You might have seen that very interesting study that said fitness trackers are pretty bad at estimating calories burned. It was carried out in Stanford University School of Medicine's labs, and found that none of the seven wearable tech devices tested were accurate at estimating energy expenditure in the form of calories burned.
There was quite a bit of difference, though, with the most accurate – the Fitbit Surge – off by an average of 27.4% (still not ideal) and the worst – the PulseOn – off by a whopping 93%.
The devices were all last gen – the first Apple Watch, the Fitbit Surge, Samsung Gear S2, Basis Peak, Mio Alpha 2, PulseOn and the Microsoft Band. So there is of course now newer, shinier hardware, as well as some improved algorithms, but calories burned are generally worked out in the same way. So what needs to be done?
How calorie tracking works now
Essentially, fitness trackers currently take information from sensors like the accelerometer, which tracks your movement. Then they use their secret sauce algorithms, which differ from company to company, to give you the actual, running total of calories burned i.e. energy expenditure so far that day. Based on those estimates, you might make decisions such as what to eat, drink or not eat or drink.
Read this: Pavlok's AI shake-up of morning routines
So, for instance, Fitbit has posted a (sort of) explanation of how it calculates this. It takes your personal data on age, gender, height and weight and uses this to estimate your BMR – basal metabolic rate. It says this accounts for half the calories you burn in a day i.e. the calories you burn at rest, and that's why you'll always start the day with some calories burned.
Fitbit says its calorie burn estimates are based on your personal BMR, activity tracked by your device's accelerometer and manually entered activity.
This will be the basic method used by most fitness tracker makers, but as the study shows, it's clearly not enough when it comes to the activity tracked part of the equation. The highest level of inaccuracy was actually from sitting tasks at 52.4%, with walking and running both around 31%.
There's the argument that as long as trackers are inaccurate by the same number of points each day, you can still make progress. So if you sit around most days, the tracker should be able to clearly show you which days in a week or month you got out and went for a walk. But again, that's not good enough when a device's whole reason for being is to track your health and fitness.
To give us an idea of how complex this estimation is, workout programmes can make calculations based on: age, gender, height, weight, workout length, workout type, your baseline fitness level, muscle groups used, muscle content and (for weights) resistance and amount of weight being lifted.
One future challenge will be working out how much muscle mass you're using – this is one of the reasons why calories burned differs so much between different sports played for the same amount of time.
So we need to add more data points when tracking this activity portion, which is added to your calorie burn at rest. When thinking about additional sensors, these could include heart rate – which some of the devices in the study did indeed track and a hell of a lot more accurately than calories – and also metrics like perspiration, or to put it scientifically, galvanic skin response.
These also apply to calories being burned at rest, not just your workouts, and as we said this is where a lot of the algorithms are running into trouble. This is because if a device doesn't take into account things like heart rate, respiratory rate and even your skin temperature, estimates are based purely on movement – which can be difficult to map to specific activities, even just pottering about the office – plus the age/gender/height/weight info.
An expensive, lab-style calorimeter which measures your breathing and costs thousands of pounds isn't practical, so we're looking forward to seeing startups and companies exploring heart rate, temperature and GSR sensors. Whoever cracks calorie burn accuracy will reap the rewards.
If you want to read more about the Stanford study, you can read the whole thing here.