Setting up geofence boundaries for multi-building sites

A geofence answers exactly one question: was this device inside this boundary at this moment? Every configuration decision is a trade between two failure modes.

Of the two, false rejections do more damage early on. A guard who cannot clock in on their first shift concludes the system is broken, and that impression is expensive to reverse. Start slightly generous, then tighten once you have seen real clock-in data for the site.

Configure for the conditions you have, not for the ideal case. Accuracy degrades badly in exactly the environments guarding takes place in.

• Open ground with clear sky is the best case, and is rarely where the entrance is.
• Dense urban sites suffer multipath error, where signals bouncing off surrounding towers can place a device across the street.
• Indoors, accuracy falls off sharply, and lobby clock-in is the norm rather than the exception.
• Underground and in parking structures, GPS is effectively unusable.
• Cold starts take longer to acquire a fix, which is precisely when a guard is standing at the door trying to clock in.

The instinct is to draw one large circle around everything. That does work for clock-in, but it makes the clock-in claim almost meaningless, because "somewhere on a 40-acre campus" is not evidence of being on post.

1. 1Anchor the geofence on the building where guards report at shift start, usually the main lobby, gatehouse, or security office.
2. 2Size the radius to that building's footprint plus a margin for local GPS error, not to the campus.
3. 3Cover every other building with patrol checkpoints, which are what actually prove the rest of the site was walked.
4. 4Watch the first two weeks of clock-in data and adjust once, rather than guessing repeatedly.

This splits the job correctly: the geofence proves the guard arrived at the site, and the checkpoints prove they covered it. Trying to make one geofence do both is what produces either false rejections or worthless proof.

Split when the buildings are operationally separate, not merely physically separate:

• Each building has its own assigned guard and its own shift.
• The buildings belong to different clients, or need separate invoices and separate reports.
• The buildings are far enough apart that one geofence covering both would also cover a lot of ground that is neither.
• Post orders differ meaningfully between them.

Keep them as one site when a single guard covers the whole campus on one shift under one set of post orders. Splitting in that case just makes a guard clock in and out repeatedly to walk their normal round.

• Underground or structure parking: do not rely on GPS below grade. Clock in at the entrance and use checkpoints per level.
• High-rise sites: a device 30 floors up can report a position well off the footprint. Widen the radius rather than fighting it.
• Construction sites: the entrance moves as the build progresses, so revisit the geofence rather than setting it once.
• Adjacent client buildings: two loose geofences that overlap mean a clock-in cannot distinguish which site the guard is on. Tighten both.
• Dead phone or denied location permission: have a documented supervisor-recorded fallback so honest guards are not forced to improvise.

Each site is stored with an address, coordinates, and a geofence radius, all editable from the dashboard at any time, so the tune-after-two-weeks step above is a quick edit rather than a rebuild. Sites are unlimited, patrol checkpoints are configured per site, and clock-in attempts outside the boundary are refused with an alert to the supervisor.