Determining Power Meter Inaccuracies

I just put a new Gen 3 Stages on my bike and have been doing some short power testing and have noticed on a couple of occasions that my cadence and power will be suspiciously high.

I calibrated both times, and always do so before every ride.

Battery is new and so is the latest firmware.

How does a person evaluate such data and decide whether or not this is accurate or inaccurate?

2175w here:

1850w here:

Devil’s advocate: do the absolute values matter, or are you looking for consistency? If the values are repeatable, then you can still base your training on those. You may need to reset your baseline, but assuming it’s not miles out, there shouldn’t be a huge issue.

1 Like

short answer, you put on two more power meters & compare readings.


I get what you are going for but if 99% of the ride is good don’t sweat the small variances. Easier said than done but after using my 4iiii left crank for almost 2 years now I only glance at it while riding and look at NP after the fact.

1 Like

My buddy uses a Pioneer L arm and always gets a 238rpm cadence spike at some point during the ride. I think little hickups like this are kind of par for the course. My Assioma don’t do it, not trying to hype them up but maybe there’s some general issues crank arm mounted strain gauges face?

I don’t really care, but it would be satisfying if these were true. They just stood out because of the high number so I thought I’d ask. I’m also new to training with power on the bike.

I’m far more concerned with consistency of the power meter to itself.

They look like glitch numbers IMO only because the 1-3 second power on both is a totally flat line which would basically never happen

Also very tough power numbers to hit in general, we have similar FTP and I know I’m a bad sprinter but if those are accurate I outright SUCK. That’s a Justin Williams looking sprint, if accurate you should be patting yourself on the back with both hands

This might be helpful; A shift…

From the latest DC Rainmaker post about SRAMs new group set.

But one of the things I’m ever-so-slowly warming to over time and doing years of power meter accuracy testing, is that I’m less convinced that on the whole power meters are as consistently accurate as people think and hope they are . And that for most people, even with slight shifting leg imbalances, I don’t think it’s as critical an issue. …

I’m coming to the conclusion that there are so many ways for power data to be inaccurate on any given day from any given vendor, that the left-only doubling thing probably is a wash in the end.

A few people took note in the comments, and he added further;

In general, people put far too much trust in their power meters providing day to day accuracy . I’ve got great examples of super reputable units that just disagree on a random day for no reason: Quarq DZero vs Tacx NEO 2 or 2T spiking 3-5% different from any other day. Or a Favero Assioma pedal doing the same. Or a KICKR V5 doing the same. … **the industry is yet to the point where one can assume that a PR day of 5w higher or even 10w higher is a legit PR.

I’m not saying single-sided is the answer for everyone, but I am saying that the vast majority of people who say single-sided is useless, would probably be surprised to find out how quietly inaccurate their power meters and trainers are day to day over months/years.

Grain of salt and all that…good luck with your training!


Ugh. The short answers are:

  1. Ray Maker is right, most riders don’t know how inaccurate their power meters are.

  2. There are ways to determine accuracy of a power meter, but most riders don’t know how to do them, and are convinced it doesn’t matter.

  3. Training (especially training of FTP) is one of the least demanding things you can do with a power meter, which is why people were able to train pretty effectively before the invention of on-bike power meters or heart rate monitors or even on-bike speedometers. Trainerroad started with the option to use a dumb trainer with a wheel speed sensor – you can still train effectively that way.

  4. People who claim that consistency is all you need from a power meter haven’t yet come across a situation where accuracy matters, mostly because they’re doing things with their power meter that aren’t very demanding. The real advantage of a power meter over a HRM isn’t that it lets you train better (see point 3) but to do things you can’t do with only a HRM. But most people don’t need to do those things very often, and many don’t know how to do them even if they occasionally would want to do them.

  5. Determining accuracy doesn’t require another power meter if that other power meter has not been checked for accuracy.

  6. To check the accuracy of your power meter, you need to do things related to point 4.

  7. I could go on, but these are the short answers.

1 Like

It’s probably an RPM glitch as others have pointed out. The PM uses an accelerometer to “guess” RPM, rather than use a really accurate device like a BB mounted rotary encoder.

Calibration on a PM is kind of a BS misnomer. It’s basically the same as hitting the “tare” button on a digital scale. Zeroes out the power but doesn’t necessarily affect accuracy at high wattages.

1 Like

Conducting Experiments workout:

  • zoom into 20-sec window from 47:45 to 48:05
  • cadence starts at 80rpm and ends at 106rpm
  • at 47:53 cadence jumps from 78 to 160rpm
  • that cadence jump is impossible
  • both power and cadence are faulty during a portion of this 20-second window

Cruiser workout

  • same impossible cadence issue with sprints at 1:00:25 and 1:03:14
  • same conclusion on bad power data on those sprints

I have a left-only gen2 Ultegra 6800 and dual sided gen3 Ultegra 8000. Can count on one hand the number of workouts per year with spikes, and none that bad.

Just curious - dual sided or single side only? Which crank?

Using another analysis tool you could dig a little deeper on dual sided data, however the conclusion won’t change. I’d share that with Stages support and ask for a replacement.


They look like power spikes to me which is quite common with the Stages PM.

1 Like

Thanks! FWIW, I’m on a single-sided Stages (SRAM GXP), and I love it so far. It’s just a tool. It’s been consistent with my KICKR and on a day-to-day basis. For that it’s perfect.

That said, when I drop a few 1500+ watt sprints I have to raise an eyebrow and ask myself…“Really…?”

1 Like

Noted. FWIW, I’m only seeing major discrepancies during really hard max efforts. There hasn’t been observable power/cadence spikes or weirdness during other sub-maximal efforts.

Same here. FWIW I do a lot of max effort short sprints, for example a 6 week block starting in August where I did sprint intervals twice a week. I’ve got a fair bit of data to back up my earlier statement. Your data would prompt me to get on the phone and politely but firmly ask for a replacement.

1 Like

As you wish, but discrepancies like this aren’t usually isolated. They’re a clue that other, less obvious, things are going on. Even if you think the error is occurring under a consistent high power situation, this is an example that consistency in error condition sometimes isn’t enough.

Here is one that I considered a spike on my gen3 dual-sided Ultegra 8000:

Time Cadence Power
3:26:12 81 624
3:26:13 86 1516
3:26:14 89 1037
3:26:15 90 961
3:26:16 94 806

That was the easiest one to find. My point is that even when I get a power spike, the cadence is normal. I can only think of one exception, when I ignored low battery warning for two days and on the third day the battery died. Immediately before that cadence went to ~200rpm and power went to ~3000W. And then it died. However the battery died on other rides without strange cadece/power, so that was exactly 1 time over 4 years of having gen2 and gen3 Stages PMs.

1 Like

Agreed, something is off. In contact with Stages—hopefully they will take care of it.

DC Rainmaker has of course seen a lot more PMs than I have, but based on RPE and HR and doing structured training and the resulting times from efforts, his statements sound exaggerated to me, at least over 5w per day. That’s coming from a Quarq and now Favero Duo.