Data will save us all

2015-07-12 00:09:12 -0400

I've done various "light quantification" experiments over the years:

• using a Makey Makey to monitor whether my standing desk is up or down
• combining that with xscreensaver status to track whether I was actually present or not)
• processing the data extracts available from google (since they're stalking me anyway, they conveniently let me look at my location history in KML form)
• using one of the manual caffeine trackers for android for a month or so (no real data value, but useful for becoming more mindful of intake, or lack thereof.)

On top of that, I'll occasionally have bursts of enthusiasm for some phone-based manual activity or mood tracker app, or even graph paper and colored pens (if I'm being really obsessive, those get detailed enough to be reasonable layout descriptions of unachievable UIs) but those approaches rarely last more than 2 weeks - usually just long enough to "see" the habit I'm trying to break, but not long enough for the data gathering itself to become a habit.

About a year ago I tried to start another round of explicit lifestyle/activity tracking. The motivation this time was a bit more sophisticated: I'm trying to find a good compromise between

• being morning-shifted enough to get out and do awesome photography - it was, after all, spring in New England, time spent indoors is time wasted
• being night-shifted enough to do awesome hardware and software hacking - I've got this great backlog of Kickstarter output, especially in the micro-arduino space, and I had some project in mind for every piece of it when I backed it...

and that means I need to establish more regular patterns, which means being more aware of the patterns I'm in now (which for me has always required some kind of external reference, even something as simple as a diary.)

I used the FitBit One for sleep tracking when I first got it, but that meant explicitly hitting buttons when going to sleep and again when waking up, plus it meant wearing it in an ill-fitting wrist-cuff (one of my many natural advantages in my chosen field has been 20cm wrists, so I grew up aware that Carpal Tunnel Syndrome was a problem... for mechanics and carpenters; I didn't see it in typists until college.) Combine that with the data itself being relatively low quality and I fell out of the habit quickly (whereas I use the step counting to this day.)

I'm giving that another try with the FitBit Charge HR - narrow, light, comfortable band, and it automatically establishes "sleep recognition" based on heart rate. Now the only thing I have to remember is to take it off (and charge it a little) when I shower, and I get decent sleep (rather than "bed") start and stop timing (and it isn't fooled by my terrible snooze-alarm habits) along with sleep "quality" based on heart-rate which seems like it would have much more potential to catch sleep disturbances than the motion of one wrist.

The next step is to pry that data out in form that I can correlate with other events - microphones for environmental noise, temperature near where I'm sleeping (or more directly, "are the blowers for the central air system running") - and not just when I'm sleeping, I'd really like to look for heart rate patterns in my driving especially correlated with location, and possibly acceleration...