Daily Rhythm Tracking w/ Nike+ Fuelband

Hello, so I’m here to give a brief talk about the Nike Fuelband. I got one at Christmas. I have no official association with Nike, so definitely don’t take this as an ad.

I actually lost my Nike Fuelband about a week ago. I left it in the shower of the gym and when I went back a few hours later it was gone.

So it looks like this and probably seen it. And actually the reason I wanted it for Christmas was not for the data, but because it has a kind of a bling, like a geek bling that I really really like.

There’s a Nike app and they show you lots of pretty graphs but in my opinion the pretty graphs are mostly useless. You can’t really tell even what time of day things were like. You can tell you know there was a walk sometime after lunch you know that kind of stuff, but it’s super frustrating how non visible your activity is in there app and there doesn’t seem to be anything else.

So you can set goals, although it’s really really granular. You can set a 2,000 fuel goal or a 3,000 fuel goal. And sometimes you achieve them and sometimes you don’t. The other this is Fuel is a weird proprietary and infact I never understood what it was. I think it ends up being roughly one third of the number of steps you took in a day but it doesn’t mean anything to me and I don’t know if it means anything to anyone else.

So mostly with playing with the app and I really wanted to see the raw data. So it turns out Nike offers a developer API. You can go to their website right here and you login and you get all kinds of interesting structured data out and I fed that into a spreadsheet. And you get basically along the rows, there’s one row for every minute of the day and along the columns there’s one column for every day of the month. So I have an entire month of data here, rows 31, 1440 columns. It’s a lot of data and then each block is basically the number of steps I took in that minute.

So when you plot it you get a plot that kind of looks like this and you can see all kinds of interesting data. So these things here are me getting up in the middle of the night and going to the bathroom. And these things are me walking to work, and you know there’s some you know walking to dinner is in here and then there’ll be a walk home at the end of the day. And depending on what you did in the day like if you run errands you’ll see them in there.

And if you average it it kind of looks like the Nike data but you lose the interesting things like the middle of the night and other small activities.

So one of the things that I was interested in was precisely like this first thing, this is like when my day release starts. Everything before that is like I’m walking around the house, taking a shower, eating breakfast, whatever. But when that long walk happens that’s usually me walking to work. And so I basically wrote an algorithm that extracts the time of the first long walk, and I plotted it. And you can see my day usually starts about noon. I’m finally out and about. It does vary. Sometimes the first long walk, usually those 10 AM ones, usually I’m out of cereal or milk and I have to walk to the store.

And if you take that first long walk and you look at the number of steps each minute and you average it I determine how fast I walk which I though was interesting. It’s about 110 steps a minute. I’m a pretty fast walker I think. And something I discovered in the data is that every so often you’ll see that I’m not walking at 110 steps a minute and I’m like what are those things. Usually they’re in the evening and it occurred to me what it probably is is that I’m walking with my girlfriend and she walks slower than I do, actually significantly slower and it’s kind of frustrating for me sometimes. But in the data it turns out that it’s only 82.9 steps per minute instead of 110. It’s only about 30% slower so really it’s not that big a deal. It just seems like we’re crawling along.

And this is another plot that’s something which Nike also offers is like the total number of steps you take per day, and my number varies widely between 6,000 to about 15,000 except for this one crazy day and I was like what in the world did I do.

And I plotted it and I think actually it’s some kind of data abnormality.

So in the previous talk they were talking about context, and theoretically I should know what that was. So I looked it up on my calendar, I looked this day up and my calendar is completely blank for that day. So I actually think it is I think I’m sitting at my desk with this kind of repetitive arm motion and it’s mistaken that as steps. It’s not steps and it doesn’t look like walking because walking is like a flat line and 100. This is like random movements. So I decided it’s working and it’s just miscounted it somehow.

Which brings us to the things that it records. The things it records are like steps, fuel, and calories and the first question is, so how good at it with calories thing and what’s fuel. And it turns out that all three of them are highly co-related. Basically when one of the readings is high, the other readings are high. Which makes me think that probably it’s not recording calories terribly accurately because there’s lots of things like the previous slide where steps isn’t really walking. I mean I was sitting at a desk. I was not sweating doing that work, even though it recorded 10,000 step doing it.

Here’s another example of where clearly it’s not going to track calories because doing yoga class, yoga class is the flat part, and the big things on the other side that’s me walking to and from yoga class.

This is not my experience of how difficult yoga class is, so clearly it’s missing that calorie count not so good.

So in summery basically it tracks walking really well and you can see the start and end of your day and really cool things like when you went to the bathroom. I measure how fast you walk and you can see distance to places. And overall I think I had a really interesting experience with the data and I kind of wish I hadn’t lost it. but I also decided that I’m not will to spend $150 to get it back.