GPS vs Wi-Fi vs cellular location accuracy

How does GPS work out a position?

GPS works out a position by timing signals from several satellites at once and solving for the one spot where those distances agree. It needs a view of the sky, and the receiver does all the work. Nothing on the ground has to know the device exists.

Under open sky it is very good. The U.S. government states that GPS-enabled smartphones are typically accurate to within a 4.9 m (16 ft) radius under open sky, and that high-quality single-frequency receivers have attained horizontal accuracy of 1.82 m or better 95% of the time (GPS.gov). The same page lists what makes it worse: signal blockage from buildings, bridges and trees, indoor or underground use, and signals reflected off buildings or walls.

Most of that list describes a construction site. A crew on an open lot gets the open-sky figure. A crew inside a concrete core, under a deck pour or between two tall buildings does not.

How does Wi-Fi positioning work out a position?

Wi-Fi positioning looks up the Wi-Fi access points a device can hear in a database of where those access points are, then estimates a position from which ones are heard and how strongly. The device does not join any network. It only listens.

Google's Geolocation API documentation says that when a request includes two or more Wi-Fi access points, the returned accuracy radius is typically around 20 meters, and that accuracy improves with more access points and stronger signals (Google Geolocation API). The same documentation requires at least two physically distinct, stationary access points, which is why a site with one router, or none, gives weak results.

A different Wi-Fi method, round-trip-time ranging, measures the distance to each access point instead of guessing from signal strength. Android's documentation says it is typically accurate within 1 to 2 meters, but it needs access points that support the IEEE 802.11mc or 802.11az ranging protocols (Android Wi-Fi RTT). On a jobsite you do not control the routers next door, so plan around the signal-strength method.

How a badge learns a jobsite's Wi-Fi over the first visits is covered in indoor location without beacons.

How does cellular positioning work out a position?

Cellular positioning estimates a position from the cell towers a device is connected to or can hear. The simplest version only knows which cell the device is in. More advanced versions defined by 3GPP, such as enhanced cell ID and observed time difference of arrival, add timing and signal measurements from several towers (3GPP TS 36.305).

For the cell-ID style lookup a device or service can do on its own, the numbers are coarse. Google's documentation says macro cells, the most common type, typically give a radius of hundreds of meters, up to several thousand meters where towers are sparse, and that a radius below 100 meters is uncommon. Small cells such as indoor repeaters can reach 10 to 30 meters, but you cannot count on one being near your site.

Cellular still has a use. It works almost anywhere there is signal, so it can show that a device stayed in the same general area between better readings. It cannot confirm that a worker arrived at a particular jobsite.

How do the three methods compare on a jobsite?

GPS wins outdoors, Wi-Fi wins indoors where there are networks to hear, and cellular is a fallback that answers which area, not which site. The table puts the sourced figures side by side.

What does an accuracy number like 20 m actually mean?

An accuracy number is the radius of a circle the true position probably falls inside, and the word that matters is probably. Android defines its accuracy value as the radius at the 68th percentile: there is a 68% chance the device is within that distance of the reported point (Android Location reference). Apple describes its value as a radius of uncertainty without stating a percentage (Apple Core Location). GPS.gov quotes one receiver figure at 95%.

So a reading shown as "±30 m" on an Android device is outside its own circle about one time in three. That is not a fault. It is what the number means.

Why are readings worse between tall buildings and indoors?

Between tall buildings, GPS loses part of the sky and hears reflected signals, and indoors it often loses the sky entirely. GPS.gov names both causes: blockage by buildings and multipath, where a signal bounces off a wall and arrives late, so the receiver thinks the satellite is farther away than it is.

• Blocked sky. With buildings on both sides, the satellites left in view sit in a narrow strip overhead. Poorly spread satellites give a poor position even when every signal is strong.
• Reflections. A reflected signal can pull a fix toward the street or across it. Because the same buildings reflect the same signals, the error tends to repeat for several readings in a row rather than averaging out.
• Indoors. Under a roof, GPS weakens or drops out, and the position comes from Wi-Fi if there are networks to hear.
• Basements and pits. Every signal is weak there. A well-built device stores readings and sends them when it reconnects, but it cannot invent a position it never measured.

How accurate does location need to be to tell two jobsites apart?

Accurate enough that the reading's uncertainty circle cannot reach the other site. A useful working rule: the gap between two site boundaries should be at least twice the 95% radius of a typical reading, because the reported point can be off by up to that radius toward the other site, and its circle reaches that far again.

This is why the honest promise for any badge or app without hardware installed inside the building is which jobsite, not which room. For the boundary itself, see how geofencing works for time tracking.

Why should a single reading not decide where a worker is?

Because a predictable share of perfectly normal readings land outside their own circle. If the radius is a 68% radius, about one reading in three falls outside it. Even at 95%, one in twenty does.

A sound system treats each reading as evidence, not a verdict. In practice that means a few rules.

• Weigh each reading by its method and its stated accuracy, not just its coordinates.
• Do not let one reading near a boundary end a site visit. Wait for a second reading to confirm or cancel it.
• Do not let a cell tower reading confirm arrival at a new site. It can only support a place already confirmed by GPS or Wi-Fi.
• Keep the method and confidence with every reading, so the office can see why a minute was placed where it was.
• Show lower-confidence stretches on the timesheet for review instead of hiding them or presenting them as precise.

How ZeroBadge handles this

The ZeroBadge badge uses GPS outdoors and Wi-Fi positioning indoors, learning each site's Wi-Fi over the first visits. Between better readings it uses cellular continuity, which is always labelled low confidence. Every reading stores its method and confidence, so a manager reviewing the week can see how each stretch of time was placed.

The accuracy is meant to answer which jobsite a worker is at, not which room. Precise indoor positioning is not promised. See multi-jobsite time tracking for how readings become hours by jobsite.