11. The Takeaways
1. Not all movement is equal.
2. We can track everything, and yet understand nothing.
3. Metabolism is more complex than we– or our algorithms–
can yet account for.
4. The technology is there.
The personalization and informatics are not. (Yet.)
5. Health is not an engineering problem.
#RITNOS
12. 1. Not all movement is equal
a step is not a jog is not a jump
Raving
Running
Fitbit
Argus
Polar HRM
#RITNOS
Raving Running
13. 2. We can track everything
…and yet understand nothing
#RITNOS
Raving
Running
RunningRaving
14. 3. Metabolism is complex
more than we– or our algorithms– can yet account for
#RITNOS
15. 4. The technology is there.
the personalization and informatics are not. (yet.)
#RITNOS
16. Photo courtesy of @jetscott
5. Health is not an engineering problem.
but engineering can advance our understanding of health.
#RITNOS
My name is Michelle Shevin, I am a senior analyst at Luminary Labs, a strategy & innovation consultancy, and I am here to tell you about how I’ve been raving in the name of science.
It all began in 2012. I had followed up a normal work day at my standing desk and an 8 mile bicycle commute with a five hour electronic dance party, also known as a rave. I synced my pedometer and heart rate monitor to find I’d burned 3,500 calories over the course of the day– 3,000 of those at the rave. I was actually able to break down how much I’d burned per DJ set—from 600 calories during Benny Benassi’s warm-up set up to 1000 calories during Calvin Harris. Unbeknownst to myself and the incredible lineup of DJs, we were doing something…healthy. The rave blew my 8 mile bike ride out of the water. Scanning back a few days, I found the rave dwarfed a three mile run. Even my ballet classes didn’t come close, at around 700 calories per 1.5 hour class. From that day forward, for health reasons, I made raving a priority in my life. At Luminary Labs, we talk a lot about sensors and health tech, and upon hearing of my hobby, my colleagues wanted to know more.
For the uninitiated, raves can appear strange. But where else do you see hundreds of young people getting together for an incredible cardio workout? Have you ever heard of a workout that can burn a pound of fat in a single session, set you in a trance, and make you feel connected to hundreds of your peers? Historically, raves have been associated with underground warehouses, teenage indiscretion, and grown people sucking on pacifiers. But raves are really about a DJ’s ability to command the crowd to dance. With the mainstream explosion of electronic dance music, raving is an exercise revolution disguised as a really good time.
My colleagues were skeptical. So Luminary’s CEO asked the question: what would happen if an analyst tested five wearable devices for three months—while raving?
I was up to the challenge-- but could any combination of sensors provide satisfyingly rich analytics to prove my case? And so began a quest for the quantified rave.
Suddenly I was living the dream: a mandate to rave, which sent me in search of new venues and previously unconsidered rave logistics. Highlights included a 7:30am pre-work dance party and an early-evening private set with world-famous DJ Afrojack at the Sirius XM studios.
Going in to the experiment, I already had the devices at the bottom of the screen– an iphone running Argus, and the Shine, a fashion-forward tracker from Misfit Wearables. I was frustrated by the analytics, as neither provided visualizations as good as the now-defunct pedometer and heart rate monitor combination that had produced the graph you saw a few slides back.
But in my research, no one device in the market stood out as having particular promise.
Using my budget, I purchased the BodyMedia FIT, the fitbit flex, and the latest and greatest heart rate monitor from Polar.
And so I set out to stress test these five tracking devices and applications under sweaty, strenuous, and loud conditions. Would the bass interfere with the sensors? wondered one colleague. Would my heart rate fluctuate around romantic prospects? Asked our chief strategy officer. Would wearables make me rave harder, or serve as a distraction, questioned my boss… and I began to wonder, would I experience wearable-overload?
When it comes to self-tracking via wearables, it matters where you wear it.
1. Argus
GPS and accelerometer on phone + optical sensor to provide an optional snapshot of your heart rate.
Will automatically kick on the GPS to provide a visualization of your run or bike ride
step counts are consistently low, probably because I tuck my phone into my shirt while dancing and it remains close to my core. Conversely, if my phone is in my hand, it will overcount steps.
2. Misfit Shine (tracks motion and sleep using the same three-axis accelerometer as most other wearables)
algorithms trained to differentiate between walking, cycling, and swimming.
it is one of few wearables designed with fashion in mind and it is also waterproof.
On the down side, it’s tiny form factor makes it easy to lose.
I’ve found that Shine’s step count is sometimes high, except for when it’s low—since I wear it on my collarbone, a high reading probably means lots of jumping.
3. Fitbit Flex (offers quite a bit of flexibility, but also comes with some friction)
There are three ways to sync the device, which is either convenient or confusing.
I like that the Flex vibrates when I hit my daily steps goal, but it it gets wet every time I wash my hands, and it counts steps even if I am just gesticulating. It’s counting steps right now.
Battery life is also an issue: it needs to be charged more than once a week, which isn’t frequent enough to make it a habit, but is frequent enough to make it annoying.
4. BodyMedia FIT (combines four sensors to offer a more accurate picture of activity)
The three-axis accelerometer, heat flux, skin temperature, and galvanic skin response sensors track motion, calorie burn, and something called METs, which has something to do with how efficiently you burn calories.
The combination of sensors makes the calorie burn estimation closer to the results you get from a heart rate monitor, but the tracker itself is bulky, a little uncomfortable, and prone to falling down your arm.
And the noise it makes to alert you it’s lost or made contact with your skin will leave the distinct impression on your co-workers that you’re raising a Tamagotchi.
5. Polar H7 heart rate monitor provided a reliable baseline of my activity level. Sure, it’s a little unwieldy for all day, every day wear, but for tracking the impact of particular activities it can’t be beat.
After three months, I anticipated having a fuller, more accurate view of my health, but I couldn’t even get a complete picture of just the raving. Even after massaging the data to compare efficacy across devices, I ended up with large gaps.
Over the course of four proper raves and three epic dance parties, I was intrigued but not surprised surprised to find that while step counts were pretty consistent, Calories varied significantly across devices and activities, demonstrating differences in the complex algorithmic secret sauces that translate the somewhat arbitrary metric of steps into the somewhat arbitrary metric of calories.
Afrojack ended up being my best workout, burning more than 600 calories per hour– no surprise to me, as I was very stoked to see one of my favorite DJs from one foot away. I raved so hard that the Shine registered 1.5 hours of “intense cycling“– coincidentally, the same analysis it gave to an accidental trip through the washing machine. (CLICK)
Most confounding for the devices was the night before New Year’s Eve, when I followed up a somewhat unsatisfying rave at pier 94 by running home down the West Side Highway. (PAUSE). My heart rate monitor, the most authoritative source of calorie counts, tracked me burning over 1200 calories, but the Fitbit only registered about 850, perhaps thrown off by the juxtaposition of raving versus running.
After all was said and done, I ended up coming to the following conclusions.
I’ve learned a lot about both raving and wearable computing, and here are the five main takeaways I want to describe:
Not all movement is equal, and our analytics don’t account for this. A step is not the same as a jog or a jump, let alone a fist pump.
We can track everything, and yet understand nothing. Tracking does not necessarily lead to true understanding of health and wellness.
Metabolism is more complex than we– or our algorithms– can yet account for.
The technology is there. The personalization and informatics are not. (Yet.)
Health is not an engineering problem. But engineering can advance our understanding of health.
Here are some analytics from that rave and run combo. Fitbit shows me the whole day, and can differentiate the intensity of raving versus running in terms of calories burned. Argus automatically picked up the run, which is probably it’s best feature, but as the readout of my heart rate during the activity shows, I was working out at the same level for two hours before the run. In fact, by 2:30 in the morning, I’d burned 1238 calories, far beyond a 1.5 mile run. But Argus didn’t pick it up the raving for analysis because it didn’t register as running. And the difference in the motion I generated also threw off the fitbit, which underestimated the calories burned by about a third.
On every device, all this nuance– a step versus a jump versus a jog—gets lost after the algorithm spits back a simplified metric of “steps. Raving is more about jumping, shaking your hips, waving your arms, and fist pumping than it is about taking steps.
In the end, what do “steps” really mean? It’s great to have a target to shoot for, but it would be better if the analytics would distinguish between different kinds of movement.
Though a consistent pattern of movement is discernible across devices, at the end of the day all I’m left with is an interesting picture of my odd sleeping habits. Problematically, the devices are designed with certain assumptions about how we live, and these aren’t true for everyone. For one thing, midnight is an arbitrary barrier. Only the fitbit and the heart rate monitor allowed me to easily view analytics for events in the middle of the night, which happens to be when I workout. Over the test period, I almost always hit my daily movement targets and often far exceeded them, but I didn’t lose weight. Tracking my activity in five different places made me data-obsessed, rave-crazy, and slightly neurotic, but it didn’t make me eat less. Health and wellness are more complex than the picture painted by my tracking devices.
All of these devices track movement, and offer a calculation of calories burned. But what about the calories we consume? Our bodies make a distinction between a calorie from soda and a calorie from lentils, and yet even if I went to the trouble of painstakingly logging every meal in the BodyMedia FIT or another interface, each calorie would count the same. Do we even have the necessary understanding of nutrition to properly equip the algorithms? As this diagram of metabolic pathways shows, these processes are complex.
We have accelerometers for movement, pulse detection for heart rate, and heat sensors for sweat, but what is actually going on inside our bodies– and how can we track it?
The holy grail of health tracking is passive tracking of both calories consumed and burned—something that may be possible with portable spectroscopy. At least two companies claim to have miniaturized spectroscopy– a technique that can identify distinct materials at the molecular level, quantifying nutrients, calories, allergens, and more.
Tellspec is promising a handheld food scanner, while Airo Health has debuted a wristband concept that could potentially provide an accurate picture of both calories consumed and burned by tracking metabolites in the blood. Airo stopped accepting pre-orders after their introduction to the market was met with a healthy dose of skepticism, but the technology is there. It just needs the right informatics.
The FDA has already approved an ingestible sensor by Proteus Digital Health that will track medication adherence– how long until a dental implant can track nutrition as we consume it?
There is work to be done on batteries. There is work to be done on informatics, visualization, and algorithmic sensitivity. There is work to be done by behavioral psychologists: Somehow we’ll have to account for the fact that seeing I’ve walked fourteen miles in a day usually just makes me want pizza. But the revolution is underway.
For better or for worse, we are witnessing the infancy of the algorithmic optimization of our lives. But health is not an engineering problem. 10,000 years ago, when we were literally chasing down our food, we didn’t need wearables to track our health. But as demonstrated by the trajectory of chronic disease, somehow it seems we’ve engineered away our intuition on when to move, what to eat and when to stop eating. Health tracking can arm us—and our doctors—with better information, and thus help us make better choices.
Though it’s still early days for wearables, this cyborg revolution will be made conspicuous by wearable computing, until computing dissolves into our textiles, and eventually into our bodies themselves.
Technology shouldn’t distract us from our experiences, it should support our enjoyment of the moment by quietly and passively keeping an eye on the bigger picture for us. And in my experience at raves, tracking did just that. Ultimately, raving provided an excellent basis of comparison across devices and gave shape to the barriers, limitations, and promise in the wearables market. The future of self-tracking may hold the key to combatting chronic disease and shifting the cultural conversation around health and wellness, but until then, I encourage you to join me, raving in the name of science.
If you’re curious what all the fuss is about, scan the QR code for a Spotify playlist that serves as a good intro to EDM