Body Fat percentage accuracy with Garmin S2 smart scales

So I wanted to present an observation about the accuracy of body fat percentage on these scales.

When I first got the S2 it said that my  BF% was 26.2. But hey I am at 18.7 on my old scales  (both sat side by side). So I spent a few hours trying to work out why the difference.

Of course like everything there is an algorithm to calculate BF% based on your age, weight , height and activity level (I believe), which you select at set up. So I set my activity level at 5. But I didn’t look at the help section which suggests what level to set it at, based on activity. But even then what is my actual activity level according to these suggestions. If  I’m disc golfing twice a week for 2 hours each and go for a walk another day for an hour, does that make me an activity level 7?

No I think not. But it certainly brings down my BF% when I weigh. So I sacked the activity level setting and went for the body composition setting instead. I bought a set of callipers and measured my skin folds using the Jackson:Pollock 7 method. After trial & error with this (watching YT vids etc), I measured at 20.42% which I entered into the Garmin settings.

And guess what, the S2 scales are more in line with my old scales. Hope this helps.

  • Garmin appears to have programmers doing all medically related calculations.  They're generally a mess.
    I suspect many of these metrics are simply correlations with people our sex, age, and evidently reported activity level.
    It certainly isn't anything the scale is measuring.

  • Thank you for sharing. I had an almost identical experience. I reasoned that the algorithm was updated and likely also the hardware too. I do notice that the S2 has much smaller variation in all its measurements. Also just looking at myself: I'm much more confident in the 21% number the S2 gives me than the 17% the original Index scale gave me. I'm just not that fit or skinny. 

    That being said, as with the other commenter, I am cautious with these readings. The correlation between overall weight, body fat and *muscle mass* (and also water, but with the opposite sign) over time is so tight despite obvious changes to my fitness and strength that I suspect the predicted values based on gender, age, height and activity levels dominate any data coming through the bioelectrical impedance hardware.

    I have yet to do a lab test for body fat, but I suspect inputting that data would allow the device to use a superior algorithm. However, I'd rather do a lab test than an at-home caliper test as you did. For data models, garbage in, garbage out.

    Also, isn't it strange that Garmin can't calculate your activity class from recorded activities and other data? A Garmin is on my wrist at almost all times... 

  • You have made a great point about the Garmin watch, surely the heart rate, gps position etc and movement from the watch. This should give the connect App all it needs to assess activity level. That said not everyone uses a Garmin watch. 

  • For a company having so much experience with sports, having bought a sports data analysis company, selling so much sports measuring equipment, they seem to be really bad at combining and utilizing all that data. Too bad. There's so much potential there.