How to: Troubleshooting Power Meter and Trainer Accuracy Issues


No matter the product, eventually, for every product review I post, sooner or later someone has a power related accuracy issue with their unit. In today’s post, I’ll be focusing on power meter and trainer accuracy issues. But perhaps down the road I’ll expand it to other categories.  Most of this post is aimed at helping you figure out if there’s an issue in the first place, and if so – whether or not you’re able to fix it.  I also cover how I go about validating accuracy and the tools I use.

All in all, there honestly isn’t a ton of complexity in any of this. It’s more about process of elimination and step by step troubleshooting than some magical wizardry.  And sometimes, the answer is simply that the unit in question is inaccurate. Either systematically (all units by that company), or simply as a one-off (your specific unit).

Generally speaking, it’s actually super rare these days to find systematically inaccurate power meters or trainers.  There are some well known ones however, even this year, by well known companies.  Instead, most of the time people have accuracy issues it’s due to either manufacturing error or consumer error (or, unfortunately, an misunderstanding of one’s existing power capabilities).

Finally, while it’s easy to assume (and post on the internet) that ‘everyone is having this problem’, in 99.999% of the cases, they aren’t. Typically speaking people that are working just fine don’t post about it, and those with issues post the loudest and most frequently (even under different usernames as I’ve discovered). That doesn’t mean they aren’t having real issues, but one should often remember just how many units are being shipped every week (often in the range of hundreds to thousands of units), and how many comments one sees online. If everyone unit of a given type was broken, then frankly you’ll see hysteria. Unless of course, nobody is buying that unit.

To Begin – Power Accuracy Primer:

I’m not going to get into how power meters or trainers determine power at a strain gauge level, quite frankly – that doesn’t much matter here. Ultimately, either it’s accurate or inaccurate.  However, there are some important aspects to look at when troubleshooting.

Generally speaking for power meters, there’s strain gauges involved. At the most simplistic level these gauges measure the force as torque, and then combined with cadence can determine your power (wattage).  In today’s power meter world, the vast majority of units use accelerometer-driven cadence.  While years ago magnet based cadence (where something would physically pass by a magnet) was considered superior, that tends not to be an issue with today’s established brands. But more on that later.

The gauges are placed in a variety of locations on different power meters, but here’s the rough rundown:

Rear Wheel: PowerTap hubs
Bottom Bracket: ROTOR INpower, Team Zwatt Zpindle, Race Face/Easton
Crank Spider: Quarq/SRAM, Power2Max, SRM, PowerTap C1, Team Zwatt Zpider, FSA (Power2Max OEM)
Crank Arms: Rotor, Stages, Pioneer, 4iiii, WatTeam, ROTOR 2INPower, Team Zwatt Zimanox, Shimano, XCadey, Magene, Cateye (4iiii OEM), Avio, Verve Infocrank
Pedals: Garmin Vector, PowerTap P1/P2, SRM EXAKT pedals, Polar/Look combo, Favero Assioma, Favero bePRO, Look, Xpedo
Pedal Spacer: IQ2, LIMITS
Non-direct Force Power Meters: PowerPod, iBike, Arofly, PowerCal

Why does location matter? Because losses.

Anytime you transfer energy you have losses of that energy.  From a cycling power standpoint, the truest output of your body’s wattage-focused energy is directly at your foot.  The further away from the foot you get, the more losses you have. That can be through things like dirty drivetrains or bendy materials. Like playing a game of telephone, each step along the way you lose information.

Taking a look at the above photo, the pedals would typically show the absolute highest value of power.  Next would be be crank arm, then the spider.  In reality, there’s very little loss between the pedals and the chainrings/spider.  So that’s essentially negligible.  Where losses are most definitely real is between the crankset and the cassette (wheel hub in the picture above).  And that’s ultimately where most of the losses will occur.

If you’re evaluating a trainer for example, you’d need to account for this. I generally budget about 2-4% drivetrain losses, but losses of upwards of 5% plus aren’t unusual with dirty chains.  Meaning that I’ll lose 2-4% of my wattage when comparing something at the crank-arm versus something at the wheel hub/cassette (like a trainer).


There used to be a site called Friction Facts that covered a lot of this in-depth, but then Ceramic Speed bought them and the data all seems less visibly found.  However, past news articles about the site outline most of the key stats you need to know. And Ceramic Speed does have some good articles left as well.  Additionally, other 3rd party sites like Cycling Power Lab has more stats on this too.

These losses are very real, and super important to understanding accuracy – especially on trainers.

Power Meters:

For the most part these days, power meter accuracy is pretty good – both in terms of what marketing specs says it should be, but also real-world usage of how accuracy actually looks.

Not only that, I definitely don’t subscribe to the theory of: ‘Race to the bottom’, when it comes to power meter accuracy.

That’s simply elitist talk, usually from power meter incumbents charging more than their younger competitors. But in reality, some of the biggest blunders we’ve seen this year in terms of accuracy have come from giants (in market size or history/longevity in the industry).  And on the opposite side, what I’d argue two of the most accurate and versatile power meters has come from startup companies in the last couple years.

Which isn’t to say that all startups are good. No, hardly so. A typical power meter research and development cycle is usually 2-4 years long – small and big companies alike.  I’m highly skeptical of any company that drops onto the power meter scene via crowd funding with no power meter experience. History shows us that it doesn’t end well. At least until they prove themselves with data gathered and published by independent sources.

But let’s pretend your past all that and have a supposedly reputable power meter that you’re having trouble with.  Here’s how to start troubleshooting.

(Oh, side note: That way I’ve formatted this piece is to give you power meter and trainer specific setup/quick steps first, and then I cover how to do tests, capture data, and then analyze the data for the most common issues. Those analysis sections are down below later on.)

Step 1: Update your firmware

Many power meter accuracy issues can be solved by simple firmware updates using the manufacturers smartphone apps. Most accuracy issues can be quickly addressed this way and usually are even noted on the company’s firmware update page.  This is especially notable if you have an older Stages power meter, where they’ve increased transmission power in firmware updates over the last year or two. While that doesn’t directly impact accuracy, it can indirectly due so by eliminating dropouts that cause overall power average issues.

2018-02-15 13.23.51

2018-02-15 13.37.29

2018-02-15 13.32.39

This is especially important for newer power meters where firmware tends to iterate quickly to resolve issues.  For example we saw this with Garmin Vector 3 this past year where firmware updates addressed dropouts and related conditions. Again, like Stages these weren’t specifically accuracy-driven, but contributed to ensuring the data was at a quality bar that was acceptable.

Step 2: Ensure it’s installed correctly:

I know, I know, you followed the instructions.

Or, you might not have.

Here’s the thing about power meters: They’re fickle when it comes to install.

Specifically, they’re fickle if they’re not nice and snug. While today’s power meters are getting much easier to install, the reality is if you’ve got a spider based power meter and the chainring bolts are not tight, you’ll get inaccurate numbers. Same goes for a pedal based power meter that’s not initially snug (it’ll eventually tighten itself though).  And while WatTeam installs are no longer an issue (out of business last month), installation was shown as critical to success. Same goes with the SRM EXAKT pedals, where installation is super finicky and needs to absolutely be nailed to get accurate results. In that case, a millimeter really does matter.

Finally, while not applicable to newer Garmin Vector 3 power meters, remember that the Vector 1/2 power meters should have been installed with a torque wrench to get the torque correct. In my older reviews of those I showed what happens when it’s not.

Step 3: Do some hard sprints, or a ride or two

Before you sound the alarms on accuracy, throw down some sprints. I prefer to do this on a trainer, since if I’ve hosed up installation somehow from a mechanical standpoint I don’t plant my face into the ground.

But why sprints?

They help to ‘bed’ the device. Which means that they get all the parts nice and snug. Most devices (and manufacturers) would agree that a simple set of 2-4 nice hard 10-15 second sprints are more than sufficient to achieve this. I can’t think of a recent power meter that needs more than that, though sometimes you’ll see things stabilize after a ride or two. I roughly remember one or two iterations ago the ROTOR power meters were like that, as were some Quarq units.

Step 4: Do a zero offset

I know this sounds silly and common – but seriously, it’s the most important thing you can do for your power meter. And you should be doing it before every ride. Most power meters will give a value back to you. The exact value isn’t comparable between brands, but it’s an offset number. For most brands this number will usually shift with the outside temperature. If this value shifts significantly day to day and nothing else has changed, that’s generally a warning sign something is amiss.

The same is true if you do 2-3 zero offsets back to back within a 60 seconds span. If you see this number drifting then, something is also amiss. Stages actually does cool stuff in that their head units track these values automatically (even for non-Stages power meters). I really wish other companies would follow this approach, as its super helpful for both athletes and coaches.

Step 5: Change the batteries

If you have a coin-cell driven unit, one of the first and obvious signs of a dying battery is poor accuracy. Usually this renders itself via dropouts, but sometimes at low-voltage levels you just get general badness. As soon as you suspect power accuracy issues, just go get a clean pair of batteries and see if live returns to normal.

Plus, if you buy coin-cell batteries in bulk, then spending the 20 cents or so on a new pair is the most ROI-efficient power meter accuracy fix you’ve got.

Step 6: Static weight test

While yesterday I noted that by and large static weight tests aren’t useful in 2018 for determining accuracy, it was pointed out that they can be useful in some level of troubleshooting. Specifically, they can establish a baseline that at a minimum your power meter can correctly identify how much weight is being applied to it.

A static weight test is when you take a known/calibrated weight and hang it from your power meter. For power meters that support this type of test, they’ll display a given value and allow you to calculate whether that value is correct.  For a dual-sided unit, you’d want the value to be identical on both sides (and thus why technically speaking you could use any weight to simply validate both sides are the same).

The reason why I note this test isn’t super useful these days is that most accuracy issues simply don’t manifest themselves in a way that would be revealed with this test. Instead, the most common accuracy issues are related to temperature drift, road conditions, and bad firmware/algorithms. But, I dive into all those nuances down below.

Ok, so you’ve done all those things, and your power meter is still not right? Sorry. That means it’s time to collect some data and dig into what kind of issue you’ve got exactly. Skip past the trainer section and continue on with data capture, analysis, and hopefully resolution.


The very first thing to know when coming to trainers is setting expectations. Most trainers have ratings for accuracy on them, and generally speaking the cheaper the trainer the less accurate it’ll be.  For example, let’s look at the Wahoo trainer lineup briefly:

Wahoo KICKR SNAP Series ($599): +/- 3%
Wahoo KICKR CORE ($899): +/- 2%
Wahoo KICKR Series ($1,199): +/- 2%

Trainers for Tacx, Elite, Kinetic, CycleOps and everyone else generally mirror these accuracy and price ranges, though below $600 and you often find +/- 5% trainers as pretty common.  And in some cases once you drop down below $500, the accuracy range is +/- 10% – which is frankly a crapton of variance.

The next thing to pretend to care about is whether or not it has a power meter in it. I say ‘pretend’, because quite frankly it doesn’t matter.  What matters is whether or not it can produce accurate power.  Wahoo learned this lesson back a few years ago when they included a power meter in their KICKR and found it didn’t hold up to shipping well (damages). When they removed it, accuracy issues largely went away.

Other trainer companies do other things, but again, I haven’t seen any evidence that it matters whether or not there’s a power meter in there. Trainer companies are perfectly capable of making power-meter less trainers.  For example, the Tacx Neo series – widely regarded as one of the two most accurate trainers out there – does not have a traditional power meter in it (instead, they measure the current in the coils). While the Elite Drivo series (the other of the two most accurate units), does have a power meter. Both produce excellent and repeatable results – just achieved different ways.

Step 1: Update your firmware

The vast majority of trainer accuracy issues can be solved by simple firmware updates using the manufacturers smartphone apps. Most accuracy issues can be quickly addressed this way and usually are even noted on the company’s firmware update page.

This is especially important for newer trainers where firmware tends to iterate quickly to resolve issues, often accuracy related ones.

Step 2: Pumping and Spindown:

So, you’ve got an unexpected power data result? Here’s the run-down:

1) Ensure wheel is properly inflated (wheel-on trainers only)
2) Ensure press-on knob is sufficiently tight (wheel-on trainers only)
3) Warm-up for 10-15 minutes
4) Do a calibration spin-down/roll-down (all trainers except Tacx NEO)

Not sure if you have a wheel-on trainer? If you leave your wheel on while on the trainer, it’s wheel-on. Whereas if you remove the wheel, it’s direct drive. At present there are no direct-drive smart trainers below $699.



One of the main reasons everyone is moving to direct drive trainers is that it removes a very significant source of inaccuracy: tire pressure.  If your tire pressure has changed since the last time you did a calibration, your results will be off. Of course, your tires are always slowly deflating – thus you have to either ensure your wheels are pumped up to the exact same pressure each time, or, recalibrate.

Additionally, wheel-on trainers tend to have roller, wheel, and sometimes even flywheel (fluid ones) that takes time to warm-up. In most cases this is achieved in about 10 minutes, but I’ve seen other trainers take at least 15 minutes (the CompuTrainer is actually such one).

The very first thing you need to do is to ensure you’ve done a spin-down or roll-down, which is effectively a calibration of your trainer. All companies do this slightly differently, and for most trainers it can also be achieved via 3rd party apps like Zwift and TrainerRoad:

This requires you spin up to a given speed (usually about 22-23MPH), and then coast down. The trainer measures the time it takes to coast and determines an offset of sorts that’s applied to the power algorithm.  This is especially critical for wheel-on trainers.

If you suspect power issues, I’d *strongly* recommend using the company’s own calibration app, rather than one from 3rd party apps like TrainerRoad or Zwift.


Because many times the trainer company itself will have high/low spindown time limits in the app that will flag something is amiss, whereas 3rd party apps generally don’t have that. So you wouldn’t see an error that you might see in the native Wahoo/Tacx/Elite/etc.. app.

Now, I know that in theory warm-up isn’t required for most direct drive trainers, but I’ve also been doing this long enough to know that it tends to for certain companies.  Thus, just do it before you begin The Great Power Accuracy Hunt.

Step 3: Tweaking settings:

The next thing you need to do if you have a Wahoo trainer is to turn off ERG mode smoothing.  While this makes for super pretty graphs, it also obscures the actual data we’re trying to evaluate.  How can you tell if ERG mode smoothing is on? Check out the below graph that a DCR reader sent me this morning:


It looks beautifully smooth. It’s also fake.  It’s actually not what the trainer was really doing, so…turn it off. You can do so in the Wahoo app settings:

2018-09-10 14.50.00

2018-09-10 14.50.29

Meanwhile, if you have an Elite Direto for example, you want to make sure the wheel circumference is correct as well. This thread from Elite discusses it in more detail.

There may be more nuances to other trainers out there – but those are the two most common ones I see by far. If there’s others specific to a given trainer to add, I’ll be happy to add them in here if folks remind me of one in the comments.

How to Test:

There’s two ways to do tests: Indoors and outdoors.


But, I’d strongly recommend starting indoors. Obviously, if you’re on a trainer – then you’ll likely be indoors. But for determining general power meter accuracy, making it work indoors is a good step before heading outdoors.

While it’s easy to just do a generic ride on Zwift and then try and do analysis on it – it’s actually much easier to see trends (and then potentially determine the source of the problem), if you have a bit of structure to your workout.  In particular, ERG style workouts are among the easiest things to identify issues in.  I tend to use two core ERG workouts, one on Zwift, and one on TrainerRoad, to look at power accuracy issues.  The first on TrainerRoad is a 30×30 workout where I alternate a very high wattage with a very low wattage.  However, this tends to focus on trainer responsiveness, and not so much pure accuracy (though sometimes I see accuracy issues).

For example, the above graph shows an issue where for the first few sets the trainer was inaccurate at the beginning of each spike. This is because of gearing in ERG mode.  In fact, TrainerRoad warns about this, and how to increase accuracy and stability in sets.  They go into great detail on it here.

Next is a test I tend to do on Zwift using their workout mode – which is the very simplistic ‘Jon’s Mix’ workout. If you had to do one test on limited time, this is a good start. This has a few nice long chunks, as well as some sprints in it:


The thing that’s nice about the long chunks (upwards of 10 minutes or longer) is that you can typically see drift-related items quickly. Drift related issues tend to come from failure to compensate for temperature shifts – either internal to the unit or external to the unit.  Internal to the unit would be heat generated by the trainer inside the shell/case, whereas external to the unit would be heating or cooling in the room/garage your in.

And that brings up a key point: If you do workouts in a garage, and that garage is cold at the start (but perhaps you turn on a heater and it slowly heats the room), that’ll likely cause accurate issues for almost all trainers. There isn’t any magical solution for that unfortunately (aside from pre-heating your garage/basement/etc…).

So what about outdoor tests for power meter accuracy?

Absolutely – do them. Certainly if that’s where your issue is. The rest of this post doesn’t much care whether your indoors or outdoors. For outdoor tests I’d recommend a route that doesn’t have a ton of starts/stops in it, so it’s easier to analyze the data. A looped route is really ideal. In Paris I used to do loops around Longchamp, as it made tracking both drift and trending super easy (also, since the elevation was well known – if you wanted to get super fancy, using Virtual Elevation methods later on were much simpler on looped courses).

Within any outdoor test I recommend a range of steady-state riding as well as some nice sprints tossed in. If you suspect issues related to poor road conditions – then find those too.

But here’s the important part: When doing these tests outdoors – mentally (or via lap markers), remember what it was that you did. If you crossed railroad tracks or a section of cobbles – write that down (with the ride time) so you can match that up against data later on.

Finally, before we get to analysis, while I do occasionally see trainer accuracy issues with differences between ERG mode and simulation mode (which is what Zwift uses in non-workout modes), it’s super rare and is more systematic than individual based.  For example I saw that this past summer on the Elite Drivo 2 in an early test, which was then resolved later through an update by Zwift and Elite.  We’ll talk about some other non-accuracy type issues later on.

Capturing the data:


Next, we need to capture the data. You cannot compare two files from two different days/times/anything’s. It needs to be two or more files from the exact same workout at the same time. If you show any company (or me) two files from two different days/times, you’ll get roughly the same response as if you showed up for your airline flight on the wrong day. I feel for you, but there’s nothing I can do with that in any meaningful way.

Generally speaking for my tests I capture data on multiple head units. I tend to use smaller Garmin Edge devices because I can plug a bunch in via USB cable and quickly download the specific files I want, and then rename them appropriately. Naming is critical here to ensure you’ve got the right things sorted for later on. Here’s how I name mine:


When capturing power data, ensure you’re capturing at 1-second recording rates. By default all Garmin units will force themselves to 1-second recording rates when power data is involved.  All Wahoo data is 1-second rates. On Suunto and Polar, all data is one-second unless you’ve got yourself in a long-form battery mode.

You can also use apps like Zwift and TrainerRoad and simply download the files later.


(Also, turn off auto-pause)

Seriously, just don’t pause it. The reason you don’t is that some devices treat pauses differently in how they record the data in the file. This can make comparison incredibly difficult if the timers don’t align because there’s a gap in one file and not in the other.  I never ever stop my timers for any of my tests. Even when I stop for ice cream and hot dogs mid-way through a 5K run.

Finally, file types. Once you’ve captured the data, you’ll usually have a few options on file types. Without question the most preferred is .FIT files. These contain the most data, including sensor information. Next after that is .TCX files, and then .GPX files. The reason .FIT files are super useful for this purpose is that they contain the exact sensor ID’s (when using ANT+), so you can keep track of which sensors are which. Whereas the less detailed .TCX and .GPX file types don’t contain these.  Everyone but Polar uses .FIT, so it’s not a big issue. For Polar, use .TCX.

Analyzing The Data:

Next, I’ll walk you through some analysis basics. Note, there’s a million ways you can get bad accuracy things, but I’m going to simplify it into the most common ways possible.

For all of these I’m using the DCR Analyzer for the graphs. It’s the tool I use for all my reviews, and you can use it too here (with more details there on how it works). There are other non-DCR options too. Golden Cheetah is one, Sport Tracks, WKO4, and even Excel  The reason I use the DCR Analyzer though (aside from being mine) is simply that it’s purpose built for comparing sports data files (power/HR/GPS/cadence/etc…). It’s entire purpose in life was to drag and drop two or more files together and instantly get a result.

Don’t use averages:

First, to begin, don’t use total average power to determine whether a unit is accurate. After all, you can be wrong 100% of the time and still have same average. In fact, I’ve seen exactly that in some cases where a power meter is so wrong that sometimes it’s way high and then sometimes way low. That results in an average that’s actually pretty darn close (if not spot on), but isn’t actually correct.

Instead, focus on sections and evaluate them individually. Be it parts of a workout, or specific events in the file. I dive into some of those below.

It’s not personal:

And next comes what I know some folks will hate: I can’t tell you which unit is right if you have only two files.

Actually, you’ll hate this more: I can’t tell you which unit is right if you have only one file.

Now, there are some boundaries to this, and you have to hope you’re outside the boundaries. For example if you show me a file that shows you at 1,500w for 3 minutes – then yes, something is wrong there (though, it might actually not be accuracy related).  Inversely, if you show me a file that says you were at 60w and you were sweating bullets – then that’s probably accuracy related.

But if you show me two files that are say, 20w apart on 250w, then it’s nearly impossible for me (or any company) to tell you which one is right.  Even if they are ‘upside-down’ (I.e. a trainer reading higher than a power meter).

And that’s the piece that tends to cause most consumers the most pain. As much as folks want to believe they ‘know’ their numbers, I can’t trust that. It’s nothing personal, but that belief assumes your original assumptions were correct. For all we know, you might have had a previously incorrect trainer/power meter before, and now you’ve got a correct one.

That happened recently with someone that asked for assistance when they went from a lower-end trainer to a higher end one….and their wattages decreased. Unfortunately, that’s the ‘benefit’ of going from a +/- 10% trainer to a +/- 2% trainer.

Finally…If you’re on a left-only unit, it’s hard to trust your numbers for power accuracy comparison purposes..

Technically, it can be a right-only unit too I suppose. The point being if you’re on a single-sided power meter (like an older Stages) – then I can’t account for whether or not you’re balanced. Again, I know it sucks, but if you’re balance is 3% heavy left (totally common), that means that right there you’re looking at a 6% difference from reality (because the numbers are simply doubled to get total power). Again, it’s not person, it’s just math.

That’s not an accuracy issue:

Next, another Debbie Downer moment coming up: Things that are often lumped into accuracy issues but aren’t accuracy issues. These can introduce errors that may look like accuracy errors, but technically speaking, they aren’t.

Here’s the most common ones:

A) Dropouts: While annoying as crap, these aren’t accuracy issues per se. They’re typically connection issues. They can be introduced by wireless connectivity issues, or battery contact type issues. The best tip for troubleshooting this is to start by getting your watch/bike computer closer and see if it continues. Next, try switching from ANT+ to Bluetooth Smart or vice versa (if you can). Or try a different device (if app based). If it’s a computer for Zwift or TrainerRoad, try using a USB extension cable for the ANT+ stick. Look carefully and see if the heart rate is also dropping out at the same time as the power/cadence, that’s often an indicator on where precisely the problem is (transmission or reception).

B) Spikes: Equally as annoying, these can be power accuracy issues, but can also be unrelated. For example there’s been occasional issues over the years where a head unit will record something like a 22,000w spikes. Most head units and even training platforms will filter these out, but sometimes they slip there. This can be caused by a timing issue as well for reception of the data. It might look like an accuracy issue – but in reality it could be caused by the receiving watch/unit/app. There isn’t a good fix here except to update your firmware on all devices, and then failing that you’d need to contact the manufacturers involved.

I’m sure I’ll think of some others to fit into this not-really-an-accuracy-issue category, but for now the above two are by far the most common.

Common Accuracy Issues:

Next, the meat of how to resolve common accuracy issues. Or at least, figure out what’s occurring. Here are the most common ones:

A) Drift: This is most easily seen when two power meters slowly drift apart over time.  Check out this graph from last winter on a trainer and power meters. Notice how by the end of the workout the two line are quite a ways apart (the issue has since been fixed).  In this case, you need to figure out which of your units is drifting. Be it multiple power meters or a trainer, one of them is drifting and one of them isn’t. But which one is right?  In my case I had two other (trusted) power meters on the bike, so the trainer was the odd duck out (that doesn’t always mean it’s right though).  The next best thing you can do is finding a friend with a power meter and putting it on the trainer.  Or, if comparing multiple power meters, than perhaps borrowing a PowerTap hub from someone and seeing how that shakes out (typically speaking the PT hub is great for finding drift issues since it autozeros every time you coast).

Further, if you see drift like this, try and figure out when the point of drift stabilizes. In the below example, it seems to be by about the ~15 minute marker that the drift has kicked in, maybe even 19 minute marker.  Thus, I’d ascertain that this particular trainer (on that older firmware) required about 20 minutes of riding time to stabilize. For some trainers (or power meters), that may be the end-game answer. It is what it is (the CompuTrainer was like this). This would also be ‘acceptable’ if you were talking 10-15 minutes when taking a bike outside from a nice warm home to the frozen tundra of Canada. Generally speaking, my guidance would be that any drift should correct itself within 15 minutes of starting a ride. I don’t find drift beyond that acceptable.


B) Cadence driven errors: This is specific to power meters, and is typically due to accelerometer based issues, albeit rare. Still, because cadence is a key part in determining your wattage (it’s part of the formula), if the cadence falters, then your power will too. Typically this happens on rough road conditions – such as cobblestones or such.

For example, check out this file where the cadence drops out when I hit some cobblestones. I’ve highlighted it in yellow:


You’ll notice at those exact same points, the power also drops out on that power meters:


In that case, it’s simply that the power meter’s accelerometer based algorithms aren’t good enough. Or perhaps, there’s a defect in your unit. Either way, it’s a call to support for that.

C) Lack of stability: In general power meters aren’t super stable in terms of readings. It’s totally normal to see 1-second values fluctuate from something like 198w, 212w, 187w, 192w, 201w, 203w, 194w, etc… That’s 100% normal and just how power meters are. Most people use the 3-second, 10-second, or 30-second smoothing options to make this easier to see while riding.

But that’s not what I’m talking about here. Instead, I’m talking about such massive differences in power numbers that it just looks crazy. Jumps of 50w or so more. That’s not correct. And frankly, it means you’ve got a crappy power meter (or I suppose, a broken one, but realistically it’s just a crappy one). Or, it could also be a crappy firmware update. One of the very first CycleOps Hammer firmware updates had this problem (resolved years ago). Here’s how it looks:

D) Inaccurate sprint data (peak or post): This is more common with trainers than power meters, though some power meters do see this. First however, you need to understand that getting identical max 1-second power numbers between multiple power meters to match is virtually impossible. This is largely due to the number of moving parts we’re talking about. Multiple head units recording multiple devices all transmitting at very slightly different rates and transmission types. It’s sorta like watching athletes on a track respond to a starting gun.

Still, you want some level of reasonableness. Below is an example of an older WatTeam power meter that shorted sprints. This is also somewhat common with wheel-on trainers in particular, as well as some newer brands.  Typically the reason for this is the company is applying a bunch of smoothing to cover up the fact that their algorithms aren’t good enough to properly detect a sprint.

Also, you’ll notice I called out ‘post’ in the title of this sub-section. That’s when after a sprint the power essentially appears to drop-out incorrectly. For example when you throw down 800w, and then back-off to about 250w. Sometimes you’ll see a trainer show 0w instead of that 250w. This happens when the algorithms of the trainer don’t correctly account for the flywheel speed catching back up to your reduced speeds. Again, not super common, but can be seen on some lower end trainers. Aside from choosing a different gear (to reduce speed), there’s not much you can do about it.

E) Dual-sided issues: Next, there’s issues where a single side of a dual-sided power meter can bring down the entire house. Remember, that a dual-sided power meter is simply the sum of the two individual power meters on each side. So if one side is inaccurate than total power is inaccurate.  This is hard to spot at first, and usually manifests itself as the power being lower or high, but you’re not sure why. The best thing to look at is left/right power balance, especially if you can compare against another left/right capable unit.

For example, in my Shimano power meter test I was able to pinpoint that one side was reading incorrectly by plotting the two dual-sided power meters atop each other, and seeing how the numbers didn’t add up. In particular, by doing a simple single-leg pedaling test I was able to show that the Shimano was giving non-zero readings even when there was no load applied:

As noted the fastest way to dive into left/right issues is by starting with a single-leg pedaling test. Pedal 45-60 seconds per leg, leaving the other leg unclipped. The unclipped leg should report 0 wattages. If it doesn’t, something is amiss (note that technically there’s some force being exerted here, but any reputable power meter will correctly zero that out). There isn’t a magical fix here however, except contacting the manufacturer and either trying a new unit or a different brand.

F) Single-sided issues: Finally, we wrap-up with what I touched on earlier – but if you have a single-sided power meter it can be incredibly difficult to do power meter comparison tests. That’s because on a single-sided units (like a left-only Stages), the power is simply doubled. There’s no fancy math involved. They take the left side and double it. So if your left leg is weaker than you’re right – it’ll underreport your total power significantly (double the actual). Whereas if it’s heavier it’ll overestimate it by double. Not only that, but most people don’t have constant power balance across all wattages.  For example when I’m fatigued I have a difference balance ratio. The same when I sprint. Up to FTP wattages (about ~300w), I’m pretty balanced, then it all goes to crap above that. Same is actually true at easier wattages for me (like soft pedaling wattages). I don’t have a solution here for you on this one, it’s just a limitation of that specific power meter technology (or price point I suppose.

G) It’s in the wrong order: Have a trainer that’s reporting higher than a power meter? That means the power numbers are effectively upside-down. That’s not right due to drive-train losses.  But first, ensure that you’re talking about tolerances that are acceptable. Meaning, if the trainer is only 1-2% above the power meter, then by time you add in all the +/- accuracy ranges, you’re still within spec – so trying to convince anyone otherwise will likely be futile.  But if you are out of range, then go back to the calibration and spin-down sections of both trainers and power meters, as well as look at aspects like temperature drift.


Ok, that’s a lot, and I’m sure I’ll think of more over time to add to the list.


Some of you may be wondering what power meter or trainer I think is the most accurate. For power meters, that’s a bit trickier, since there are so many conditions and frankly, far too many brands to list here. Instead, I cover that in my power meter recommendations guide here.

For trainers, the selection is a bit less. If I were to look at the best of the best from an accuracy standpoint, it tends to be the Elite Drivo series and the Tacx Neo series. But you have to understand the gap between those and most of the other trainers that are priced from $799+ are miniscule, if present at all. For example $849 Elite’s Direto trainer is astoundingly accurate, and IMHO basically just as accurate as the Drivo series.

It totally depends on the nuances of your specific ride. For some people (99%) they may never notice the difference between those two price points, whereas for others it may become more apparent. Either way, I summarize all those in my trainer recommendations guide here.

As far as how to fix issues beyond what I’ve noted, it’s mostly going to take calling up support of the company of the product you’re having accuracy issues with. In many cases, a simple unit swap out might resolve the issue. Whereas, if you’re looking at a low-end trainer, then you’re likely at the limits of accuracy of that device.

With that – thanks for reading!