I recently got a new tracker, Basis. Before This I only had one all-around tracker, a Fitbit. When I only had one tracker, the Fitbit, I regarded the number I got from it as the truth. If it said I had walked 8355, I had walked 8355. I knew that it was not a direct truth. The number could be off but I had no way of proving it so I let it go. When I added Basis to the tracker stack, I wanted to know how they differed and how much they varied. So inspired by Gary Wolfs article about how his Nike+ FuelBand and Fitbit correlated we at sympho.me decided to do the same exercise on our devices; Fitbit and Basis.
What did I do
I analysed how Basis and Fitbit stack up against each other. They both use an accelerometer as their way of gathering data for steps. However Basis is a wrist watch and Fitbit is meant to be in the pocket or around the waist. Which means that they will have different algorithms for tracking the steps. A hand moves a lot when you are not walking, e.g. gesticulation. Basis solved this by not tracking the smaller movements you do.
How did I do it
I exported my Fitbit data into Excel, this can be done though sympho.me or a Google Drive script thats has been covered on the Quantified Self blog before. Basis does not yet have an API or export function so that had to be done manually. (Basis has said they are going to release an API in the future). The sample data is located in the bottom of the post.
What did I learn
The first thing I learnt was that the two devises where never 100% in sync and I decided that if they where within 300 steps of each other thats ok. I usually work on a standing desk and noticed that my Fitbit was tracking more steps that Basis, one time the Fitbit had a lead of over 800 steps. If this was the result of 3 hours in front of my desk, how many steps would I "lose" per year? However further investigations showed that they were not that different after all at the end of the day. On average (over 11 days) the steps are 283 different per day, favouring Basis. Thats a 3.55% difference, a difference I find perfectly acceptable. Some days where rougher than others with the biggest gap on 1568 steps on one day, which is far too high for one day but it does even out in the end so I’m satisfied with that, for now. In addition to the steps, there was one major fault between the devices, calories. Basis uses more sensors that just the accelerometer when it is calculating calories. Basis uses galvanic skin response (sweat) and a temperature. Fitbit only uses its accelerometer. Thus there is a big difference in the results here. With a total of 2544 calories difference over the 11 tracked, a average difference of 6.56% calories per day. There were days with as little as 12 calories difference, which is as close to perfect as you can get it. But overall the difference was an average of 212 calories per day, this time the favouring the Fitbit. This difference is very different from the steps, because if you walk 283 steps one way or the other its is not going to change much. It is less that 4 minutes worth of walking therefore not vital to your health. However 212 calories is a doughnut or half a litre of Cola, and imagine if you ate or drank that much per day because your tracker said it was all right. Check out this webpage to see how much 200 calories is in different food types. Remember everything in moderation!
Have anyone else found the same with their data? I will update with more data when Basis releases their API for a more longitudinal study.
By Erik Holland Haukebø