How the rule works at punch time
Each site in the system has a geofence: a centre point and a radius, or a polygon for irregular plots. When an employee punches, the app takes a GPS fix and computes the distance from the site centre. Inside the radius, the punch is accepted normally. Outside, the system either blocks the punch, accepts it with an 'outside geofence' flag, or asks for a reason – which of the three depends on your policy.
A useful refinement is to compare the GPS accuracy radius as well: a fix 60 m from the centre with 50 m accuracy is genuinely ambiguous, while a fix 2 km away with 10 m accuracy is not. Flagging rather than blocking the ambiguous case avoids punishing honest guards for bad signal. The GPS attendance entry covers accuracy ranges.
Choosing the radius and handling multiple sites
A 50 m radius suits a small shop or office entrance with clear sky. A 100–150 m radius is the practical default for housing societies, warehouses and hospital compounds where the guard may punch from a rear gate or under a canopy. Large factories, campuses and construction sites may need 200–300 m or a polygon that follows the boundary wall. Start slightly generous, look at where honest punches actually land in the first two weeks, then tighten.
Multi-site companies need a geofence per site and an assignment that says which employees may punch at which sites. A guard rostered at Site A should not be accepted at Site B without the roster changing – that is how proxy attendance hides in multi-site agencies. The multi-site attendance guide walks through the setup.
- Shop or office: 50–75 m
- Society, warehouse, hospital: 100–150 m
- Factory, campus, large construction plot: 200–300 m or polygon
- Review actual punch scatter after two weeks and adjust per site
Exceptions and common mistakes
Legitimate exceptions exist: a supervisor visiting several sites, a guard escorting a cash van, an engineer called to a client not yet in the system. Handle them with an 'outdoor duty' or regularization request with approval, not by widening every fence to a kilometre. Keep the exception count visible in a report; a site with 40% out-of-fence punches has a fence problem or a fraud problem.
Two mistakes recur. First, geofencing alone without identity: it proves a phone was on site, not that the rostered person was. Pair it with selfie or face verification. Second, ignoring GPS spoofing: a mock-location app can put a phone inside any fence, so the system must read the mock flag.
A logistics warehouse in Bhiwandi is 220 m long. The original geofence was a 75 m circle at the main gate; guards at the rear dock gate 190 m away failed 30% of punches. The admin changed it to a 150 m circle centred mid-plot, and out-of-fence punches fell to 2%, all of which turned out to be a guard punching from the tea stall across the road.
Attend Mitra supports multiple geofences per company with a configurable radius per site, applies the geofence check at punch time alongside face or selfie verification, flags mock-location on Android, and lets admins see punch locations on the live monitor and in per-site reports.
