What the GPS accuracy figure actually tells you
Every satellite fix comes with an estimate of how good it is, and phones present it as a number of metres. It is one of the most misread values in the whole business. It is not the distance you are currently wrong by, it is not a measure of how good your speed reading is, and it is not a promise. It is the radius of a circle around the reported position within which the true position probably falls, with probably meaning about two thirds of the time. Learning to read it turns a speed display from a number you either trust or do not into an instrument with a stated confidence.
A radius, and a probability
Android reports horizontal accuracy as a radius at sixty-eight per cent confidence. If the screen says five metres, roughly two thirds of fixes taken under those conditions would fall inside a five metre circle around the truth, and about a third would fall outside it. That means the occasional fix is worse than the figure suggests, and it is entirely normal for a track to show a single point jumping sideways while the accuracy number looks respectable. It also means the figure is an estimate the receiver produces about itself, computed from satellite geometry and signal quality, rather than a measurement of actual error, which nobody in the car can know.
Why position accuracy and speed accuracy are different things
Speed does not come from comparing successive positions. It is derived from the Doppler shift of the carrier signals, and that calculation is far better conditioned than the position solve. In practice a receiver reporting a five metre position radius is typically holding speed to within a few tenths of a kilometre per hour, and even a poor twenty metre fix often still gives a usable speed. This is why a satellite speedometer works well in situations where a mapping application is visibly confused about which street you are on.
What makes the number go up
Geometry comes first: the satellites need to be spread across the sky, and a receiver that can only see a narrow strip between two buildings gets a poor solution no matter how strong the signals are. Then obstruction, which is why the figure climbs in tunnels, under dense trees, in multi-storey car parks and beside high walls. Then multipath, where the same signal arrives twice, once directly and once bounced off glass or concrete, and the receiver has to decide which arrival to believe. Weather has a small effect through the ionosphere and troposphere, and a windscreen with a metallic heating or coating layer can cost a surprising amount all by itself.
Cold starts, warm starts and why the first minute is bad
A receiver needs to know which satellites are up there before it can use them. From completely cold that information has to be downloaded from the satellites themselves, which is slow. Phones shortcut it by fetching the same data over the network, which is why a fix appears in seconds with a data connection and can take a minute or more without one. During that first period the accuracy figure is large and falls steadily as more satellites join the solution. The practical rule is simple: let the number settle before you take any measurement seriously.
How to read it while driving
The table below gives a working interpretation. Treat it as a guide to confidence rather than a specification, and pay more attention to changes than to absolute values: a figure that suddenly triples means something about your environment has changed, usually that you have entered a built-up canyon, gone under cover, or that the phone has been moved somewhere with a poorer view of the sky.
Where you put the phone matters
The antenna in a handset is small and it points upward when the phone is lying flat. A phone face up on the dashboard with a clear view through the windscreen gets the best fix available in a car. Slid into a door pocket, buried in a bag, or mounted low behind the steering wheel, it is looking at metal and its solution degrades. Heated or coated windscreens attenuate the signal noticeably, and a few vehicles have a small uncoated area near the mirror precisely to let devices through. Dual-frequency receivers, which use the newer civil signal alongside the traditional one, handle reflections in cities much better, and phones that support them show visibly steadier numbers between buildings.
| Reported accuracy | Situation | How much to trust the speed |
|---|---|---|
| 3–5 m | Open sky, phone on the dash | Fully; speed is good to a fraction of a km/h |
| 5–10 m | Normal driving, some obstruction | Good; brief wobble on sharp corners is normal |
| 10–20 m | Urban streets, trees, poor mount | Usable for speed, weak for position and distance |
| 20–50 m | City canyon, heavy cover, cold start | Treat distance and track as unreliable |
| Over 50 m or no value | Tunnel, car park, no fix | Nothing is being measured; the display is stale |
Frequently asked questions
Does a 5 m accuracy mean my speed is wrong by 5 m/s?
No, and they are not related in that way. The metres describe the position solution. Speed comes from a different calculation based on the Doppler shift of the signals, and it is normally far more precise than the position radius suggests.
Why does the figure get worse in the city?
Buildings block part of the sky, which weakens the geometry, and they reflect signals, which adds paths that did not travel in a straight line. Both effects push the estimate up, and both are why a track can drift a street sideways downtown.
My phone says 3 m but the map shows me on the wrong road. Why?
The accuracy figure is a confidence radius, not a guarantee, and roughly a third of fixes fall outside it. Map matching can also snap you to the wrong parallel road when two run close together.
Do I need a data connection for accurate GPS?
Not for accuracy, only for speed of acquisition. Network assistance provides the satellite almanac quickly. Without it the receiver reads the same information from the satellites, which takes longer but ends up in the same place.
Does a windscreen mount block the signal?
A plain screen is fine. Heated or metal-coated glass attenuates satellite signals measurably, and some cars leave a small transparent window near the rear-view mirror for this reason. If your fix is poor in one car and good in another, the glass is the usual suspect.
What is dual-frequency GPS and does it matter?
It is a receiver that uses a second civil signal alongside the traditional one, which lets it identify and reject reflected paths. In open country the difference is small; in a city it is the difference between a steady reading and one that wanders.
Should the display keep showing a speed when the fix is lost?
No. A number that is no longer being updated is worse than no number, because it looks alive. The honest behaviour is to mark the reading as stale and say so, which is what this app does when fixes stop arriving.
Does using GLONASS or Galileo improve accuracy?
Usually yes, because more constellations mean more satellites visible and a better spread across the sky, which is what the geometry term in the solution rewards. Phones use them automatically without any setting, and the benefit is largest exactly where you need it, in streets where only part of the sky is visible.
A speedometer you can read at a glance, and a head-up display your windshield reflects. NitroHUD shows GPS speed with the accuracy figure printed beside it, mirrors the readout for windshield use at night, logs the trip you are on, and times acceleration runs with your personal bests kept. No ads, no account, and trips stay on the phone.
Get NitroHUD on Google PlayAbout NitroHUD