With no network the floor has two real choices: stop and wait, or write it in a notebook and backfill later. A system that cannot submit offline and also forbids backfill does not exist in their hands. Offline is not a tech preference. It is whether this document can be “taken now, sent when there is signal.” If it can wait, build a queue. If it cannot—and you must deduct stock or block double redemption—offline will scramble the books. Improve signal or move the work point. Do not force offline.
Patrol photos, arrival check-in, meter reads, finish notes. Their value is evidence left at the time. The back office seeing them a few minutes later is usually fine. Finishing on site cuts rework. See Finish patrols on site, cut rework.
Stock deduct, punch-card redeem, release outbound, pay. Two offline phones each deduct once and reconciliation breaks first. Store redemption should confirm online. Design is in How to design store redemption and reordering.
Mini programs and H5 are weaker offline than a native APP. If offline is hard, selection leans APP. See When you actually need a native APP. Office staff approving on 4G do not need an offline engine. Spend the budget on field roles.
The queue must be visible: which items are still on device, whether failure can retry, whether photos were dropped. Silent failure is the worst—users think they submitted; the back office has nothing. Write the conflict rule: two people edit one ticket offline, whose wins. Accept this on a weak-network device, not only a simulator. Steps are in From brief to launch: how a mobile build runs. DaXi walks signal dead zones with the floor first, then decides on offline so complexity is not piled on people who do not need it. To assess field network, go to the Apps & Mini Programs and say the roles and whether you have basements or warehouse zones.
Walk a half shift before you decide on offline. Note where the phone shows no service, 2G only, or a Wi-Fi name that will not actually connect. If dead zones are only the elevator and B1 parking and work happens on the floor shop, skip offline—a “please submit at the door” prompt is enough. If the work point itself is a shielded room or cold store, no queue means you are asking them to fake a later record. The reason a PC system does not match the floor is often this blank. See If field staff never sit at a PC, the PC system does not exist.
Cut the offline scope: cache today’s ticket list and blank forms. Do not stuff three months of history and the full material master into the phone. Bigger cache means more conflicts and privacy risk. When someone leaves or changes phones, how unsynced queue items hand over must be written into policy or you get “I submitted it—it is on the old phone.”
Photos are the heaviest offline payload. Compress first so the device does not freeze when signal returns. Failure must be tappable retry—not a one-shot “network error.” After signal, sync order must be stable: document then attachments, or the whole ticket. Half-success makes QA and finance distrust the system together. If the floor still will not use it after launch, follow Launched, unused: how to diagnose to see whether they are stuck on submit failure.
Whether the timestamp is submit time or sync time must be written on the report. QA chasing sync time makes the floor feel the system is blaming them.
Basement, cold store, field—which documents can wait for signal, which must deduct stock on the spot. Send roles and the network as it is. We will recommend a queue or a moved work point.