Geofencing sounds binary — they're inside or they're not. In practice it's a judgement call, and getting it wrong in either direction costs you.
Set it too tight and your staff stand at the gate in the rain, tapping a button that keeps refusing them. They ring the office. Someone marks them in manually. Within two weeks, half the team is being marked in manually and the geofence has stopped meaning anything.
Set it too loose and a check-in from the car park across the road looks identical to one from the workshop floor.
Start with GPS reality, not site size
Phone GPS is typically accurate to somewhere between 5 and 20 metres outdoors with a clear view of the sky. Under a canopy, between tall buildings, or just inside a doorway, that degrades quickly — 30 to 50 metres is common, and worse isn't unusual.
So a 20-metre radius isn't precision. It's a guarantee of false rejections, because the phone's own uncertainty is bigger than the box you've drawn.
A rule of thumb
Take the distance from your pin to the furthest point staff legitimately check in from, then add 30 to 50 metres for GPS drift.
For most single-building sites that lands somewhere between 100 and 200 metres. Timix defaults to 150 metres, which is deliberately forgiving — it's the value we'd rather explain than have you fight.
Larger sites need more thought. A yard, a campus or a multi-block facility isn't one circle. If your radius has to reach 400 metres to cover the far gate, it also covers the petrol station next door, and you've bought nothing.
When one circle isn't enough
If a site is genuinely spread out, the answer is usually more locations, not a bigger radius. Two 150-metre circles centred on the two gates people actually use tells you far more than one 500-metre circle that covers both plus the road between them.
It also gives you something a single circle can't: which entrance people are using.
Watch accuracy, not just distance
Distance alone can mislead. A check-in recorded 40 metres from the site with a ±5 metre reading is solid. The same 40 metres with a ±90 metre reading tells you almost nothing — the phone was guessing.
This is why Timix refuses readings worse than 100 metres outright and stores the accuracy figure alongside the distance on every geofence check-in. A rejected reading is more useful than a recorded guess, because a guess in a column looks like a fact six months later.
Tune it after a fortnight, not on day one
Set something forgiving, run for two weeks, then look at your out-of-range rejections. If a site is generating them daily from people who were genuinely there, the radius is too tight or the pin is in the wrong spot — very often the pin is on the building's postal address rather than the door staff actually use.
Fixing the pin usually beats widening the circle.