Weighing up body fat scales (part 2)
Following on from my DEXA scan versus body fat scales experiment, I wanted to see how the body fat scales performed over a (short) period of time, to see how reliable their measurements were. I figured measuring them once against a known 'truth' was useful, but what if the next day they gave me entirely different readings, so the initial measurement was a complete fluke and no actual indication of accuracy. My grand plan... weigh myself every morning for 9 days straight on each set of scales, note the measurements and make pretty graphs.
Side note: if you ever decide to do something similar, do not underestimate how long it takes to weigh yourself five times every morning for a week and a half and record and collate all of that data!
The scales
For reference, in case you haven't read part 1 of this experiment (which, of course, I highly recommend!), the scales I bought to compare were:
FitBit Aria
Kamtron cs20m
Renpho es-cs20m
Tanita bc-602
Withings body+
You may note in the results that there is a measurement missing for the FitBit Aria (day 8); I swapped out my home router and basically forgot I need to re-sync the FitBit Aria to the network, so that day's data wasn't saved. Entirely my fault and no indication whatsoever of the reliability of the scales.
The results
For each of these graphs, I took one measurement (body weight, body fat percentage, etc.) and calculated the average of all of the scales on each day (since I couldn't do a DEXA scan every day to determine 'truth', I used the average of all of the scales). I plotted this average on the graph and then plotted the measurements for each set of scales on each day. Based on this, I could then calculate the difference between the average and that scale's measurement on each day and determine the range of these differences. I deemed the most accurate scales to be those where the range of differences between the average value and the scale's measurements was the lowest.
Body weight (kg)
In order, from the lowest to highest difference from the average:
Tanita bc-602: 0.114 (range: -0.08 to -0.2)
Renpho es-cs20m: 0.159 (range: -0.36 to +0.12)
Kamtron cs20m: 0.209 (range: -0.126 to +0.08)
FitBit Aria: 0.375 (range: +0.08 to +0.457)
Withings body+: 0.443 (range: -0.24 to +0.2)
Body fat percentage
In order, from the lowest to highest difference from the average:
FitBit Aria: 0.4% (range: +1.1 to 1.5%)
Renpho es-cs20m: 1.49% (range: +2.3 to 3.8%)
Kamtron cs20m: -1.49% (range: -1.6% to -3.1%)
Tanita bc-602: -1.5% (range: -1.28% to -2.8%)
Withings body+: 1.68% (range: -0.84% to +0.84%)
Muscle mass (kg)
In order, from the lowest to highest difference from the average:
Tanita bc-602: 0.96 (range: -0.5 to +0.46)
Kamtron cs20m: 0.895 (range: 0.5 to 1.398)
Renpho es-cs20m: -0.898 (range: -1.59 to -2.49)
Withings body+: 0.96 (range: -0.5 to +0.458)
Bone mass (kg)
In order, from the lowest to highest difference from the average:
Withings body+: -0.066 (range: -0.18 to -0.25)
Kamtron cs20m: 0.075 (range: 0.25 to 0.32)
Renpho es-cs20m: 0.08 (range: 0.05 to 0.128)
Tanita bc-602: -0.11 (range: -0.1 to -0.21)
Conclusions
All of the scales fluctuated with body weight. Even bone mass measurements changed daily, so they all clearly take body weight into account to calculate bone mass and muscle mass, rather than actually measuring it. I guess this is expected of home-grade scales. I did try and get hold of some medical grade scales to add to my comparison, but they were out of stock at the time.
None of the scales came out on top in every category, and which one may be best for you may well depend on which measurements you're most interested in. Personally, I think I'd probably opt for the Renpho scales out of this selection as they were the most consistent if you take into account both body weight and body fat measurements.