Position Accuracy at Building Scale

Everything about geofenced attendance depends on one figure: how wrong the phone's position estimate is at your workplace.

Typical error

Open sky, standing outside: five to ten metres.

A street with buildings either side: twenty to fifty metres, because signals arrive after bouncing.

Inside a building: no satellite fix at all. The phone falls back to mobile masts and known wifi networks, giving anywhere from thirty metres to several hundred.

A basement, a warehouse with a metal roof, a hospital interior: worse still, and sometimes a position derived from a mast a kilometre away.

Why this is the central problem

People clock in indoors, because that is where work starts.

Which is exactly where the position estimate is worst.

A shop frontage is fifteen metres wide. An error of fifty puts the estimate on the far side of the street.

So the boundary has to be drawn large enough to absorb the error, and a boundary that large also includes the car park, the pavement, the café next door and sometimes the bus stop.

That trade-off is the whole design problem, and no product solves it — they only choose where to put it.

What the phone actually reports

A position and an accuracy radius.

The radius is a confidence estimate, not a guarantee, and it is frequently optimistic indoors.

Good apps record it alongside the position. Ask whether yours does, because a refused punch with an accuracy of two hundred metres is a different event from one with an accuracy of eight.

A system that stores only the point has thrown away the information needed to explain any dispute.

Measuring it at your own site

An afternoon, and it replaces every vendor claim.

Stand at the entrance, inside the door, at the back of the building, in the car park, across the road.

Record the position and accuracy at each, several times, on two different phone models.

Do it again at a different time of day, because mast and wifi conditions change.

The spread you see is your design constraint, and it will differ between sites more than most people expect.

What the numbers mean for the design

If indoor error is under about twenty metres, a tight boundary can work.

If it is fifty or more, a boundary that accepts indoors will also accept the street, and the system is not proving much.

If there is no usable fix indoors at all, geofencing is the wrong method for that site and one of the alternatives should be used.

Knowing which case you have is the first task, and it precedes choosing a product rather than following it.

Two phones, not one

A detail that changes the measurement.

Android and iOS estimate position differently, and models within each differ again.

An older handset with a weaker receiver behaves worse indoors, and someone on your team has one.

Measure on the worst phone anyone actually uses, not on the newest.

Because the radius has to work for them too, and a system tuned to a flagship refuses everyone else.

An implementation prompt

During configuration, employee productivity reports can prompt questions about fields, ownership and output. Confirm current capabilities and document each plan, integration or policy assumption.

Independent reference

For an external point of reference, see the official GPS information service. Its published guidance provides useful context for testing the assumptions in this note.