Should an event geofence be a circle or a polygon?

Use a circle for a single open space where a radius roughly matches the venue. Use a traced polygon for anything L-shaped, long and narrow, or sharing a building with unrelated businesses, where a circle large enough to cover the venue also covers places attendees are not.

How large should an event geofence be?

Large enough to include the queue outside the door. Attendees waiting to get in are legitimately at the event, and a boundary drawn tight to the building refuses exactly the people about to walk through it.

Does GPS accuracy affect check-in?

Yes, and it should. A phone reports a position and an accuracy radius, and using only the position treats a confident reading and a wildly uncertain one as equally trustworthy. A 100-metre accuracy reading inside a 30-metre fence is a phone indoors, not proof someone is elsewhere.

Event geofencing: circle or polygon, and what accuracy means

A circle takes seconds to set up and fits almost no real building. A traced polygon fits the venue but takes a few minutes. Choose by the shape of the space — and remember the boundary matters less than how you treat an uncertain reading.

Choose by the shape of the space

Venue	Use	Why
Standalone club, hall, warehouse	Circle	One footprint, roughly round; a radius covers it without covering much else
Field, park, festival site	Circle	Boundaries are soft anyway, and a generous radius costs nothing
L-shaped or long, narrow venue	Polygon	A circle covering both ends also covers the street between them
Unit in a mall or mixed-use block	Polygon	A circle admits the shop next door and the food court
One room in a conference centre	Polygon	Otherwise every parallel session is inside your event

The rule underneath: a circle is right when the smallest circle containing your venue does not contain much that is not your venue. For a lot of urban buildings that is simply false, and the polygon is worth the extra minutes.

Include the queue

The most common setup error is drawing the boundary tight to the walls. It feels precise. It also means the twenty people queuing on the pavement are outside your event according to the software, and they are the ones most likely to be checking their phones.

Extend the boundary over the entrance, the queue, and the smoking area if you have one. Being slightly generous costs you a handful of people who walked past; being tight costs you the people actually arriving.

Accuracy matters more than the boundary

A phone does not report a point. It reports a point and a confidence radius — "somewhere within about 12 metres of here", or "somewhere within about 90 metres of here". Those are very different statements and a check-in decision that ignores the second is guessing.

What ignoring accuracy does

Indoors, it rejects real attendees. Inside a building, phones fall back to wifi and cell positioning. The reading drifts and the radius widens — a phone that says "within 90 metres" is telling you it is indoors, which is exactly where your attendee should be.
Outdoors, it admits people who are not there. A confident reading 200 metres away is much stronger evidence of absence than an uncertain one 40 metres away, and treating them the same wastes the signal.

The useful behaviour is to fold the radius into the decision: an uncertain reading near the boundary gets the benefit of the doubt, a confident reading well outside it does not.

Practical setup, in order

Trace the building if it is anything other than a simple footprint.
Extend over the entrance and any queuing area.
For indoor venues, be generous rather than tight — you are fighting positioning, not fraud.
Do not use the fence as an entry control. It decides what counts as present, not who gets in; blocking check-in at capacity makes your own numbers worse.

What a geofence will not do

It will not tell you which room inside a building someone is in — that is below the resolution of consumer positioning. It will not work for virtual events, obviously. And it will not produce a number for a fire officer: it sees attendees running the app with location enabled, which is a strong sample and not a census.

Within those limits it answers the question ticket scans cannot: not who arrived at some point tonight, but who is inside the boundary now.

Last reviewed August 2026. If something here stops being true, it is a bug — tell us.