Guides · Treadmill error over a long run
How much does a small treadmill error cost over a long run?
You finish the 18 miler, the screen ticks over to 18.00, and you drop it into the week’s total. That number is printed to two decimal places, which is not the same thing as being right.
A treadmill belt running two percent slow turns a screen showing 7.0 into a real 8:45 mile instead of 8:34, and it turns your 18 into about 17.6. Ignore that once and you lose nothing. Ignore it for a training block and it piles up, because the error lands the same way every session instead of scattering.
The screen is not reporting on the belt
You pick a speed, the machine works to hold it, and the display shows you what you picked. Some machines check a spinning part under the deck and some check nothing at all, but none of them watch the surface under your feet. That is a design choice rather than a fault.
So the treadmill is not lying to you. It answers the question you asked, and it has no way to answer the one you did not.
A steady error is not a small error
Runners are used to numbers that wobble. Your watch loses satellites under a bridge and hands back a mile that was really 0.97. The next one comes back at 1.02. Across a two hour run that kind of error mostly cancels itself, and you learn to trust the total even when you distrust any single split.
A miscalibrated treadmill does not behave like that. It is not scattering around the truth. It is sitting to one side of it, in the same spot, every time you step on.
I have checked that directly. I put a handheld tachometer on the same treadmill twice, nine days apart, and all three speeds came back within six tenths of a percent of the first set of readings. That machine’s error was not wandering. It was sitting still, because it belongs to the machine rather than to the run.
Picture a mile marker painted in the wrong place on a loop you run every week. Every lap is short by the same few yards. Twenty laps do not fix it and a hundred do not either, because the mistake is built into the loop and each lap inherits the whole of it.
That gap between wobble and offset is the reason a couple of percent is worth your time. Wobble gets smaller as you add miles. An offset gets bigger.
The same 18 mile long run, at four amounts of drift
The screen reads 7.0 miles per hour the whole way, which it calls an 8:34 mile.
| If the belt runs | Your real mile | Ground you covered | Missing |
|---|---|---|---|
| 1% slow | 8:39 | 17.8 miles | 0.2 miles |
| 2% slow | 8:45 | 17.6 miles | 0.4 miles |
| 4% slow | 8:56 | 17.3 miles | 0.7 miles |
| 6% slow | 9:07 | 16.9 miles | 1.1 miles |
Worked arithmetic on typical amounts of drift, not machines I measured. The rule travels to any pace: a belt one percent slow adds one percent to every split you run and takes one percent off every mile you thought you covered. In lab tests, four treadmills ran their belts between 98 and 103 percent of the set speed, so the first two rows sit inside that range. The 4 and 6 percent rows sit outside that range. They are here because the widest gap I have personally run into was wider still.
Now run it forward. Sixteen weeks at fifty miles a week is eight hundred miles on paper. At two percent it is about sixteen fewer on the ground. You taper believing your legs have a full block in them, and they have one long run less.
The half that costs you on race day
Distance is the part people notice. Pace is the part that picks your race.
If the screen said 8:34 and the belt was two percent slow, every one of those long runs actually happened at 8:45. You did the work. Your log is what you used to choose a goal, and it was never the pace you actually held.
A build full of 8:34 long runs points you at 3:44 on race day. The pace you have really rehearsed for four months is 3:49. You trained honestly. You picked a number you had never run, off a screen that was never measuring.
I have run a machine whose display read five miles when I had really run four, so I definitely under-ran my marathon training. That is a wider gap than anything in the table above, and it is the reason this page exists.
What you can actually do about it
One interval on a machine still compares fairly with the next interval on that same machine, because the machine’s error is the same both times. The absolute pace is the part you cannot carry outside. A foot pod or a known outdoor route can tell you something is off, but neither can tell you whether it is you or the belt. Adjusting the machine is not the answer either. Most gym treadmills are not yours to touch, and on one you own that work belongs to your manual or a technician.
None of those give you the belt’s real speed, and that is the number every line on this page is built from. Treadwell reads the moving belt directly. You hold an iPhone Pro over the surface for a few seconds and its camera and depth sensor watch the belt itself, with nothing stuck to the machine and nothing typed in. Nothing in the answer came from the treadmill, so there is nothing for the treadmill to be wrong about.
On a hotel treadmill I had never seen and had never calibrated, the app’s reading matched my handheld tachometer within half a percent. Because the error is steady, you only need to measure it once. Check your own machine and you learn which way it leans and by how much, and from then on you set the dial to the pace you actually meant to run. When the reading is not clean, the app says so instead of inventing a figure.
Find out which way your machine leans
The belt’s real speed in a few seconds, measured directly. Treadwell needs an iPhone Pro.
Quick answers
How much does a small treadmill error cost over a long run? A belt running two percent slower than the screen turns a displayed 7.0 miles per hour into a real 8:45 mile instead of 8:34, and an 18 mile long run into about 17.6 miles of actual ground. At four percent the same run lands near 17.3 miles and every mile is over twenty seconds slower than the screen claimed.
Does a treadmill error average out if I run enough miles? No, and this is the part that catches people. A miscalibrated belt is not noise that scatters either side of the truth. It is an offset that lands the same way every session, so running more miles multiplies it instead of cancelling it. Random error shrinks with volume. A steady offset grows with volume.
Is two percent enough to worry about? For one easy run, no. For a training block, yes, because two percent changes both the pace you ran and the ground you covered, and it does it in every session rather than once. Sixteen weeks at fifty miles a week is about sixteen miles you logged and never covered, and a goal pace about ten seconds a mile quicker than anything you actually held.
Does a treadmill error affect my pace or my distance? Both. A slower belt means you covered less ground in the time you ran, so your real pace and your real distance move together. Whether the distance on your screen shows that gap is a separate question, and nothing on the machine tells you. I have run on a machine whose display sat well above the ground I actually covered.
Sources.
- That some machines check a spinning part while others check nothing at all comes from manufacturer service documentation: Precor’s Experience Series service manual, part 20039-180 rev A05, describes a motor design needing no speed sensor at all, and Horizon and Matrix documents on Johnson Health Tech’s own domain describe a magnet or an optical disc turning near the motor or the front roller. Each is scoped to those product lines rather than to a brand.
- That none of them watch the belt’s top running surface is our own reading of every motorized treadmill manual we could obtain. It is an absence across the documents we hold, not a manufacturer statement, and we do not claim it covers every machine ever built.
- The 98 to 103 percent figure is from Willwacher et al. 2021, Scientific Reports 11:2244, which measured average belt speed on four treadmills with seven people walking and running on them. Those were laboratory machines, two of them force-instrumented research treadmills.
- Every row of the drift table, the eight hundred mile block and the marathon times are worked arithmetic from a display reading 7.0 miles per hour. They are illustrative examples of typical drift, chosen as ordinary cases, and none of them is a machine we measured.
- The repeat tachometer check is ours. We measured the same treadmill on two days nine days apart, and the three settings we tested read within six tenths of a percent of the first set of readings. That is one machine over one short window, and we offer it as evidence that a machine’s error holds still rather than as a figure that describes treadmills in general.
- The display reading five miles against a real four is Bryan’s own run on a machine he trains on, reported first hand. It is our widest recorded gap rather than a typical one.
- The half a percent is ours, measured on a hotel treadmill we had never seen and had never calibrated, against a handheld tachometer taken the same day. It describes the accuracy of our own reading on that machine. It is not a claim about whether that machine’s display was right, and it is one machine rather than a range.