What is GPS drift?

Leave a GPS tracker on a table and watch its reported position for an hour. It will not sit still. The plotted point wanders in a loose cloud, sometimes tens of metres wide, occasionally jumping much further. The device has not moved. Its estimate of where it is keeps changing.

That wandering is GPS drift, and it is the single most common cause of geofencing systems that alert for no reason.

What causes it

A GPS receiver calculates position by timing signals from satellites. Every step of that has error in it:

• Atmospheric delay. Signals slow down passing through the ionosphere and troposphere by an amount that varies with weather and solar activity.
• Multipath. Signals bounce off buildings, so the receiver measures a reflected path that is longer than the direct one. This is why drift is far worse in cities than in fields.
• Satellite geometry. When visible satellites are clustered together rather than spread across the sky, small timing errors translate into large position errors.
• Clock error. The receiver’s own clock is far cheaper than the atomic clocks aboard the satellites.

The receiver has no way to distinguish “I moved two metres” from “my estimate shifted two metres”. Both look identical in the data.

How much drift to expect

Conditions	Typical error
Open sky, clear view, good satellite geometry	3–5 m
Suburban, some tree cover	5–15 m
Urban, tall buildings	20–50 m
Dense urban canyon, indoors near a window	50–200 m
Indoors, no sky view	Position may be unavailable, or wildly wrong

Cheap 2G trackers sit at the worse end of every row. They use older chipsets with less capable antennas, and many fall back to cell tower positioning when they lose satellite lock — which is accurate to hundreds of metres, not tens.

Why it breaks geofences

Picture a boundary drawn tightly around a building, and a tracker parked just inside it. Drift pushes each reported position a few metres in a random direction. Some land inside the boundary, some outside.

A naive system compares each position to the boundary and alerts on every change from inside to outside. Over an afternoon, that produces dozens of exit alerts from a vehicle that never started its engine. The user disables notifications, and the system is now worse than useless — because it will also be ignored on the day something real happens.

How to prevent it causing false alerts

Draw boundaries larger than your error. This is the fix that requires no software at all. If your device is accurate to 20 metres, a boundary drawn exactly on the property line will misfire. Add a buffer of at least 50 metres, and include the car park and driveway. See GPS accuracy for choosing a sensible radius.

Debounce alerts in software. Drawing a bigger zone reduces drift crossings; it does not eliminate them. A device parked near any boundary will eventually produce one. The system has to be built assuming this happens.

What Geoblip does

Geoblip treats drift as the normal case rather than an exception.

Each device-zone pairing carries an alert status of ok or alerting, maintained separately from the device’s physical inside/outside position. An exit notification is sent only when that status transitions from ok to alerting. If the device is already alerting and drifts back and forth across the boundary, the physical position is updated correctly but no further notification is sent.

Getting a second alert requires a full alerting → ok → alerting cycle — the device has to genuinely re-enter the zone and then leave again. Drift produces tiny oscillations right at the boundary, so it cannot complete that cycle. Real movement can.

The mechanism is described in more detail under alert debouncing and geofence hysteresis.

Drift is not the same as inaccuracy

Worth separating two ideas that get muddled. Accuracy is how far the reported position sits from the truth. Drift is how much that reported position moves around while the device does not. A receiver can be consistently wrong by 30 metres with very little drift — that is a systematic offset. Or it can average out to the correct position while scattering wildly around it. The first breaks boundaries that are drawn in the wrong place; the second breaks alerting. They need different fixes.

Last reviewed 7 August 2026.