Can you really trust your fitness tracker?
Why your activity tracker is getting it wrong
To understand why our fitness tech is feeding us questionable stats, you need to delve into the technology inside.
All trackers rely on a combination of sensors, from GPS and optical heart rate, to thermometers, accelerometers and sensors that measure galvanic skin response, bioimpedance and even ambient light. The Microsoft Band 2 has no fewer than 11 of these working tirelessly to monitor your every move. But that’s just the start of it. Once the sensors have done their best to capture what you’re doing in the real world, each manufacturer uses its own proprietary algorithm to estimate things such as your calorie-burn and step stats. Different algorithms mean you’ll see different results on your wrist and in your app.
“The results you get aren’t necessarily bespoke to the way your body works.”
For things such as calorie burn, the picture gets even more complicated. Each fitness band takes basic information about you such as your height, weight, age and gender, then uses that to generalise about how many calories someone like you will burn just sitting around being alive – also known as your resting metabolic rate. The algorithms work on population averages, so the results you get aren’t necessarily bespoke to the way your body works.
According to Dr Cedric X Bryant, chief science officer at the American Council on Exercise (ACE), it’s a tough thing to get right.
“Predicting caloric expenditure is a relatively complicated process,” he said. “There are certain assumptions that are made when developing the algorithms that translate movement activity detected by the devices into calories burned. Even devices with the best prediction equations will have some margin of error due to natural biological variability.”
His team’s research backs this up. According to the results from an ACE study that looked at the accuracy of calorie-burn tracking, leading trackers were off by anywhere from 13 to 60% with some devices overestimating and others underestimating.
Speak to the manufacturers and they’ll tell you it’s still early days, and that Big Data will eventually help us get more accurate tech.
“I think Big Data analysis will improve the knowledge… improve the algorithms.” Morten Urup, Jabra
“I think Big Data analysis will improve the knowledge,” said Morten Urup, director of product management at Jabra, whose Pulse headphones monitor heart rate in-ear. “It will improve the algorithms, thanks to the fact that you can start grouping people and analysing for patterns when you have data from hundreds of thousands of people. It’s something we’ve not been able to do before on a scale like this. Once you can, it’ll be possible to start improving the algorithm for gathering the data, and also the devices we’ll see in the future. You’ll see a general improvement.”
Steps too far?
Counting steps and monitoring general day-to-day activity levels ought to be simpler to keep tabs on, but even here we’re still not entirely nailing it.
The most common bit of kit inside an activity tracker is the accelerometer, which is used to measure movement – or acceleration. The accuracy of your data will depend first on the quality of the sensor inside your device (and not all accelerometers are created equal), and second where you wear the device.
Hip-worn trackers tend to be more accurate because it’s essentially impossible to move any part of your body without engaging your hips, whereas wrist bands can misinterpret movement of the arm for steps. For the same reason, most activity trackers track some types of exercise better than others. Walking and running tend to be more reliable than yoga or strength training. It’s also why I clocked loads of steps in the shower this morning when what I was really doing was scrubbing my unmentionables.
Continues on page 3