What's missing from the Basis B1 "quantified self" watch
I've now used the B1 watch from +Basis
since August 1 last year. The watch contains a bunch of sensors that track perspiration, pulse, motion and temperature, and it syncs to an online web-service where you can see and "analyze" your data.
Personally, what I find most lacking is the website. You can choose between a "heat-map" showing how one indicator (e.g., average temperature) varied across 1-hour blocks over a two week period, or you can see a graph showing how a bunch of indicators changed during a day or a subportion of a day. There is - currently - no API and no way to export your data.
I find these "analytic" choices puzzling. They basically let you see "how things look now," whereas the point of such a watch for most users would be to learn something that can be used to improve some specified outcome. Off the bat, I can think of several things I hope to see that might make the watch more useful and exciting than it currently is. Basis is rolling out a new version of the webpage on the 21st of January, so what follows can be thought of as my wishlist (in addition to an API and data export option, of course):
1. Increased flexibility in terms of defining indicators and determining time-scale of a graph.
For instance, I might want to specify a couple of outcome measures (e.g., "average pulse while awake and at rest," or "total hours of sleep during a night") and see how these change over larger timescales such as weeks or months. "I was so stressed at work last October, let's see how that impacted my sleeping patterns and my three-day moving average daytime-pulse-while-at- rest ?"
2. Open and adjustable/shareable aggregations of indicators into outcomes
The Basis website does try to aggregate and combine the various measures into more meaningful indicators such as sleep, and "activity" and even "walking" and "cycling" - but the way it does this is hard to understand. My "daily insight" page, for instance, often tells me that I've slept 2 hours - perhaps because interrupted sleep causes some portions of a night's sleep to be allocated to the previous calendar day. I don't know.
I would ideally like to see the "code" or weighting/combination "recipe" that shows how the physiological measures collected are combined to estimate higher level outcomes such as sleep as opposed to waking, or "walking" as opposed to "riding a bike." If these are seen as proprietary trade secrets, it would still be good if users could themselves develop their own such recipes or algorithms and share them amongst themselves. This would also make it possible for a community of users to develop indicators of particular use to people with certain health conditions etc (e.g., number of times heart rate is outside of a specified bound during a day, or variability measures).
3. Ability to analyze experiments.
The point of a watch such as this is to learn something about yourself. We might want to know, for instance, how we would be affected by quitting coffee. Or a smoker might want to track the rise and fall of "withdrawal" effects as he or she quits. By allowing users to define and manually enter "potentially causal" variables such as coffee use, attendance at exercise classes, "meetings at work" etc., it would be possible to see how these are correlated with various outcome measures we are interested in. (E.g., "how is my sleep improved if I don't drink coffee after 15:00?", "How is my self-scored productivity at work correlated with sleep quality the night before?")
4. Ability to conduct crowdsourced experiments with other users
Related to the last point, it would be cool if a bunch of people with similar questions about, e.g., coffee in the afternoon and sleep, or alcohol drinking and various hangover cures, could be recruited and join together to make their own randomized experiment. This could be an "opt-in" thing where you are asked whether you would like to be informed of such initiatives, and if you are, what areas you would like to follow and be informed of.
5. Info on what Basis users have learned
This will just be correlational, but it could still be fun if Basis could analyze the aggregate data collected from all users and point out surprising correlations. While correlation does not equal causation, causation does often imply correlation - and surprising patterns could trigger interesting hypotheses that users might want to examine using crowdsourced experiments of the kind described in point 4.