Analytics Blog

Get Your Data in Shape in 2014!

Have you given up on your New Year Resolution yet?  Despite all the efforts put in marketing wearable devices and the education provided by movements like the Quantified Self , data still shows that about 75% of us give up on New Year Resolutions within the first week!

Why is it so?  I had the opportunity to present at the first QS conference a few years ago and the passion and technology I saw there was the origin for everything we now know as the “wearable device industry” (a market bound to reach $12B in 2018).  For a few hundred dollars now, anyone can build a system that tracks their health and exercise goals.  This should make it easier for all of us to keep our resolutions, right?

Well, think again.  Even with good data and good systems, our ability to learn from data quickly is still encumbered by very basic challenges such as referencing data points across multiple sources, or easily establishing correlations across datasets that might not be obviously related.

Companies are constantly wrestling with such challenges, but since I’m quite a data geek when it comes to collecting my own data (not as much as Stephen Wolfram though), I thought I’d explore some of the potential that hides in my zeroes and ones.  Let’s see if this can serve as an example of what might happen with data, in your life or at your work.

Mi Data Es Tu Data?

To make it easy, I’ll focus on 4 devices and techniques I use: a Withings scale for weight, Nike+ and Jawbone bands for activities/sleep, and IFFTT to automate the collection and mash-up of my data in the cloud.  (I use Alpine software for predictions and correlations, more on that later).

#1: My runs.  I keep records of every single run in the Nike+ system and measure my speed, pace as well as time of exercise.  What I know is that I recorded 106 runs in 2013, 64 in 2012 and 70 in 2011.  I can tell what day and what time I run the fastest and the longest.  But what about the factors that impact my performance the most?  Well, for that, I’ll have to look somewhere else…and so it starts…

#2: My up/down activities: I use Jawbone for non-running activities as well as my sleep data.  I can tell what nights I’ve slept the most (Wednesdays) and the ones I’ve slept the least (Saturdays).  That’s great but what if I could identify that if I slept 7 hours and ran the day after that, preferably after 4pm, my running performance would be best and therefore positively impact my weight?…that’s in a 3rd source, which, luckily I also have 🙂

#3: My weight: I use a Withings scale to keep track of my weight, muscle and fat numbers (I’m not just interested losing weight, I’m interested in fat and muscle levels too).  By looking at my data in Withings, I noticed that that my biggest weigh gains have occurred on Sundays and Fridays.  Week-ends seem to be big for weight fluctuations but, again, it’s difficult to draw any conclusion from this data in isolation.

Wouldn’t it be great to determine what happened the times I lost or gained the most? Did I sleep more or less?  How much did I exercise prior to those changes?  What would be the optimal exercise routine if I wanted to lose 5 pounds for instance?

Most of the answers I’m looking for will come from joining and modeling all this data in one place.  Doing so, I’ll be able to understand the consequence of my activities and also plan better for maximum impact.

Do you have the same challenge, at work or at home?

This January, our Head of Products, Joel Horwitz, will be introducing a program to “get your data in shape”.  The program will be applicable to such problems as the above but also should work for predictive and advanced analytics issues at home and at work!

Hope to hear from you soon!  Have a great 2014!

PS:  If you want to find out more about The Quantified Self, watch my interview of Gary Wolf below!