A GPS app measures distance by joining position fixes with straight lines and adding them up. If it records points too rarely, those lines cut across every bend and the total reads short. If it records noisy points while you move slowly, the lines zigzag and the total reads long. How far off you end up depends on your speed, how twisty the route is and how clean the satellite signal is.
That’s why a watch, a phone and a bike computer can finish the same ride with three different distances, and why none of them is necessarily broken.
What is GPS sampling rate? #
Sampling rate is how often a device records your position. It’s usually quoted in hertz (Hz): 1 Hz means one point per second, 5 Hz means five.
Two different rates can matter:
- How often the receiver computes a fix. Most phones deliver location about once a second when an app asks for the highest accuracy.
- How often the app keeps a point. Many fitness watches and apps offer a battery-saving mode that stores a point only every few seconds, or only when your direction changes. The receiver may still be working every second, but the distance is added up from the thinner set of points.
The second one is the usual reason two devices disagree.
Why does a low sampling rate make distance read short? #
Because a straight line between two points on a curve is always shorter than the curve. Mathematicians call it a chord and an arc, and the gap grows fast as the points get further apart.
Take a 90° bend with a 50 m radius, driven at 36 km/h (10 metres per second). The road around the bend is 78.5 m long. Here’s what different recording intervals measure:
| Recording interval | Distance between points | Distance lost on the bend |
|---|---|---|
| Every 1 second | 10 m | 0.2% |
| Every 2 seconds | 20 m | 0.7% |
| Every 5 seconds | 50 m | 4.1% |
| Once for the whole bend | 78.5 m | 10% |
At one point per second the loss is negligible. At one point every five seconds, a road full of bends like this one reads about 4% short.
Speed makes it worse, because at a higher speed the same interval puts the points further apart. The saving grace is that fast roads have gentle curves. A highway bend with a radius of several hundred metres loses almost nothing even at 100 km/h. The worst case is a moderately fast ride on a very twisty road: mountain switchbacks on a motorcycle, a technical descent on a mountain bike, or a 400 m running track, where GPS commonly reads short.
Why can a high sampling rate make distance read long? #
Every GPS fix is off by a few metres, in a direction that changes from one fix to the next. When you’re moving quickly, that error is tiny compared with the ground you cover between fixes. When you’re moving slowly, it isn’t.
Walking at 1.4 metres per second, you move less than the error in each fix. If one point lands two metres to your left and the next lands two metres to your right, the line between them can be several times longer than the step you actually took. Record ten points a second and that wobble gets added ten times over. Stand still at a traffic light and a naive app will keep adding distance while you go nowhere.
In practice phone receivers already smooth their output, so the effect is smaller than pure randomness would suggest. It’s still the reason good apps filter out tiny movements rather than add every one.
How do speed and terrain change the error? #
| Activity | Main risk | Usual direction |
|---|---|---|
| Walking or jogging in a city | Scattered fixes near buildings | Reads long |
| Running on a 400 m track | Corners cut between fixes | Reads short |
| Cycling a twisty trail under trees | Corners cut, weak signal | Reads short |
| Driving a highway | Very little | Close to true |
| Driving mountain switchbacks | Corners cut | Reads short |
| Boating in open water | Very little | Close to true |
| Sitting still with the app running | Drift | Reads long unless filtered |
Hills matter much less than most people expect. Most apps measure horizontal distance, so a slope adds a sliver of extra road that GPS doesn’t count. On a 10% grade that’s about half a percent. There’s more on it in our guide to why bike odometers and GPS differ on hills.
Is distance calculated from position or from speed? #
There are two ways to do it, and each has a weakness.
- Joining positions. This is what most apps do. It’s affected by corner-cutting and by scatter, as above.
- Adding up speed over time. GPS receivers calculate speed from the Doppler shift of the satellite signals, which is often steadier than position. Multiplying speed by time doesn’t cut corners, but any speed error and any dropout go straight into the total.
Some devices blend the two. Whichever method your app uses, a clean signal matters more than the maths.
How do you get more accurate GPS trip distance? #
- Record every second if your app or watch offers the choice, especially for twisty routes or track sessions.
- Give the phone a view of the sky. A dashboard or handlebar mount beats a jacket pocket, and a jacket pocket beats the bottom of a backpack.
- Wait for a good fix before starting. Standing outside for 20 or 30 seconds after opening the app lets it settle, and starting in a garage or under a bridge gives a bad first point.
- Use an app that filters drift, so traffic lights and coffee stops don’t add distance.
- Check it against a known distance. Measured highway markers, a 400 m track or a surveyed trail give you a reference. We explain how to do this for a car in trip odometer vs GPS distance.
How Speedometer GPS samples your trip #
Speedometer GPS asks the phone for its most accurate fix once a second and adds up distance from every one of them while a trip is running. It only counts distance while you’re moving, ignores movements under 5 m between fixes (10 m once you’ve been still for a while), throws out single jumps of more than 200 m in one second, and doesn’t use fixes with an accuracy worse than ±20 m. A pill under the speed shows the fix quality from Excellent to Poor, so you can see when the signal is good enough to trust.
That 5 m filter shows the trade-off this whole page is about. It’s tuned for vehicles. Below about 18 km/h (11 mph) you move less than 5 m between one-second fixes, so at walking, running and slow cycling speeds, and when crawling in traffic, the app’s trip distance can read short. The speed readout isn’t affected, because it comes from the receiver’s own speed measurement. On foot or on a slow bike ride, use the app for speed and a watch, bike computer or known course for distance.
When you save a trip, the app keeps the route for its map, speed chart and GPX export, trimmed to at most 500 points so the history stays light. The distance shown for the trip was counted from every fix while you drove. If you import a long trip’s GPX file into another app that recalculates distance from the points, it’s working from a thinned-out copy and may come out a little shorter on a winding route. Trust the app’s own figure for that trip.
The app reads location only while it’s open, so keep it on screen for the whole trip. Our battery guide covers keeping a phone going on a long day.
Frequently asked questions #
What sampling rate is best for GPS distance? #
One point per second is enough for walking, running, cycling and driving on almost any route. Faster rates help little with a phone’s accuracy, and slower battery-saving modes start to cut corners on twisty routes.
Why do my watch and phone show different distances for the same run? #
They often record at different rates, filter drift differently and sit in different places on your body. A watch in a battery-saving recording mode, for example, will usually read shorter on a route with lots of turns. A few percent of difference is normal.
Why does my GPS app add distance when I’m standing still? #
Each fix is a few metres off in a random direction, so a stationary phone appears to wander. Apps that add every movement count that wander as distance. Better apps ignore small movements and only count distance while you’re actually moving.
Is one GPS fix per second enough for driving? #
Yes. At highway speeds the curves are gentle enough that one-second sampling loses almost nothing. Tunnels and dense city centres cause far bigger errors than the sampling rate does.
Why does GPS measure a running track short? #
A 400 m track has two tight bends, and straight lines between one-second fixes cut across them. Scattered fixes near the stands can make it worse. Many runners use the track’s lap count, not GPS, for track sessions.