How Accurate It Actually Is

A map showing a vehicle at an address looks definitive. It is an estimate, and knowing the size of the error changes what the data can be used for.

Where the error comes from

Satellite geometry. Fixes are better with more satellites spread across the sky and worse when they are clustered.

Obstruction. Urban canyons, tree cover, multi-storey car parks, warehouses. Signal reflects off buildings and arrives late, which the receiver interprets as distance.

Assisted positioning. Where satellite signal is poor, devices fall back to mobile masts and wireless networks, which is far less precise and is frequently not labelled differently in the interface.

Update interval. A fix every two minutes leaves the position between fixes as an interpolation drawn on a map, not an observation.

What that means in practice

Open road: a few metres. Good enough for anything.

Dense urban: tens of metres, routinely, and occasionally a position on the wrong street.

Indoors or underground: unreliable to absent, with the last known fix sometimes reported as current.

Which means "the van was at number 42" may mean "the van was somewhere in this block", and in a dispute that distinction decides the outcome.

The decisions this affects

Proof of attendance at a specific address: usually fine, provided the geofence is sized for the error rather than drawn tightly around the building.

Time spent at a site: less reliable, because entry and exit are detected at the fence boundary and the error moves both.

Whether someone stopped where they should not have: unreliable, and this is the use where the consequences are most serious for the individual.

Speed: derived from position differences and noticeably noisy at low speed; a device reporting speeding needs corroboration before anything follows from it.

What to ask a vendor

What is the update interval, and is it configurable?

Is the reported position an observation or an interpolation — is that visible in the interface?

Is accuracy recorded with each fix, and can it be seen?

What happens when signal is lost: does the device buffer and backfill, report a gap, or repeat the last position?

The last one matters most. A system that repeats the last known position without saying so produces a record showing a vehicle stationary for an hour when it was in a tunnel.

Using it honestly

Never act on a single fix. Patterns across many are far more reliable than any one point.

Size geofences for the environment, larger in dense urban areas.

Show accuracy in reports where a decision depends on it.

And where a person disputes what the system says, start from the assumption that the system might be wrong, because in the marginal cases it frequently is.

Measure your own

An afternoon that replaces every specification figure.

Drive a known route with known stops.

Compare the recorded positions against where you actually were.

Repeat in the difficult places: the multi-storey, the industrial estate, the city centre.

Record the result, because when a figure from this system is challenged, "we measured it at approximately thirty metres in urban conditions" is a much stronger position than a vendor's brochure.

A concrete product reference

When translating this principle into a buying test, read the feature details provides a concrete workflow reference. Verify the current behaviour in a trial and judge it against the purpose and limits described above.

Independent reference

For an external point of reference, see the official GPS information service. Its material provides useful context for the limits and operating conditions behind a reported position.