Walking up and down a flight of 17 wood stairs barefoot with the iPhone on my waist and the FitBit in my pocket.
Up and down is 34 steps, perhaps 35 on occasion when I turn to go back up.
Doing up and down 10 times therefore should be 340-350 total steps and approximately 5-10 floors of steps since the 17 steps is probably around 17 feet.
iPhone actually in my hand, FitBit in right pocket
FitBit Before 2,585 After 2,870 TOTAL 285 Steps Floors Before 6 Floors After 12 Total Floors 6
iPhone Before 1,328 After 1,627 TOTAL 299 Steps (From Withings App)
iPhone on my holster right hip, FitBit in right pocket
FitBit Before 2,870 After 3,223 TOTAL 353 Steps Floors Before 12 Floors After 21 Total Floors 9
iPhone Before 1,422 After 1,979 TOTAL 557 Steps Floors Before 5 Floors After 17 Total Floors 12 (From Apple Health App)
iPhone Before 1,627 After 1,978 TOTAL 351 Steps (From Withings App)
Conclusions: Barefoot on steps is NOT likely to be the problem with the iPhone even when it is NOT on the waist. However, the Health App clearly DOES NOT sync well with the motion sensor and so sometimes your readings can be off (see above) and NOT be the current reading. To be sure the App is current you need to cross reference it with another app reading the data from the motion sensor (in this case Withings) and when the numbers match you can be sure the Health app is actually reporting the correct data. I believe the data on floors for the iPhone and the FitBit are probably about right, although it is hard to know what each of them considers to be a floor (ie what height difference. FitBit uses 10 feet). The overcount on the iPhone is more likely from a failure to sync in a timely fashion. If we assume approx 200 extra steps, the correct iPhone floor count probably would have been 351/557 * 12 or 7.5 which is essentially the same as the FitBit. (perhaps slightly lower).
Walking flat barefoot on a stone kitchen floor. 400 Steps nonstop (this does not replicate normal walking around the house which is start and stop)
FitBit Before 3,267 After 3,671 TOTAL 404 Steps Floors Before 21 Floors After 21 Total Floors 0
iPhone Before 2,015 After 2,393 TOTAL 378 Steps Floors Before 17 Floors After 17 Total Floors 0 (From Apple Health App)
iPhone Before 2,014 After 2,392 TOTAL 378 Steps (From Withings App)
Conclusion iPhone Under Reports barefoot steps by approximately 5.5% (this is a crude estimate and the numbers are small, but it is suggestive). This is NOT sufficient to explain the much larger under-reporting observed during the day.
Walking flat barefoot on a stone kitchen floor 200 steps with 10 10 second stops
iPhone carried in hand 200 steps with stop and go every 20 steps
FitBit Before 3,696 After 3,911 TOTAL 215 Steps Floors Before 21 Floors After 21 Total Floors 0
iPhone Before 2,392 After 2,545 TOTAL 153 Steps (From Withings App)
iphone off by about 23% under count
iPhone carried in holster 210 steps with stop and go every 20 steps
FitBit Before 3,911 After 4,139 TOTAL 228 Steps Floors Before 21 Floors After 21 Total Floors 0
iPhone Before 2,557 After 2,742 TOTAL 185 Steps (From Withings App)
iPhone off by about 11% under count
My conclusions are as follows:
- iPhone motion detector is likely as good as the FitBit, however, the interpretation of this data is at the heart of the discrepancy between actual and recorded step counts.
- Apple's Health app exacerbates this problem, because it reads the data from the motion chip and DOES NOT do so real time and often will crash. This gives the appearance of a lower number, unless you are sure it is synced to the motion detector when reading. The only way to do this for sure is
- Turn on and off the phone (usually works, but it means truly turning it off where you press and hold the power button)
- Checking to see that the data matches what you see on another app that DOES NOT crash. Withings seems to work pretty well, but will NOT record the floor data
- iPhone seems to capture data BEST when placed at waist level on a holster (no testing was done in the pocket), but clearly if you are holding the thing and walking around you will be UNDER COUNTING by easily 20-30%.
- iPhone app seems to be pretty good on steps even if you are barefoot (my guess is despite what I originally thought steps actually create more motion and vibration so this is NOT the setting where you tend to get under counts).
- iPhone WILL NOT be as accurate when you are barefoot around the house (probably lose about 5-8% under count)
- iPhone WILL NOT be accurate if you stop and go (which is what we do when we are at home and in the office). This appears to DOUBLE the undercount to 10-25% would be my guess.
- In light stepping situations (ie barefoot around the house) FitBit may overcount by 5-8% in the at home situation described above (be nice if someone else did some recording and checking to see what they get). It's hard for me to be sure, but it clearly is NOT under counting and the over count could be as little as 1-5%, but I think there is some (at 1-5% it's really NOT very significant).
- Therefore the combined difference between the two devices is easily 25-35% in the setting of your home. This is made worse if you put the phone down. Therefore if 4-5000 steps are taken at home you can easily see a difference of 1-2000 steps over this period between the two devices.
- It is likely that the difference will be consistently different so that you can still use this as a point of reference for calorie tracking and progress. However, if you dramatically change how much of your steps occur at home versus while you are exercising the iPhone is most likely to be erratic and incorrect.
- So the reason some people get good comparative results and others do not has a lot to do with where they are wearing the iPhone, how much of their steps come from a real activity, versus rummaging around in an office setting or at home. So the more your steps come from home or the office the more your iPhone will UNDER COUNT.
- All of these conclusions are based on very small step numbers and a single device. To test for yourself and report what you find to this thread simply take a screenshot BEFORE and AFTER your test and then subtract the data and report.