Introduction & Prelude To A Health & Fitness Data Manifesto
In support of open data formats, open algorithms, realtime hardware, multi-dimensional data acquisition and the destruction of the health and fitness company “walled-garden” paradigm; the most limiting factor in creating novel systems that provide new insights to produce better outcomes from health and fitness data.
“When we set out on this journey, There were no doubts in our minds, We set our eyes to the distance, We would find what we would find…”
Sting, “Something The Boy Said” from “Ten Summoners Tales”
Where do I start?
Where DID we start? Well, when we originally “set out on this journey” our market was mobile location based gaming in a company named ZoneVR. Our purpose (pre Pokémon Go, BTW) was to get people of all ages more active by developing a Serious Gaming platform that enabled the player to use mobile device sensors combined with fitness and health wearables to control live action, location based games.
Instead of check-ins and simple GPS aided proximity detection for game pieces and enemy tracking and location, we wanted to augment this experience with motion gestures (jumping, slashing, etc…) and exercise recognition using device accelerometers and gyroscopes and magnetometers.
Instead of opening a virtual portal by moving into a small geofence, we wanted the player to also have to increase their heart rate by 15% to open the portal. To freeze an enemy we wanted the player to use their Bluetooth enabled EEG headset to activate the weapon by increasing or decreasing their level of concentration. And so on and so on.
Inadvertently, we began to experience the pain points associated with these hardware integrations. We also began to realize that the industry adoption of the wearables (as a feature) was secondary to the issues with player adoption of location based gaming in general. It would take time (thanks Pokémon Go!). And in the end, we realized that we had built a mobile wearable hardware stack that was capable of multi-modal data acquisition for health and fitness data, as a complete side effect.
Over the course of time, and with some creative pivoting and a new company we began to create apps, initially for physical therapy, using that software stack named “WearKit” and “FitKit”, and we entered into the digital health, fitness and telemedicine markets.
Where did it go and what did we learn?
Throughout the development and integration of these wearables, we saw that equipment manufacturers fell into one of two buckets:
1. The realtime open wearables with open formats that provided the low-latency communication necessary for realtime health and fitness measurements and player-character control (Polar Heart Rate Monitors, Pebble Watch, etc…) via Bluetooth
2. And the “walled garden”, batch oriented communication based wearables that were simply not well suited for realtime gaming, as the actual data was accessible after 15 minute delays (and if the device happened to sync), and needing to access the manufacturers (Fitbit, etc…) individual custom REST based cloud API’s
I found it interesting that the biggest industry players ALWAYS used the walled garden approach. You couldn’t access the original data in the original format in realtime. They always post-processed the data and made it as hard as humanly possible to export that data from their cloud into other systems. They knew that the data was going to be their bread and butter, and through subscriptions they’d make thousands of times more revenue from that data than from their hardware offerings.
And so, the name of the game for the big boys became:
1. Provide basic reporting and metrics for the lowest common denominator consumer to appease them.
2. Prevent the consumer from exporting raw device data in a timely fashion, make it difficult to interoperate with other systems as to prevent effective mining from their walled garden.
3. Buy out, consolidate, disable and prevent realtime open hardware manufacturers that provide the same types of sensors, features and functionalities from gaining a market share (e.g. Fitbit buying out Pebble and abandoning the hardware)
At the time I couldn’t blame them, we are in business to make money. But over time as their popularity increased and as I began to see real people with real health issues beginning to request and then demand access to their live raw data. Their requests fell on deaf ears and I began to realize the flaws in the walled garden approach. The walled garden reports and data points were not really useful to the customer and could only be taken so far (as far as metrics and reporting), and the customer should have a right to access their own data in realtime, for whatever uses that they chose.
It started with the Quantified Self community, a group of enthusiasts and self experimenters and reporters that had found a myriad of uses for that data, THEIR data. Eventually, that slowly began to expand from individuals into other health and fitness based companies that also wanted to provide more value, more alternatives, more innovation in these health and fitness markets by utilizing popular health and fitness trackers and avoiding having to delve into designing, manufacturing, distributing and selling custom hardware themselves. Some ended up having to do exactly that.
And so the walled garden wearable companies exploded on to the scene, modeled after Fitbit, each with custom hardware, each with custom cloud API’s with reporting and delayed/post processed data access, mirroring the rival they once railed against (and in some instances, admired).
They thought: “What can we provide that is better than the market leaders? Better algorithms, actionable intelligence, reporting and metrics that will blow them out of the water, that’s the new frontier!” And they all collected loads and loads of low dimensional data, so much so that we, apparently, have neither the processing power or real capability to do much more with than what currently exists and is readily available.
So where does that leave us? What “new” can we do with these systems, this data?
Digital health is nothing new, but personal health and fitness tracking is relatively new and has managed to remain in the spotlight over the last few years despite sluggish innovation in new wearable product features and functionalities. Compare any fitness band or watch and tell me what makes any of them distinct from the next?
We are witnessing a crippling stagnation in health and fitness data reporting metrics and limitations in actionable intelligence based on the health and fitness data that has been collected from hundreds and hundreds of millions of individuals utilizing this wearable hardware.
We continue to read negative reports regarding big data and analytics companies not having the processing power or even the know-how to produce the insights that we were really hoping for. But why?
Why can’t we do more with current wearable health and fitness data?
As I stated before, there are a glut of wearables on the market, from the big boys (Fitbit, Garmin, Withings, Wahoo, Polar, Apple, Google, etc…) to the no-name band wagoners that also believed “hardware is king” and that the collected data will unlock (outside of loads of recurring revenue) new insights into our personal health and that somehow, with the proper algorithms, we’d witness better outcomes.
We have watches, bracelets, straps and all sorts of gear with the same sensors producing the same generic low-dimensional data points. Motion sensors for step counting and generic body motion for sleep quality, weight scales and heart rate monitors and medical devices capturing blood glucose, pressure, and oxygenation measurements; it seemed as if recording this data over a period of time would unleash new practical insights into our general level of health and some how, through magical algorithms, we could mine this data and it would somehow provide actionable tasks and recommendations for improvement in some particular area. All by wearing that wristband or watch everyday.
We’ve collected petabytes of crap data, not because of faulty sensors or any issues related to data precision, it’s crappy because the data that we have collected:
1. Is limited in the dimensionality necessary to produce proper correlations in relation to the context that the readings were taken.
2. Is lacking semantic binding regarding the purpose and the environmental factors present during (and potentially affecting) the digital reading.
3. Is partitioned into multiple proprietary databases protected behind the device manufacturers “walled-garden”, none of which are easily interoperable or aggregated.
4. Is presented back to the user without retaining the original resolution or data points necessary to reproduce the original event when the readings were taken; reports are generated on behalf of the account holder using unknown algorithms and only this post processed data is allowed to “escape the garden”.
We’ve convinced ourselves that by recording these one-off measurements, and somehow by adding one extra dimension (time), that all of these magical insights would become apparent and would help to produce better health outcomes. We were sure of it, we set our eyes to the distance in the hopes that sometime in the near future we’d “find what we would find”.
Unfortunately that has not been the case, and the mistake that we made is not simply lack of multi-modality (from a sensor and signal standpoint) during data acquisition (DAQ), it has more to do with context and collection protocol itself, and it all starts in DAQ.
In our next article we discuss:
Data Acquisition — The issues of context, purpose, data dimensionality and fidelity in the Walled Garden
/blee/
Brian Lee
blee@quanticsystems.io
© 2017 Quantic Systems / Motion Solutions — http://quanticsystems.io
