A phone that loses the sky still shows a location, often accurately enough to identify a floor of a shopping mall. Several fallback systems produce that estimate without any satellite involvement.
Satellite signals fail indoors
Positioning satellites transmit from medium orbit at very low power by the time signals reach the ground. Roofs, steel and concrete attenuate them below the level a receiver can decode reliably.
Reflections make matters worse. A signal bouncing off a nearby building arrives later than the direct path, and a receiver treating that as direct computes a position offset by a considerable distance.
Urban canyons produce this error outdoors too, which is why a phone can place a walker on the wrong side of a downtown street.
Wireless networks act as landmarks
Every wireless access point broadcasts a unique identifier continuously. Phones passing by record which identifiers were audible and how strong they were, alongside a satellite position obtained outdoors.
Aggregated across millions of devices, this produces a database mapping network identifiers to approximate locations. Indoors, a phone hears several identifiers and looks them up.
The method works because access points rarely move. When one is relocated to another city, devices near it can briefly report the old location, a known failure of the approach.
Sensors carry the estimate between fixes
Phones contain accelerometers, gyroscopes and magnetometers. Together these detect steps, turns and heading, allowing the device to extend a known position by tracking movement.
This technique accumulates error steadily, since small heading mistakes compound over distance. It is useful for seconds and minutes, not for an afternoon.
Barometric pressure sensors add the vertical dimension. Pressure falls measurably between floors of a building, which is how a device distinguishes levels within a mall or airport.
Short-range beacons fill specific spaces
Venues sometimes install small transmitters that broadcast an identifier a phone can detect within a limited radius. Museums, terminals and large stores have used them for wayfinding.
Because the range is short, detecting a beacon localizes a device tightly. The tradeoff is that every space needs hardware installed and maintained.
Newer chips measure signal round-trip time rather than strength, which resists the interference that made strength-based estimates unreliable.
Accuracy is reported, not guaranteed
Operating systems return a position together with a confidence radius, drawn on maps as a circle. A wide circle indicates the estimate came from network lookups rather than satellites.
Applications that ignore the radius and treat the center point as exact are the source of many apparent positioning errors that are really presentation errors.