AI quick summary
- Consistency beats absolute accuracy — a meter reading 2% high every ride trains you perfectly; a meter that drifts between rides is useless for tracking progress
- Zero offset (the 'calibrate' function on your head unit) is a 10-second pre-ride ritual that zeroes the strain gauge. Do it before every ride, on a flat surface, with the pedals level
- True calibration (setting the slope) is a factory or lab process — most riders never do it. The zero-offset ritual is what keeps a meter consistent ride-to-ride
- Signs of drift: your numbers suddenly jump vs heart rate/RPE, or the left/right balance shifts. A recalibration or return is warranted
/ 01
Calibration vs zero offset — the key distinction
Most riders use the word 'calibration' loosely to mean the pre-ride ritual of zeroing the power meter. Technically, that ritual is a zero-offset (or 'tare') operation: it tells the strain gauge what zero load looks like, so it can measure torque accurately. It takes 10 seconds and should be done before every ride.
True calibration — setting the slope, the relationship between measured strain and output watts — is a factory or lab process. It's done once (at manufacture) and occasionally re-done (if the meter drifts). Most riders never perform a true calibration; they rely on the factory slope and the ride-to-ride zero offset to keep the meter consistent.
The reason this distinction matters: the zero offset is what keeps a meter consistent ride to ride. A meter with a slightly-off factory slope but a consistent zero offset will read (say) 2% high every ride — which is fine for training, because you're tracking changes in yourself, not chasing absolute truth. A meter with a perfect factory slope but a drifting zero offset will read inconsistently — which is useless, because you can't tell whether a change in your numbers is you or the meter.
/ 02
The pre-ride zero-offset ritual
Before every ride, on a flat surface, with the pedals level (cranks vertical, so neither pedal is loaded), trigger the 'calibrate' function on your head unit. Wait for the confirmation. This takes 10 seconds and is the single most important thing you can do for power data consistency.
Why it matters: temperature changes shift the strain gauge's zero point. A meter that was zeroed at 20°C will read differently at 10°C if not re-zeroed. Modern meters (especially the Favero Assioma and newer Garmin Rally units) auto-compensate for temperature, but the pre-ride zero offset is still the reliable baseline. Do it every ride, every time.
When to re-zero mid-ride: on long rides with big temperature swings (a mountain descent after a hot climb), a mid-ride zero offset can correct drift. Stop, unclip, level the cranks, re-zero. This is rarely necessary on rides under 3 hours or in stable conditions.
/ 03
Why consistency beats accuracy
For training, you are tracking changes in yourself — your FTP, your interval watts, your fatigue. A meter that reads 2% high every ride shows you those changes accurately (a 5% improvement in your FTP shows up as a 5% improvement in your numbers, even if the absolute watts are 2% high). The offset is invisible to the training signal.
A meter that drifts — reading 2% high one ride, 1% low the next — destroys the training signal. You can't tell whether your 'bad' ride was you or the meter. This is the failure mode that makes a power meter useless, and it's almost always a zero-offset or temperature-compensation problem, not an accuracy problem.
The exception: if you race on numbers (draft legal, where you need to know absolute watts to pace a breakaway) or you're rehabilitating an asymmetry (where left/right balance matters), absolute accuracy and dual-sided measurement matter. For everyone else, consistency is the whole game.
/ 04
Signs of drift — when to recalibrate or return
Sudden number jumps: if your power suddenly reads 5-10% higher or lower than usual, for the same heart rate and perceived effort, the meter may have drifted. Check the zero offset, re-zero, and recheck over a few rides. If the drift persists, the meter may need a factory recalibration or replacement.
Left/right balance shifts: on dual-sided meters, your left/right balance should be stable (most riders are 48-52%). A sudden shift to (say) 40/60 that persists across rides suggests a sensor problem in one side. Single-sided meters (which double one leg) can't detect this — which is why dual-sided meters exist for serious training.
Inconsistent torque values on the same climb: if you ride the same climb at the same effort regularly and the watts vary more than 3-5% for the same perceived effort, the meter is drifting. Time for a recalibration or a return.
/ Sources