Underground, inside a steel-frame plant or behind a hill, the app shows a spinner and the worker gives up; the register says absent.
Apps that record time at upload show 09:40 for a worker who reached at 07:55, creating false late marks and overtime disputes.
Some apps drop face or location checks when there is no server, so the offline mode becomes the proxy-attendance mode.
At 09:00 the dashboard shows 40 absent at a site of 120; the supervisor cannot tell who is absent and whose phone has simply not synced.
Every check-in is written to the phone first with the device time and GPS fix at that moment; upload happens later and never changes the recorded time.
The enrolled face descriptor is cached on the phone, so face match and liveness run without a server; a failed match is recorded as such, not waved through.
Site geofences and shift timings are stored on the device, so a check-in outside the site is flagged offline exactly as it would be online.
The dashboard shows pending-sync devices per site, so 'not yet synced' is distinguished from 'absent' before anyone is marked LOP.
Check-ins, check-outs, patrol scans and incident reports stored locally with timestamp, location and verification result.
Queued records upload in the background when any network appears, including a brief signal on the way out of the site.
Duplicate or overlapping records from retries are merged by rule; genuine conflicts are flagged for supervisor review, not silently overwritten.
Encrypted face descriptors, not photos, stored on the device for on-device matching with liveness detection.
Device time compared with server time at sync; check-ins from a device with a manually changed clock are flagged.
Per-site count of devices with unsynced records and the age of the oldest record, so supervisors chase syncs, not people.
Geofences, shift timings and enrolled face descriptors are stored on the device while it has network.
Employee marks attendance without signal; face match and geofence check run on the phone.
The record is saved locally with its original timestamp and location, marked as pending sync.
Queued records upload automatically when the network returns; duplicates and conflicts are resolved by rule.
Dashboard shows check-ins at the time they happened, with a pending-sync count per device.
The shaft, the basement and the plant floor become valid check-in points instead of dead zones.
Original timestamps mean a worker who reached at 07:55 is never penalised because the sync happened at 09:40.
Face and geofence checks run on the device, so offline mode does not become the easy route for proxy attendance.
Supervisors see pending syncs before month-end, so nobody loses a day's wage to a phone that synced late.
Guards and site staff record checkpoint scans and incident reports underground and sync them with the attendance.
Sync completes long before the payroll lock, so payable days come from all check-ins, not the ones that had signal.
Shift crews entering a pit or underground section where the last signal is at the surface office.
e.g. a 06:00 shift crew marks in at the pit entrance with face verification offline; records sync when the crew bus passes the surface tower at 14:20, all with 06:00 timestamps.
Finishing crews working inside a tower's basement levels and core where signal does not reach.
e.g. 80 workers mark in at the B3 level; the site engineer sees them as present once the records sync from the ground-floor office, with the geofence flag showing none were outside the plot.
Solar parks, telecom tower maintenance and hill infrastructure with patchy coverage across the day.
e.g. a technician checks in at a hill tower site at 10:10 without signal; the record uploads when he reaches the highway at 12:30, and the supervisor sees the original 10:10 time.
Book a free demo and see how Attend Mitra helps Indian businesses track attendance, manage shifts, and export payroll-ready reports.