Offline sync must define conflict rules, authoritative fields, and retry queues. Marketing works-offline without policy causes silent overwrites after launch.
Part 1 — Classify documents: which may create offline, which must be online.
Part 2 — Conflict examples: server wins, last-write-wins, or human merge.
Part 3 — Make local queues observable: failure reason, retries, human intervene.
Prioritize verifiable scenarios: main path, auditable permissions/logs, and clear interfaces.
| Decision lens | Common trap or cost | Safer cut |
|---|---|---|
| Rules first | Conflict/authority fields | Then build |
| Observable queues | Human intervene | Fewer incidents |
| Idempotent post-back | No duplicate docs | Safer integration |
Offline is a product policy, not a toggle
Fully offline viable?
Stock reservation and payments usually no; use allowlists.
How to test offline?
Real weak-net and airplane toggles with conflict cases, not lab Wi-Fi only.
Multi-device on one job?
Lock the work order or partition fields to stop mutual overwrites.
MES/WMS integration tip?
Idempotency keys on post-back to prevent duplicate reporting or issues.