Monitoring, not just for your IT infrastructure

No, I'm talking about that fancy wristband some of you wear, the one that talks to a smartphone. That's a monitoring system, but for your body.

We're IT. We do monitoring systems, so lets take a look at this one!

Figure out the SLA

What is the SLA of your body? Only you can really know that.

Figure out the Service Level Objectives

Ah, some meat. What things define meeting you internal SLA?

  • Able to climb 4 flights of stairs without tunnel vision.
  • Able to eat dessert once in a while.
  • Able to sleep more than 2 hours in a row.
  • Get pregnant.
  • Cholesterol below a certain level.
  • Weigh less than 200lbs/90kg.
  • 95th %ile blood sugar numbers don't go above a certain level.
  • Menstrual cycle longer than 24 days.
  • Able to bike 10 miles a day for a week.

Whatever it is. Define what metrics need to pass in order to determine if your SLA is being met.

Figure out what needs to be monitored to track the SLOs

Some of these SLOs are more concrete than others. Being able to climb 4 flights of stairs without tunnel-vision may take monitoring several things, where 'able to sleep more than 2 hours in a row' may only need one monitoring point. You may not even know what needs monitoring, it may just be an experience of running the system and you figure out whether or not you're meeting an SLO by aggregate analysis of time-series data (a.k.a. reports).

Some monitorables:

  • Blood pressure
  • Heart rate
  • Blood sugars
  • Blood oxygenation levels
  • Lipid profiles
  • Triglycerides
  • TSH levels
  • Estrogen & progesterone levels
  • Testosterone levels
  • Menstrual flow tracking
  • Sleep tracking
  • GPS tracking
  • Activity type logs (walk, run, bike, canoe, kayak, rock climbing, etc)
  • Mood tracking

Figure out how to monitor them

Ahm...

For a lot of these, monitoring is done once a year (lipid profiles, triglycerides), or if they're being actively tracked once a month. A few can be cheaply done continually with today's technology (GPS tracking, sleep tracking, heart-rate), a few others aren't cheap but still can be done continually (blood oxygenation, blood sugars). Some are self-reported and don't lend themselves to automation very well (activity type logs, menstrual flow tracking, mood tracking).

This is where the health-bands come in to play since the human body doesn't come with built-in performance monitoring hooks. They're increasing in sophistication, but the under $200 market only automatically monitors a few things:

  • Sleep tracking
  • Paces taken
  • Heart-rate
  • GPS tracking

Fancy analytics allow these four streams to be a proxy for a lot of the others, but there are only a few SLOs this set of measurements can support. This is why the Health Apps we're seeing explode on the market are all about tracking weight with maybe some other things on the side; they're using these four measurements the same way CPU/RAM/Disk/SWAP is used by IT infrastructures as a proxy for actual application issues, and like IT infrastructures that rely on proxy measurements for everything a lot of nuance goes missing. They will need a lot more direct-measurement to get that nuance.

These do not support comprehensive SLA monitoring, so there is no wonder they're getting a lot of flak for not being comprehesive.


Unfortunately, they're being advertised as comprehensive health monitors when they're not. Even for self-reporting based monitorables such as menstrual flow and mood tracking, they're not getting attention within the applications; for actual-comprehensive monitors they would still be self-reporting so may as well start building that feature-set now. I can only speculate why these features aren't present, but I strongly suspect Minimum Viable Product has a lot to do with it. In the mean-time, speciallist applications will pick up the slack.