iPhone 6 Plus M7 Chip Step Data Accuracy OFF compared to FITBIT One

Karrell,

Thanks I was beginning to think it was just me. The step tracking in the apple health app is somewhat awkward in that the display doesn't always change real time, seems to crash a lot (perhaps that's why it under estimates) when other apps are communicating with it (like when I'm testing multiple running apps), doesn't have a good historical list display (ie doesn't have one at all and looking at the graph is pointless), and can be confusing when there are multiple sources of data (the algorithm for choosing and resolving data conflicts for the same time period is pretty much a black box). If you check to get source data from multiple apps the Apple Health app seems to add all of them together giving you a massive amount of steps that is then later resolved in some unknown way.

I am curious to know what your tests are on accuracy of the fitbit for steps. I've found it to be so accurate when I walk or run around a track that its better than me trying to count the steps (because it's so easy to mess up). Inside the house going up and down steps and walking around the office I'm not as sure. However, outside in the open with a run I think the apple health M7 M8 chips are reasonably close. My impression is that the problem is primarily when you are in the house, shopping, or at the office (ie not regular same movements) that it is off. Sadly I think for most of us 30-50% of our steps are this way.

As far as I know FitBit refused to link up with Apple's Health Application, so the only way I know of to get some of the data across is to find an application that communicates with FitBit and which also communicates with Apple Health's application. I don't have any ready answers to that, but I know FitBit communicates with Withings and MyFitnessPal. Withings does communicate with Apple Health, but I don't think the step data crosses (oh well). If you had a good way of transferring data I'm all ears. BTW Fitbit has a VERY GOOD return policy and if your device is NOT working they will replace it without much fanfare.

Meg, Karell,

Thank you for your thoughts and observations. I have a medical background. I am not a programmer, but maybe I'm a closet data geek. I agree with Meg's point that the goal is to workout more. However, I also agree with Karell that if the devices are not at least precise (meaning they give the same or approximately the same number when you do the same activity) then you are kind of screwed. Being different is okay so long as it's consistent and so long as one understands how it's different. From a psychological perspective having the number be different on the high side gives a better emotional boost and encouragement, however, in the end being able to effectively use the data with calories in and out to figure out how to adjust your activity and diet to achieve your goals is what I think I most want to get out of this.

I like the idea of the wrist manager program, does it actually require that you have some sort of wrist device or is that just the name?

Being that I am in the medical field I did test two different fitbits, one jawbone and the Nike Fuel. I really didn't like the jawbone interface, but I did like that it could charge with a regular USB cable. The Nike Fuel was very disappointing to me and since the two fit bits weren't used at the same time I can't confirm whether the numbers were the same, but that subsequent products appeared to get similar results for similar activities and so gave the impression that they were equally accurate. I suspect there is always some slight variation between products. I had the impression the jawbone was not as accurate and I won't even comment on the Nike Fuel (yes it was that bad)

Oddly enough the same apps getting the same information from the M8 chip are not always generating exactly the same number, but pretty darn close (off by less than 1%). I was going to guess originally that the M7 and M8 chips send the raw data to an app that can then use an algorithm to adjust and refine the actual steps. I now think that it actually just uses its own algorithm and reports the outcome to the other programs (any thoughts on this?). I'm going to bet that the quality of the motion chips between M7 and M8 is similar with the M8 being somewhat better, but that the real issue is the quality of the algorithm used to establish and validate steps. Also there is the possibility that memory is shared across apps and that it goes out from time to time if something else is active (just a guess be nice to test, I surmised this from getting the health app to crash while it was reading and sending data to 3 other apps). I wear my iphone 6 plus on a holster at the hip and my fitbit is usually in my pocket or it's on my headset clipped to the front of my shirt. Today I observed that on the Elliptical (yes I know they aren't supposed to work well there) that I actually got a higher reading on the iPhone 6 plus for the very first time. This, however, might not be useful as I observed that despite the Fitbit undercounting on the Elliptical the actual calorie estimate was darn close to what I get using a heart monitor with an app called Wahoo Fitness. Which way of calorie counting is the most accurate? Who knows, but I suspect that your heart rate is a pretty darn good measure.

There's another interesting thing I noticed on the Health App and that is that it is possible to add entries so in theory I could adjust the numbers on that. I haven't wanted to do that until they improve the data input so that it's more useful and fun to work with. I've kept historical data back for about 6 months and perhaps years for other apps. I'm not sure I'm actually finding it that useful to keep. The only data that is fun for me to keep is the weight data which has nothing to do with Fitbit, but Withings collects it automatically every morning I step on the darn thing.

My best guess as to why the iPhone under counts is that the software parameters being used are not as well fine tuned and established as those in the fit bit and that it has little to do with the chip itself. Fitbit has a higher vested interest in what they are doing and spent more serious dollars and time on this. That is my guess. Since the iPhone has GPS and tracks motion in several directions there really is no reason why it couldn't be a lot more accurate. If they open up the data and allow other programs to tinker with algorithms you might actually get data a lot closer to FitBit. In fact, Fitbit's app could then actually use the original source data in their algorithm's to accurately calculate the steps. They don't have a lot of incentive to do this as it makes using their product unnecessary. Even so I do like that they now allow you to accept entry of data into FitBit from the iPhone although as we all now know that data is undercounted.

One last thing I did notice that instead of deleting data I think you can just de-select a source that you don't want to use in the tabulation on the Health App. Not sure that works but I had the impression that this eliminated some of the double counting that I perceived. Be interested in your thoughts.

Karrell, The link works! Thanks very much that is extremely useful. Unless the fitbit program actually uses the data from the M8 chip differently, I would be surprised if setting your FitBit to use that as the source of its data improves its accuracy. With that said it should be easy enough to compare the FitBit step number from the FitBit app using the iPhone's sensor with the native number from the iPhone M7 chip (you might have to find an app that pulls this data from the Health App, because otherwise the Apple Health App might be getting inputs from multiple sources and not tell you what it pulled from the chip itself). If you do this and get a DIFFERENT number please let me know, I am curious. I also have a friend who builds health monitors and chips and may ask him if he knows how the data is reported. I am betting that the M7 and M8 provides raw data, but that what is being provided to apps is software processed using an Apple derived algorithm that may not (being generous here) be as good as FitBit's algorithm.

Meg, Completely agree Calorie counting is very crude at best and I'm not sure which algorithm is best. I think the heart rate is a better method. What is most useful is to get some consistency in the way its is measured so you can determine what works for you. So in other words if you are negative 1000 calories and still not losing weight you need to go more negative or at least begin thinking that perhaps there is something else that needs to be done. Most people forget that unless you are exercising a **** of a lot (ie more than a few hours a day) MOST of your calories burned are NOT from your exercise but from your basal rate. Therefore you are actually much more effective in losing weight by calorie restriction (very hard to do) and doing things to change the basal rate (ie strength building exercises (yes I know that sounds counterintuitive, but muscle burns weigh more basal calories than fat).

I would be very interested in knowing how any of you reconcile your observations with the article below which suggests that the 5s with the M7 actually tracked very close to a FitBit and UP.

http://www.tuaw.com/2013/12/19/how-the-iphone-5s-measures-up-as-a-fitness-tracke r/

My only explanation is the observation (not with sufficient data points) that the following may be happening:

1. iPhone tracking software (emphasis on the software) has been optimized for getting steps from regular repetitive walking and/or running
2. iPhone tracking software works especially well when the iPhone is on an Arm Band and reasonably well when placed on a holster (which is what I do on my waist)
3. iPhone software probably does a much poorer job when the thing is in your pocket (ie sensitivity for defining a step is too low and misses steps)
4. iPhone software probably is not well designed to count steps when they are of different weight, different torque (ie turning, changing directions, odd positions around the house)
5. To the extent that your steps are inside the house or in the office the software is not properly calibrated and will undercount (this would mean that the more your steps are from a run the closer the devices will be, it also means if you put the thing on an arm band you might get better results)

These are just guesses. The other possibility is that one or all of us have a bad M7 or M8 sensor and should get the thing replaced (don't know how to determine this for sure).

Based on this article the FitBit uses a three dimensional accelerometer. Do you know whether the M7 or M8 is three dimensional this would explain why the FitBit might be better around the house/office etc.

Karell,

Yes I know that your tests like mine consistently show a lower read for iPhone than the FitBit.

The article I showed above http://www.tuaw.com/2013/12/19/how-the-iphone-5s-measures-up-as-a-fitness-tracke r/ seems to support the idea that the iPhone should provide data similar to that of the FitBit, however, my read of it is that this is true when they are wearing the iPhone on an armband (probably the most ideal location to measure movement given the torque of your arm) and when doing a consistent and regular activity like running, treadmill, walking, etc. I can imagine that any start and stop might result in a re-calibration calculation that loses a step or two before it realizes that this is a real step. Not a big deal for a 3 mile run, but a huge deal when you are at home. At the bottom of the article he describes a 13,800 step day for which the iPhone only registered about 8,800 steps (sound familiar?). He claims that this is because at home he would sometimes forget to keep the iPhone with him. Now the step difference described which is about 5,000 steps easily represents 3 miles. That is a HUGE difference. He also notes that he doesn't use it on his arm at that point. I think it more likely that 1-2,000 of those steps are his failure to keep the device on him and the remainder is that the iPhone simply is not as sensitive when used in a different location on the body and for constantly changing and varying activities with different movements and rotations. In other words, I am going to bet there is nothing wrong with our iPhone motion chips. Sadly I think it is likely the Apple Genius people won't know much either (I asked them the other day and got this "Huh, oh it measures steps by itself..." I think they aren't up to date on this yet so unless you are lucky and get someone who really took the time to look this up you likely won't get a useful answer. You might do better simply calling Apple that way they can at least try to refer it to someone more knowledgeable.

Based on the link below the FitBit uses a three dimensional accelerometer. Do you know whether the M7 or M8 is three dimensional this could explain why the FitBit might be better around the house/office etc?

http://en.wikipedia.org/wiki/Fitbit

I'd hate to think that the M7 M8 are simply not as good and doubt that this is the case, but perhaps if they aren't three dimensional they might not be as accurate. Who knows?

When I am comparing the Fitbit to the iPhone the iPhone is almost always on a holster by my waste and occasionally in my pocket, the Fitbit is always on my headset which is around my neck or in my pocket. Generally I find the FitBit is most accurate when in a pocket, on my shirt neck, but NOT hanging around my neck so if anything I would think the iPhone positioning should be in its favor.

Karell if you are able to wear the thing on an arm band all day and see if the measurement is closer to the FitBit that would be helpful to know. Also I would be interested in knowing if you find as I do that if I go for a walk and check both the FitBit and the iPhone for steps, that they are a LOT closer and sometimes within 10-20 steps.

Thanks