I’ve been blogging and sharing various graphs of my diet and fitness data for several months in my Fitness and Food series on the JMP blog. There, I shared how I collected years of activity and food log data using BodyMedia FIT activity monitor armbands. You can also hear more about my observations about this project on the inaugural episode of QS Radio here. My first armband was a Core that synced via USB and the second was a Bluetooth-enabled Core BW. BodyMedia’s app and web site provided simple visualizations of weight, steps and calorie information that helped keep me motivated to track my data on a day to day basis for several years before I ever explored my options for exporting my data. Eventually, I wrote JMP scripting language (JSL) add-ins to make is easier to import my activity and food log data history into JMP, available on the File Exchange on the JMP User Community.
My visualizations of my BodyMedia data were also the topic of my poster from JMP Discovery 2014, and you can see a copy on the JMP User Community or view a copy here. While examining my activity data in JMP, I quickly saw that the biggest downside to the armband was my reluctance to wear it during the summer months. I didn’t like wearing how obviously it showed on my upper arm in short sleeve weather. Also, if I wore it outside during exercise in the spring and summer months, I ended up with a very obvious tan line on my upper left arm!
Of course, I was well aware of my choice to take off the armband during the day in the summer months, but I was really dismayed to see this decision had such an obvious impact on my data collection efforts. Clearly I could not trust measurements of my activity and number of steps during warm weather unless I had worn the band the whole day. BodyMedia’s software auto-generated an average calorie burn for the time I was not wearing the band, and I could fine-tune the estimate by copying an equivalent time segment from a day in the past week with similar activity patterns. But I found I rarely took this extra step. Even if I did, the software didn’t estimate steps for periods where I was not wearing the band. Still, I kept using the armband because I liked its sleep-tracking features and I had invested many hours entering custom recipes and food items into the food logging software.
I’ve been reading reviews of newer activity monitors with great interest recently, waiting for one with a strong enough feature set to overcome the inertia of switching from my familiar device and software. My requirements were relatively straightforward:
- I needed a simple food logging interface including an iPhone app.
- I was not willing to sacrifice functionality I already had. I wanted to continue tracking my sleep, and my BodyMedia FIT auto-detected my sleep/wake status. I ruled out several promising monitors which lacked automatic sleep tracking features.
- I wanted to be able to export my data at minimal (or preferably no) cost, aggregated at the day level or preferably hourly
Choosing a New Monitor
After considering various different devices on the market and a few that were coming, I decided to purchase a Fitbit Charge HR. I bought it on Saturday, March 28th at a local Target. I purchased the black band, which was the only color they had. Black is a more practical color, but that didn’t stop me from checking around on my phone to see if the plum color was available anywhere (unfortunately, it wasn’t). I bought the device replacement plan from Target for $17, which hopefully will make it easy to swap out a damaged or malfunctioning device should I run into any issues.
I bought a size small band, and initially I found the top of the display to be slightly uncomfortable, as it was a bit wider than my wrist at the top and didn’t fit around it as closely as I expected, but as the band became broken in, I quickly became used to wearing it and don’t find it uncomfortable. I have to take it off to charge it, as the charging cord fits to the underside of the band, but it doesn’t take long to charge and I only need to do it every 4-5 days. Using the heart rate function appears to drain the battery more quickly than regular use for step and activity monitoring.
I considered waiting for an Apple watch, but opted not to for two main reasons. First, I was primarily looking for an activity monitor and not a smart watch. I almost always have my phone with me on the go and even at the gym. Second, reviews hinted that the Apple watch would have a relatively short battery life and might require overnight charging through a plug on the underside of the watch. Sleep tracking was not projected to be part of the initial feature set. I love my iPhone, but its need for frequent charging is my single biggest frustration. I have a charging cords at home, at work, and in my car to make sure I don’t ever run too low to use it.
I also considered waiting for a Jawbone Up3, built with the technology Jawbone acquired when they bought BodyMedia. I love the observations and visualizations made by Jawbone data scientists on their blog and I really like the looks of the Up3, especially future concept drawings of additional colors expected after the initial launch of black and grey. But the Up3 was not available at the end of March, and in fact many customers were waiting on bands ordered months prior because of manufacturing delays. I knew from reading reviews of the device that I would have to manually initiate sleep tracking, which was a nonstarter for me.
I’m not alone in selecting this device-in fact, a few weeks after I purchased it, a prominent review of the Fitbit Charge HR in Forbes declared it “The Best Fitness Tracker You Can Buy.”
Side by Side Comparison
After purchasing the Charge HR, I wore both devices for 2 ½ days, during which I checked the BodyMedia app against the Fitbit readout frequently. The calorie burn estimates seemed extremely similar at any given moment. I found this interesting because I remembered comparing armband estimates to those from my daughter’s Fitbit Zip (which clips to a belt) a few summers ago. I had the impression that the Zip overestimated my calorie burn and number of steps, even though I had adjusted her account settings to reflect my own height and weight. Perhaps this was due to the repeated laundering the Zip in question had endured at the hands of a forgetful 10 year old, or perhaps the Fitbit algorithms are improved with the extra information available from devices with direct skin contact.
Here is a graph of hourly step count (exported using the Zenobase service) from the period where I wore both devices. You can see on Friday, I wore my BodyMedia armband to work, and on Saturday afternoon, I purchased and started wearing the Charge HR alongside the armband. I tracked with both devices for a full weekend day (Sunday) and work day (Monday), then with the nice weather on Tuesday, switched over to wearing the Charge HR alone.
Although the exported step counts were largely similar for some hours of the day, it seemed that the Fitbit was quicker to start tracking steps for intermittent or non-exercise activities, so it counted more steps than BodyMedia during those activities. There seemed to be a bit of a delay between when I started moving and the armband began to track my steps, so when I was not moving consistently, it counted fewer steps.
While running the monitors side by side, I also tried out the Fitbit food logging software and the MyFitnessPal (MFP) logging tools alongside the BodyMedia food logging app I’d been using. I decided on using MFP going forward because of its extensive food database, although I remain wary about checking user contributed entries. Many of the items I had to add as custom food items to the BodyMedia database were already in the MFP database, which cut down substantially on the need to add custom items. Additionally, I had already written an add-in to import CSV files exported from MFP using the free My Fitness Pal Data Downloader Chrome extension. The social aspect of MFP food logging intrigues me. Although I don’t currently share my log publicly, I know others who do, so that remains a possibility.
I had intended to wear the FIT and Charge side by side for longer, but between the improving temperatures outside and the consistent readings, I switched over to using the Charge HR alone within a matter of days. One major contributor was being able to use the Charge HR to track heart rate, which is not supported by any BodyMedia device. I’ve used heart rate monitors with chest straps in the past and jotted down heart rate ranges in my notebook. Though using the chest strap was inconvenient, I never did enough heart rate training to make a bigger investment in an automated tracker. The Charge HR auto-detects my resting heart rate, and when I remember to start up the timed workout function by holding down the side button for 3 seconds, I can monitor my heart rate during workouts and afterwards see a graph of the changes. I can label the type of the tracked workout (e.g., weights, walk or bike, among others). I look forward to being able to retrieve this kind of heart rate data from the device in the future!
The only downside to the Fitbit app that I have noted so far is that if I go to sleep and my sleep is interrupted for more than a few minutes before midnight, the sleep total for that period will count in the previous night’s total instead of contributing to that night’s sleep total. For example, if I go to bed at 9:30 pm on a Monday and then get up to tend to my son before midnight, the sleep hours from 9:30 till when I get up will count towards Monday’s sleep total, not towards the total for Tuesday. I never had this issue with BodyMedia’s sleep export, but now I know I will have to go back and find some way of recategorizing sleep stretches to be sure they count towards totals on the correct night. I hope there will be a way to export summary information displayed in the app, which notes information like “2x awake, 16x restless” about stretches of sleep.
All in all, although it was difficult to imagine switching devices after all the time I spent accumulating data with one device, the switch itself was remarkably simple once I decided to do it. I’m very happy I did, and am looking forward to better data collection this summer.