How to calibrate your power meter (and how often)

A power meter is the one piece of equipment on your bike whose output you cannot sanity-check by feel. A slack headset announces itself; a meter quietly reading 5% low just tells you that you have got slower. So it is worth knowing what calibration actually does, which of the two procedures you are performing, and how often either of them is genuinely needed.

The short version: the thing you do before every ride is not calibration. It is a zero offset, and it fixes a different problem from the one most riders think it fixes.

Two procedures, commonly confused

Almost every head unit calls both of them "calibrate", which is where the confusion starts. They are not the same operation and they do not correct the same error.

The distinction matters because they fail in opposite ways. A drifting zero makes your power wrong by a roughly fixed number of watts at every intensity — 15 W high whether you are soft-pedalling or sprinting. A wrong slope makes it wrong by a percentage, so the error grows with the effort and your sprint numbers rot faster than your endurance numbers.

Temperature is the main villain

Strain gauges measure the tiny deformation of metal under load. Metal also deforms with temperature, and the electronics reading the gauges have their own thermal behaviour, so the unloaded baseline moves as the meter warms or cools. Reviews of the power-meter validity literature single temperature out as one of the conditions that most reliably alters what a meter reports, and the practical recommendation that falls out of it is simple: zero the meter once it has reached the temperature it is going to ride at, not while it is still carrying the warmth of the house.

What else moves the number

Calibration state is only one of the inputs. Several of these are documented in the same body of laboratory work, and knowing them saves you from chasing a phantom fitness problem.

How accurate is accurate?

Manufacturers generally claim something like ±1% to ±2%. Independent testing suggests that claim holds for precision more often than it holds for trueness. When 54 power meters from nine manufacturers were checked against a physics-based reference, mean deviation across the sample was around −1%, but the spread was wide enough that six units were out by more than 5%. Their repeatability — giving the same answer twice for the same effort — was much tighter, at roughly 1%.

That combination is the important insight. Your meter is probably a good stopwatch and a mediocre ruler. It will track your own changes faithfully while reporting a number that is not quite the same number someone else's meter would report. Which is fine, as long as you never compare your FTP with a friend's and conclude something about either of you.

Is it the meter, or is it me?

This is the question that actually sends people looking for a calibration guide. A few things separate a drifting meter from lost fitness, and none of them require new equipment.

• Check the shape of the loss. Lost fitness usually shows up as an effort you cannot hold as long. A calibration error shows up as an effort that feels the same and reads lower — same breathing, same cadence, same heart rate, smaller number.
• Use heart rate as the referee. It is a slow, noisy signal, but over several similar rides a stable heart rate against falling power is a strong hint that the power, not the rider, has moved.
• Watch for a step, not a slope. Fitness changes over weeks. Hardware changes overnight. If the drop appeared between two consecutive rides, suspect the kit — especially if you changed a battery, updated firmware or moved pedals between bikes in between.
• Compare against a second device. If you also ride a smart trainer, a familiar indoor session is a decent reference point. It will not match exactly — see why your indoor FTP is lower — but a sudden 20 W gap where there used to be a 5 W gap is informative.
• Do the static torque test. It is the only one of these that gives a real answer rather than an inference.

Trainers: spindown is the same idea in a different dress

A wheel-on trainer estimates power from the speed of the roller and a stored model of how much drag the mechanism produces. Tyre pressure, how hard the roller is pressed into the tyre and — again — temperature all change that drag. A spindown re-measures how long the drum takes to coast to a stop and updates the model, which is functionally a zero offset for the whole system. Do it after a proper ten-minute warm-up, with the tyre pressure you actually ride, and repeat it if you change tyres or clamping force.

Direct-drive trainers need it far less, and several models self-calibrate. They also hold their numbers well over time: one popular unit tested fourteen months apart, after regular use, showed a typical error between the two test sessions of about 3 W, or 1.6%. If your direct-drive trainer's numbers have moved noticeably, look at firmware, a loose skewer, or the possibility that your fitness is genuinely different before you blame the strain gauges. There is more on this in smart trainer power accuracy.

What to do with all this

Practically: zero offset before every ride at riding temperature, a static torque test once a season and after anything violent, and a settled acceptance that your meter's absolute number is your own private currency. Consistency is what training models need — and consistency is the thing power meters are genuinely good at.

That is also why it is worth keeping one meter for one purpose. Moveee builds your power profile and FTP estimate from your ride history, so a meter that quietly drifts halfway through a season doesn't just corrupt one ride — it bends the baseline that every subsequent session is prescribed from. If you do change or recalibrate a meter mid-season, expect the first couple of weeks of any plan to look slightly off while the new numbers settle.