This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

S2 Body Fat not accurate

  • Consent to be added to an internal case - yes, I consent
  • Consent to access your Garmin Connect account (if needed) -  yes, I consent
  • What device/test are you comparing your Index readings to - Tanita Inner Ccan
  • What does your Index S2 display vs what does your reference device display -  S2 Fat = 17.1 Tanita =10.6
  • What is your activity class set to, and is this appropriate for your activity level - 8 - Yes
  • - Please update the body fat% calculation already, please give us an edit button for body metrics as well. This has been on-going for way too long.

    If you want to make things simple if you aren't having much luck - add an 11 activity function and a factor for us to fill in. My body fat my 5% less than my reading on your scale. If I mark 11 and 5 - then it will just subtract 5 from the current algorithm. 

  • Hi,


    I've had for a few weeks my Forerunner 255S music and am fully satisfied,great product. I also have a edge touring for years.
    I love my watch so much, and I would really want a true Garmin scale for Garmin connect. But, for 150 bucks, I expect top-grade quality. I would gladly have bought S2, but, for what have I have read there, it's not possible.

    Please let me now when the problem is (fully) fixed. Otherwise I'll wait for a decent Index S3. It's a shame, for such a great brand like Garmin, that values quality sport electronics overall, to have such a bad device in its catalogue.

  • There wont be an S3 unless they fix this as its not a hardware issue but an algorithm issue. They should do like they did with FirstBeat and just license the Tanita algo. They should also do as I, and many others, have suggested, and let people either plug in a correction factor or edit the results. I agree though - this has been going on for so long and hard to understand from a company that values algorithmic health measures and is probably best in class. I dont get it. 

  • From the outside I'd say it's an example of what not to do with machine learning : putting any variables you can find and letting the algorithm use anything to guess the target.

    The problem is that as a user you don't want to be stuck to a score, even if 99,99 of your age class have this score and therefore it maximizes the accuracy in Tensorflow.

    They'd better keep it simple, I think. You merely want your weight to come from the weight sensor and the % of fat to come from the other sensor, as on any cheap device. Not from your age, skin color, haircut, shampoo brand or whatever.

    If I'm right I hope they won't do the same error with the patch : thinking too much about minimizing the metric when the point is about the application as a whole

  • Well, it's not only the body fat (which depends on hydratation more than I'd like) - in my case the body composition seems to be completely wrong.

    If I take total wieght and subtract fat (from fat%), muscle mass and bone mass I'm getting remainder of almost 36kg. 
    Seems I have heavy brain, which is supported by extra strong spine as my bone mass shows 7,6kg.

    If I compare the results with Tanita's InBody - the fat% difference actually is not so bad (remember it is affected by legs only vs hands&legs measurement) but the body composition is useless in absolute values, hoping that at least the trend will yield some feedback.

  • Scales completely inaccurate and are used in Garmin 6s for Vo2 max and other calcs. Outside of weight, the composition info is useless. 

  • I think if we all @ mention  everyday he might muster up the courage to reply to everyone with an update. 

  • to be fair, i don't think it is 's fault. we need engineers within garmin to come up with a fix and if they can't do that in reasonable time (which is already the case) then it is up to garmin management to front up with compensation. don't shoot the messenger

  • Garmin-Kevin can't give you an update without Garmin's approval.