(Another) Wearables Accuracy Study

[NoBoMac]

Saw this the other day. Link in story to the paper for those that want to put to use their stats class knowledge.

How accurate are wearable fitness trackers? Less than you might think

Wide variance underscores need for a standardized approach to validation of devices.

arstechnica.com

ADD: not going to stop using my AW as not getting too OCD about stats from it as not using anything else that want/need to compare against. It's THE baseline for me and as long as they don't totally change the underlying algorithms, I'm good.

 

[turbineseaplane]

I saw that also

The comments captured my feelings

Raw accuracy isn't really relevant (to me) so much as tracking the trends

Step count (as a random example) might not be real world hyper accurate, but the device is accurate against itself and can convey to me if I'm walking less, as much or more than normal -- which is more what I care about

Of course there are many other metrics touted by various devices where raw accuracy is more critical ... I've honestly always been suspect about those types of claims, depending upon the device and manufacturer

 

[Howard2k]

Makes total sense. All about the trend. Individual data points in isolation don't have as much value without a larger view.


Another good example is the 10,000 steps thing. If you're someone looking to improve your health and you regularly do 15,000 steps, then 10,000 steps is probably not right for you. And if you usually do 2,000 steps then 10,000 is probably not right either. These things seem obvious but they get lost in the larger "10,000 steps" mantra that gets recited so often.