My Fitbit has rarely left my belt, or wrist with the latest trackers, since I got it in July 2012. The devices aren’t perfect but for what I use them for it’s perfect. They are a reminder that I need to stay active otherwise I could revert back to my heavy years.
With the Fitbit I have:
- taken 18,555,575 steps
- moved 10,651.24 miles
- climbed 42,381 floors
The Fitbit website will give you these totals. It will even give you some great charts with friendly filters to drill into your data. They also sell a Premium service to give you access to more of your data. I did buy it one year and I did love that it ranked you against your peers. Yet, I still wanted to do deeper analysis and review of just my own personal data.
A couple months ago I started to explore the Fitbit API. When I got to the section on the Intraday Time Series I went into brainstorm mode. To be able to get the level of granularity of total steps taken every minute I’ve had a Fitbit!
So I registered a personal application with Fitbit and began development. I wrote a series of Python scripts that executed API calls and stored the data into a Sqlite database. In a future post I will do a write up on those scripts. For now let’s start looking at the data.
My first goal is to look at the steps by minute to find out when I’m most active. My data set contains:
- 2,026,261 data points collected from Fitbit
- 99.57% complete (Fitbit had a record of 18,475,902 steps out of 18,555,575)
- starting date of 2012 July 21
- ending date of 2016 May 27
The first question I wanted to answer was: How active am I throughout the day and how does the seasons affect it?
I suspect that:
- I’m not active when I’m asleep (on a normal night 21:30 to 05:30) regardless of the season.
- I’m more active later in the night during summer.
- I’m less active later in the morning during winter.
- I’m hard as nails and will be active after work (16:00 – 18:00) regardless of the season.
I decided to use just full years of Fitbit use (2013, 2014 and 2015). I also decided to create a heatmap of the data so any trends I didn’t suspect will stick out.
Here is the chart:
The chart confirms my suspicions. Yet, there are some interesting things I didn’t expect:
- In September I become active at noon. I cross referenced my running log and found I don’t run as much during the summer. When September brings the cooler temperature I start running during lunch to get it out of the way.
- A couple hours after lunch I go into a coma and my activity drops around 14:00 regardless of the season.
I have not answered my question with satisfaction. I’m now left with more questions and a desire to compare my steps with other data: like the weather of those days. I will re-visit this question in the future. For now I need to look at the other data I’ve collected from Fitbit.
If you would like to download the aggregated dataset for the chart, I have made it available (XLSX).