Why is my phone's GPS only accurate to 3–5 metres?

Because the phone was built to find a restaurant, not a stake.

That answer is glib, but it is also the shape of the real one: everything about a phone’s positioning hardware is a compromise made in favour of size, cost, and battery life. There is no hidden setting, no better app, and no calibration ritual that changes it. This guide explains what the actual limits are, so you can tell the difference between advice that helps and advice that just sounds technical.

Where the metres come from

Satellite positioning is a timing problem. Each satellite broadcasts a signal stamped with the time it left; the receiver measures how long it took to arrive and converts that to a distance. Light travels about 30 cm per nanosecond, so a nanosecond of timing error is 30 cm of position error.

The errors come from four places, and a phone is at a disadvantage in all of them.

1. The antenna is the biggest problem

This is the one people skip, and it dominates.

A GNSS antenna wants to be reasonably large, to sit above a ground plane that shields it from reflections coming off the earth, and to be pointed at the sky with nothing in the way. A phone antenna is a few millimetres of metal squeezed into the edge of a slab that also holds cellular, Wi-Fi, Bluetooth, UWB and NFC radios, and it spends its working life tilted at a random angle inside a human hand.

Two consequences follow:

• Weak signal. GNSS signals arrive below the thermal noise floor — weaker than the background hiss of the electronics receiving them. A small antenna recovers less of a signal that was already almost nothing.
• Multipath. With no ground plane, the phone happily receives signals that bounced off a wall, a wet road, or a car roof before arriving. A reflected path is longer than the direct one, so the measured distance is too long. This is why your position is worst in exactly the places you most want it: town centres, farmyards, tree lines.

A dedicated receiver’s antenna is a completely different object — larger, purpose-shaped, mounted with sky above it and metal below it. That single difference is worth more than any software.

2. Most phones listen on one frequency

The ionosphere delays satellite signals by an amount that depends on their frequency. That is enormously useful: a receiver listening on two frequencies can compare them and measure the delay directly.

Most phone chips are single-frequency. They cannot measure it, so they apply a broadcast model instead — a rough average that removes maybe half of a metres-scale error. Some recent flagships do have dual-frequency GNSS and are genuinely better for it. They still have the phone’s antenna.

3. Nothing is correcting the rest

Even with a perfect antenna and two frequencies, satellite clock and orbit error and tropospheric delay remain. Removing those needs an external correction stream — the thing this whole industry is built on, and the thing a phone on its own has no source for.

Corrected properly, those errors go from metres to centimeters. Uncorrected, they set the floor.

4. The “accuracy” number is a guess

The accuracy radius your phone reports is the receiver’s own estimate of its uncertainty, computed from the same measurements that produced the position. When those measurements are corrupted by reflections, the estimate is corrupted too — usually optimistically. A phone confidently reporting 3 m while sitting 15 m off is not lying; it is wrong in a way it cannot detect.

What the blue dot is not telling you

Consumer maps do a great deal of quiet work to make positioning look better than it is:

• Snapping. If you are near a road, you are drawn on the road.
• Smoothing. Successive fixes are filtered so the dot glides instead of hopping.
• Sensor fusion. Accelerometer, gyroscope and compass keep the dot moving plausibly through gaps.
• Network positioning. Wi-Fi and cell databases fill in where satellite reception is poor, at accuracies of tens of metres.

All of that is excellent for navigation and useless for measurement. It makes the display steadier without making the position truer, and it is why recording a field boundary by walking it with a phone produces a shape that looks tidy and does not close.

Do tablets or rugged handhelds do better?

A little, and not for the reason people expect. A tablet has more room for an antenna and more space away from the other radios, so it often behaves slightly better than a phone in the open. A rugged handheld is built to survive being dropped, which is worth a lot in a field and worth nothing to the position.

But both are still a small antenna with no ground plane inside a case held by a person, and unless the specification explicitly says multi-band GNSS — two or more frequencies — the ionospheric error is still being guessed from a model rather than measured. “Professional”, “survey” and “high-precision” appear on plenty of handheld data sheets that describe metre-class devices.

The only reliable tells on any spec sheet are the number of frequency bands, the number of constellations, and whether the device can accept a correction stream at all. Everything else is enclosure.

Things that do not help

• Turning off battery saver, toggling location mode, “calibrating the compass”. The compass affects which way the arrow points, not where the dot is.
• GPS “booster” or “repair” apps. There is nothing for them to repair. At best they trigger a fresh almanac download, which shortens time to first fix and does nothing for accuracy.
• A case, or holding the phone differently. Marginal. Real, but marginal.
• A newer phone. Small gains, unless it brings dual-frequency — and even then, see the antenna.

Test your own phone in ten minutes

You do not have to take any of this on trust. Two experiments, both free:

The scatter test. Put the phone flat on a fence post in the open, start an app that records position, and leave it for ten minutes without touching it. Nothing moved, so every metre of spread in that track is error. In the open you will typically see the dot wander over a few metres. Repeat beside a building or under a tree and watch it get much worse — that is multipath, visible.

The return test. This is the one that matters for anything you will come back to. Mark a specific point — a gatepost, a drain cover. Walk 200 m away, wait a couple of minutes, and come back to exactly the same physical spot. Compare the two recorded positions. The gap between them is your phone’s repeatability, and it is usually noticeably worse than its claimed accuracy.

Repeatability is the property that decides whether last year’s recorded boundary is any use this year, and it is the property phones are weakest at. It is also the reason the specification quoted for a dedicated receiver on this site is “±2.5 cm repeatable” rather than a bare accuracy figure.

What does help

An external GNSS receiver, paired over Bluetooth. It replaces every weak link at once: a real antenna with a ground plane, multi-band and multi-constellation tracking, and the ability to accept a correction stream. The phone keeps doing what it is good at — being the screen, the app, and the data connection.

The step from 3–5 metres to centimeters is not incremental. It is the difference between “somewhere in this field” and “this corner of this field, again next week”.

For how the correction side works, see how Northing RTK works; for what external receivers can and cannot do with an iPhone specifically, see RTK receiver for iPhone.

How much better does it get?

With a dedicated receiver and a correction stream, measured on our own logged drive sessions:

• ±2.5 cm repeatable accuracy, RTK Fixed, open sky.
• Fixed within seconds of a cold start, with no warm-up.
• Fixed for 88% of a 40-minute drive including deliberate tree cover, a bridge and buildings, recovering in 1–2 seconds after each.

And without any subscription at all, using the free mode built into the receiver: 50% of positions within 7.7 cm and 95% within 22 cm once converged, parked. Even that free fallback is roughly an order of magnitude better than the phone in your hand — though it takes minutes to settle and behaves quite differently while moving, which is set out honestly in RTK vs PPP.

So is phone GPS bad?

No — it is superb at what it was designed for. Finding a junction, hailing a car, recording roughly where a photo was taken: 3–5 metres is plenty, and it does it in a device that also fits in a pocket and lasts a day.

It is simply the wrong instrument for measurement. Asking a phone to mark a drain to the nearest few centimetres is asking a tape measure to weigh something. The fix is not a better setting; it is a better antenna, another frequency, and a correction stream — which is exactly what an external receiver is.

Common questions