recovery rather than menu hunting.
- Reason and affected promise remain visible.
- Resolution history feeds analytics and prevention.
Improvement loop
Close the cause, not only the case.
Stable reason taxonomy
Separate symptom, root cause, source process, equipment, supplier, carrier, client rule, and data-quality class.
Time and recurrence
Detect, acknowledge, own, resolve, reopen, repeat, impact, and cost by workflow and cause.
Controlled corrective action
Change rule, mapping, slot, packaging, training, maintenance, connector, or commercial policy with approval and outcome review.