POWER METER ACCURACY: WHEN YOUR NUMBERS LIE AND HOW TO FIX IT

Power meters lie. Not maliciously. Not all the time. But often enough that if you don't understand why, you'll spend weeks chasing phantom fitness losses — or worse, making training decisions based on data that isn't telling you the truth.

You did the same climb, same legs, same effort. Twenty watts lower. Your first instinct is that you've lost fitness. Your second instinct is to do another FTP test. Both instincts are wrong. The problem might be the thing measuring you, not the thing producing the watts.

This is the troubleshooting guide I wish someone had written ten years ago. Every source of power meter error, how to identify it, and how to fix it. Most of this is fixable. Some of it just needs to be understood and accounted for.

Zero-Offset Calibration: The Ten-Second Habit That Fixes Half of Everything

Let me be really clear about this: if you are not calibrating your power meter before every ride, you are not using a power meter. You are using an expensive random number generator.

A zero-offset calibration — sometimes called a "zero reset" — tells the power meter what "zero" looks like right now. No force on the pedals, no torque on the crank. This is the baseline from which every watt is measured. If that baseline is wrong, every single number in your ride file is wrong by the same amount.

Here's where it gets really interesting. The zero-offset value drifts. It changes with temperature. It changes with humidity. It changes when you tighten your pedals or adjust your crank bolts. It changes if the bike has been sitting in a cold garage overnight or baking in a hot car for two hours. The strain gauge is a physical thing, and physical things respond to their environment.

Hunter Allen and Andrew Coggan made this point years ago in Training and Racing with a Power Meter — the zero-offset is the single most overlooked source of error in power data. It takes ten seconds on your head unit. Garmin, Wahoo, Hammerhead — they all have a calibration option in the sensor menu. Use it.

The protocol is simple:

1. Clip in and unclip. Let the crank arm hang vertically under its own weight with no foot on the pedal.
2. Go to your head unit's sensor menu and select "Calibrate" for the power meter.
3. Wait for the confirmation number.
4. Ride.

If the number is dramatically different from your usual zero-offset value — say, more than 50 counts off from your norm — something else is going on. We'll get to that.

Temperature Drift: The Invisible Error You Cannot Feel

Temperature is the single biggest environmental factor affecting power meter accuracy. And it's the one most riders don't account for.

Every strain gauge has a temperature coefficient. As the gauge warms up or cools down, its resistance changes, and that change gets interpreted as force — even when there's no force being applied. Modern power meters have internal temperature compensation algorithms that try to correct for this, but the compensation is never perfect.

Here's what that looks like in practice. You calibrate in a 5-degree garage at 6am. You ride for forty minutes, the ambient temperature climbs to 15 degrees, and the power meter warms from pedalling. The meter's internal temperature has shifted 15-20 degrees from the calibration point. If the temperature compensation is off by even a fraction, you're looking at 1-3 per cent drift. On a 250-watt effort, that's 2.5 to 7.5 watts of error that you cannot feel and cannot see on a single ride.

Comparative testing — the kind that Shane Miller (GP Lama), Ray Maker (DC Rainmaker), and others have done extensively — shows that this drift is real and measurable. When they run multiple power meters simultaneously on the same bike, the divergence between devices often correlates with temperature changes during the ride rather than actual power production differences.

What you can do about it:

• Calibrate after the bike has been at riding temperature for five minutes. If you've just wheeled the bike out of a cold shed, ride easy for five minutes, then calibrate.
• On days with extreme temperature swings — a cold start warming into a hot afternoon — consider re-calibrating at a natural stop point.
• Be aware of the effect when comparing data across seasons. Your winter numbers and summer numbers may not be directly comparable if you're calibrating in vastly different ambient temperatures.

The good news: for most riders, in most conditions, temperature drift is small enough that it doesn't change your training decisions. But if you're chasing a specific FTP number or comparing data across months, it's worth knowing the error exists.

Battery Voltage: The Silent Data Killer

This one catches people every time. Your power meter is still turning on. It pairs with your head unit. It shows numbers. So it must be working, right?

Not necessarily. As battery voltage drops below the optimal range, the excitation signal to the strain gauge weakens. The meter still functions — it still measures something — but what it measures becomes progressively less reliable. You'll see it as noise: unexplained spikes of 50-100 watts on flat roads, sudden drops to near zero during efforts, or a slow drift in readings that doesn't correlate with anything you're doing on the bike.

Here's where most riders go wrong. They wait for the low battery warning on their head unit. But by the time that warning fires, the data has often been unreliable for several rides. The threshold for the warning is set at a voltage that keeps the meter operational, not at the voltage that keeps it accurate.

The fix is boring and it works: replace batteries on a fixed schedule. Every 150-200 hours of riding, or every 3-4 months, whichever comes first. For coin cell batteries (CR2032, commonly used in Stages, 4iiii, and others), that's a cost of about two quid and five minutes of your time. For rechargeable units (Favero Assioma, Garmin Rally), charge before every ride or at least every two rides.

If you're seeing erratic data and you suspect batteries, this is the first thing to rule out. Fresh battery, calibrate, ride a known route at a known effort, and see if the numbers stabilise. If they do, it was the battery. If they don't, keep reading.

The Single-Sided Assumption: When 50/50 Isn't 50/50

Single-sided power meters — Stages, 4iiii single-leg, Favero Assioma Uno — measure force from one leg and double it. The assumption is that your left and right legs produce equal power. Fifty-fifty.

The assumption is almost always wrong.

Most riders have a measurable left-right imbalance, typically in the range of 48/52 to 46/54. This is normal. It doesn't mean anything is wrong with you. It doesn't need fixing. But it does mean your single-sided power meter is systematically over- or under-reporting your total power by 2-5 per cent, depending on which leg is measured and how asymmetric you are.

Let's put numbers on it. Say your true total power output is 300 watts. Your actual split is 46/54 — your left leg produces 138 watts and your right produces 162 watts. A left-side-only meter reads 138, doubles it to 276. That's 24 watts below your actual output. Every ride. Every interval. Every test.

Here's the critical distinction: this error is consistent. If you test your FTP on a single-sided meter and train with the same meter, your training zones are internally correct. You'll still be training at the right intensity. The numbers might not match what a dual-sided meter would say, but that doesn't matter — you're measuring from the same reference point every time.

Where it breaks down:

• Comparing your power to someone else's. Their meter measures differently. Their leg balance is different. Raw watt comparisons between riders on different equipment are meaningless.
• Swapping between a single-sided meter and a trainer. Your trainer measures total power. Your single-sided meter estimates it. They won't agree.
• After an injury or change in bike fit. If your leg balance shifts — which can happen after a knee injury, a cleat change, or a saddle height adjustment — your single-sided meter will reflect that shift as a power change, even if your total output hasn't changed.

If you own a single-sided meter, it's a perfectly good training tool. Just understand what it's measuring and what it's assuming.

Indoor vs Outdoor: Why Your Trainer and Your Power Meter Disagree

This is possibly the most common source of confusion in cycling data. You ride indoors at 250 watts. You go outside and the same effort reads 265 watts on your crank meter. Or vice versa. Something must be broken.

Nothing is broken. The discrepancy is physics.

A smart trainer measures power at the rear axle — after your chain, cassette, derailleur, and bottom bracket bearings have all extracted their share of friction. A crank or pedal power meter measures before those losses. The difference between the two — called drivetrain loss — is typically 3-8 per cent, depending on chain condition, lubricant, derailleur alignment, and temperature.

That means if your crank meter reads 260 watts, the trainer might see 245-252 watts of that arriving at the rear axle. Both numbers are correct. They're measuring at different points in the system.

There are additional factors:

• Tyre pressure and roller contact. Wheel-on trainers are particularly sensitive to tyre pressure and how firmly the roller contacts the tyre. Small changes produce measurable power differences.
• Calibration state of the trainer itself. Most smart trainers have their own spindown calibration. If you haven't done one recently, the trainer's power reading may have drifted.
• Heat. Trainers generate significant heat during long sessions. Internal resistance in the brake mechanism changes as the unit warms up. Some trainers compensate for this better than others.
• Power smoothing. Some trainers apply aggressive smoothing to the power signal, which affects how power is reported moment to moment — though it shouldn't change the average over a full interval.

The practical solution is simple: pick one power source for indoor training and stick with it. If you use your trainer's power for indoor sessions, test your FTP indoors and set your indoor zones from that number. If you prefer your on-bike meter's numbers, use ANT+ or Bluetooth to transmit from the meter to your training app and ignore the trainer's reading. Don't mix sources. Don't try to apply a correction factor. Just be consistent.

If you are specifically trying to compare indoor and outdoor performance — say, to check whether your indoor FTP and outdoor FTP are actually different or just a measurement artefact — the best approach is to ride a sustained effort at the same target RPE and heart rate both indoors and outdoors, using the same power meter, and compare.

Firmware Updates: The Hidden Variable

This is the one that will make you properly paranoid, and rightly so.

Power meter manufacturers push firmware updates that can change how your power is calculated. Not the hardware — the software that interprets the strain gauge signal. Temperature compensation curves, signal filtering, cadence detection, zero-offset algorithms — all of these live in firmware and all of them affect the final watt number.

Most of the time, firmware updates improve accuracy. But "improve" means "change," and a change means your numbers shift. If you've been training with a meter that consistently read 3 per cent high due to a firmware quirk, and the update corrects that, your FTP just dropped by 3 per cent overnight. Your fitness didn't change. Your training zones didn't change. The number changed.

Favero, Garmin, Stages, 4iiii, SRAM — they've all pushed updates over the years that affected reported power. Some were documented in release notes. Some weren't.

How to handle this:

• Read the release notes before updating. If the notes mention calibration changes, power calculation adjustments, or temperature compensation improvements, treat the update as a potential FTP-altering event.
• Re-test after major firmware updates. Run a benchmark effort on a known climb or a controlled indoor test. Compare to your recent data. If there's a meaningful shift, adjust your zones.
• Don't update the morning before a race. Obvious, but people do it.
• Keep a record of firmware versions. When you look back at your training data in six months and wonder why there was a sudden shift in August, the firmware log will save you hours of confusion.

Crank-Based vs Pedal-Based vs Hub-Based: Accuracy Differences That Actually Matter

Not all power meter types are created equal in terms of where they sit in the measurement chain, and that affects what they're actually measuring.

Crank-based (Stages, 4iiii, Quarq, SRM): Measures deflection of the crank arm or spider under load. Well-proven, good temperature stability in high-end units (SRM is still the reference standard for a reason), and positioned early in the drivetrain. Spider-based meters measure total power directly and are inherently dual-sided. Arm-based meters are typically single-sided.

Pedal-based (Favero Assioma, Garmin Rally): Measures force at the pedal spindle — the earliest possible measurement point. The advantage is portability between bikes. The challenge is that pedal meters live in a mechanically complex environment — pedal bearings, cleat float, and installation torque all affect the measurement. Favero has established a reputation for exceptional consistency at a reasonable price. Garmin's Rally series has improved significantly with firmware updates.

Hub-based (PowerTap): Measures at the rear hub, after all drivetrain losses. This gives you the truest picture of what's actually reaching the road, but the trade-off is that any efficiency gains from a new chain or better lube won't show up in the power number. Hub-based meters are less common now but were the benchmark for total power accuracy for years.

Smart trainers (Wahoo KICKR, Tacx NEO, Elite Drivo): Measure at the resistance unit. Modern direct-drive trainers are accurate to within 1-2 per cent, but they're measuring something slightly different again — power absorbed by the trainer's braking mechanism, corrected for flywheel inertia.

The practical takeaway: the differences between meter types are smaller than the differences caused by poor calibration habits. A well-calibrated Stages will give you better data than a poorly calibrated SRM. Choose your meter based on budget, convenience, and bike compatibility — then calibrate it properly.

When to Ignore Your Power Data and Ride by Feel

Here's a sentence you won't read often on cycling data websites: sometimes the right thing to do is ignore your power meter entirely.

Power data is a tool. It gives you precision and repeatability. But it can also become a cage — a number you chase at the expense of listening to the actual signals your body is sending.

Situations where RPE and heart rate should override power:

• You're ill or coming back from illness. Your power output will be reduced but your RPE at any given wattage will be elevated. The power meter doesn't know you have a cold. Ride to RPE.
• Extreme heat. Cardiac drift will push your heart rate up while your power output drops. The power meter says you're underperforming. Your body says it's 38 degrees and you're doing exactly what you should be doing.
• Late in a long event. After five or six hours, the relationship between power and RPE shifts. Holding your target watts starts costing more perceived effort. If you're locked to the number, you'll blow up. If you're reading your body, you'll adjust.
• Early season or after a break. Your FTP may be temporarily depressed. Riding to stale power zones will either be too easy or too hard. Ride by feel for two to three weeks, then retest.
• When the data looks wrong and you can't fix it mid-ride. If you've noticed erratic numbers and can't diagnose the issue on the road, just ride. Log the ride for heart rate and RPE analysis later. Bad power data is worse than no power data because it leads to bad decisions.

The ability to ride without data — to know what threshold feels like, to sense when you're in Zone 2 versus Zone 3, to pace by breathing and muscle sensation — is a skill. It's a skill that gets weaker the more you depend on the screen. Practice riding without the data page occasionally. It'll make you a better cyclist when the meter works and when it doesn't.

How to Spot a Dying Power Meter

Power meters don't usually fail catastrophically. They degrade. The readings get noisier. The zero-offset drifts further between calibrations. The data starts telling a story that doesn't match what your legs and heart rate are saying.

Warning signs, roughly in order of seriousness:

1. Increasing zero-offset variability. Your zero-offset value should be fairly consistent day to day (within 20-30 counts on most units). If it's jumping around by 100+ counts between rides, something is changing mechanically or the strain gauge is degrading.
2. Unexplained power spikes. Sudden jumps of 200-500 watts on flat, steady riding — not sprint efforts, not accelerations, just phantom spikes that don't correspond to anything physical. This is usually a failing strain gauge or a connection issue.
3. Progressive drift during rides. Power that slowly climbs or falls over the course of a ride without corresponding changes in effort, heart rate, or terrain. The meter is drifting and not self-correcting.
4. Dropouts. Intermittent loss of signal — zero-watt readings for a few seconds followed by a return to normal. Can be a battery issue, but if fresh batteries don't fix it, it's likely a communication or sensor failure.
5. Gross inaccuracy against a known reference. If your crank meter suddenly reads 15 per cent different from your trainer on a controlled effort — and it didn't used to — something has changed in one of them.

If you've ruled out batteries, calibration, firmware, and environmental factors, and the problem persists across multiple rides, the meter likely needs servicing or replacement. Most reputable manufacturers offer warranty support and will either repair or replace a failing unit.

The Diagnostic Checklist: When Power Feels Wrong

Print this. Stick it on the wall of your pain cave. Run through it before you start questioning your fitness.

Step 1 — Calibrate and re-ride. Zero-offset the meter. Ride a known effort — a climb you've done fifty times, a flat segment at threshold, a controlled indoor test. Compare to recent data from the same effort.

Step 2 — Replace the battery. Fresh CR2032 or a full charge on rechargeable units. Calibrate again. Ride the same effort.

Step 3 — Check for firmware updates. Go to the manufacturer's app. If there's an update, read the notes. Update if appropriate. Calibrate. Ride.

Step 4 — Compare against heart rate and RPE. On a sustained effort — ten minutes at threshold, say — your heart rate and perceived exertion should align with power in a way that's familiar to you. If power says 250 watts but your heart rate is where it usually is at 280 watts and you feel like you're at 280 watts, the power meter is the problem.

Step 5 — Check the hardware. Is the crank arm tight? Are the pedals properly torqued? Is there any visible damage to the strain gauge housing? Is the battery compartment seal intact and free of moisture?

Step 6 — Dual-reference test. If you have access to a second power source — a trainer, a second meter, a friend's meter you can borrow — ride the same effort on both simultaneously. A consistent offset is normal. An erratic or changing offset suggests a fault in one of them.

Step 7 — Contact the manufacturer. If Steps 1-6 don't resolve the issue, the strain gauge may be degrading. Most power meters carry a two-year warranty. Even outside warranty, manufacturers often offer discounted service or replacement for units with confirmed sensor failure.

The Bottom Line

Power meter accuracy is not a binary state. Your meter is not simply "accurate" or "inaccurate." It's producing data that's shaped by calibration habits, temperature, battery state, firmware, measurement position, and the inherent assumptions built into the device. Understanding these factors doesn't make you paranoid — it makes you a better user of the tool.

Calibrate before every ride. Replace batteries on schedule. Read firmware release notes. Pick one power source per environment and stay consistent. And when the numbers don't match how you feel, trust the process of elimination before you trust the screen.

Most power problems are fixable in five minutes. The ones that aren't are still diagnosable. And the rare case where the meter is actually failing? That's what warranties are for.

If you want to talk through your own power data — whether something looks off, how to set up your tracking properly, or just to get a second opinion on what the numbers mean — the Roadman Cycling community on Skool is where those conversations happen daily. Coaches, physiologists, and riders who've been through every one of these issues. Come ask.