GPS speed versus the speedometer in the dashboard

Put a satellite speed readout next to the dial in almost any car and the dial will be the higher of the two. At an indicated hundred kilometres per hour the phone will typically say ninety-four to ninety-seven. This is not a fault in either instrument, and the phone is not being generous with you. It is the direct consequence of a rule that governs how vehicle speedometers must be built, combined with two completely different ways of working out how fast something is moving. Once you know which instrument is answering which question, the gap stops being mysterious and becomes useful.

The rule that makes the dial exaggerate

Type approval regulations for vehicles say that the indicated speed may never be lower than the true speed, and may exceed it by no more than ten per cent plus four kilometres per hour. Manufacturers therefore aim deliberately high: an instrument that read exactly true would fall foul of the rule the moment a tyre wore down or a customer fitted a slightly smaller wheel. The comfortable engineering answer is to build in a permanent optimistic bias of roughly five to seven per cent, which is why the number you see at motorway speed is almost always a few per cent above what you are actually doing over the ground.

What the car is really measuring

A modern speedometer counts wheel rotations, usually from the same sensors that feed the anti-lock braking system, and multiplies by a rolling circumference stored in the car's software. That circumference is a constant, but the real one is not. It changes with tyre pressure, with load, with temperature, and slowly with tread wear, since a worn tyre is a smaller tyre and turns more times per kilometre. Fit wheels that are a size larger or smaller than the factory fitment and the error moves again, this time permanently. None of these effects are visible from the driver's seat, which is another reason the calibration is set optimistically.

What a satellite receiver measures

A phone does not count wheel turns; it measures the Doppler shift of the signals arriving from satellites overhead, and derives a velocity vector from that shift. Because the calculation depends on frequency rather than on successive positions, it is far more precise than you would guess from the position accuracy figure: a receiver that only knows where it is within five metres can still know its speed within a fraction of a kilometre per hour. It measures true speed over the ground, independently of the car, the wheels or the tyres, and it does not drift as components wear.

Where satellite speed struggles

It is not perfect, and the failures are specific. In a tunnel there are no satellites at all, and the readout goes stale or disappears. Between tall buildings, signals arrive by reflection as well as directly, and the receiver can briefly compute a velocity from a path that bounced off a facade. On a steep gradient the receiver measures speed along its actual three-dimensional path, which is very slightly longer than the distance covered on the map. And there is latency: the fix arriving on your screen describes a moment perhaps a few tenths of a second in the past, so during hard acceleration or braking the phone lags reality while the wheel sensor does not.

Which number to use for what

For the question am I within the speed limit, the honest answer is that enforcement is based on true speed, which is what the satellite figure estimates and what the dashboard deliberately overstates. That does not make the dial useless: it makes it conservative, and a driver who obeys the dial is always inside the limit. For measuring performance, comparing two runs, or checking whether a fitted wheel size has changed the calibration, the satellite figure is the fairer measurement because it is independent of the car. For instantaneous reaction in traffic, the dial wins, because it never goes blind under a bridge.

Checking your own car's error

The measurement takes two minutes on a flat, quiet stretch of road with a clear sky, with a passenger reading the phone so the driver watches the road. Hold a steady indicated speed and note both numbers, then repeat at two or three different speeds. The error is proportional rather than fixed, so you should see something like ninety-five against an indicated hundred and forty-seven or forty-eight against an indicated fifty. Write the percentage down: it is the correction factor for your car with those tyres, and it explains the mismatch between the trip computer's fuel consumption and the figure you calculate at the pump.

Reading accuracy alongside speed

This is the reason a serious speed display prints its accuracy estimate next to the number rather than hiding it. A reading of ninety-six with a five metre accuracy figure under open sky is trustworthy. The same reading with a thirty metre figure, in a city, in heavy rain, or with the phone lying face down in a cupholder, deserves much less confidence. The app this guide belongs to shows the figure permanently for exactly that reason, and greys the display when a fix goes stale rather than continuing to show a number that is no longer being updated.