Greetings from our semi-regular1Whenever I have a few good thematically-related questions and nothing else that’s more immediately on my mind ask-me-anything series where I answer reader questions. As always, if you’d like me to answer a question, you can send it to email@example.com.
Today, I’ve got just one question to go through, because it’s actually three (good!) questions in one:
I was wondering if you had any thoughts on the accuracy of fitness trackers like Fitbit et al, especially in relation to their calories used calculation – are they a better metric than the usual estimate of 2100 for a reasonably active man, for example? Should one use this as a guide to how much you can eat?
The three questions I’d like break out and answer here are:
- Do wearable “fitness trackers” have any utility at all? Which ones?`
- Is their calorie math, specifically, any good?
- Should I use this calorie math (or some other calorie math) as a guide for how much to eat?
Do wearable “fitness trackers” have any utility at all? Which ones?
Information is only useful if you use it, and devices that gather information are only useful if you use that information.
That is to say: all fitness trackers are valuable if they provide you with information you will act on — even if that information isn’t particularly accurate. But also, none of them are valuable if you can’t or won’t do anything with the information. If seeing your daily step count or sleep score doesn’t get you to do anything differently, then devices that provide that information are just expensive, normally ugly watches.
This is my big beef with most of the “quantified self” movement, in fact — collecting data is pointless if you cannot (or just will not) use it to course-correct your actual behavior. Unless you’re going to make some Feltron-style art with it it’s all junk.
Now, if you’re the kind of person that is trying to increase your passive activity AND you’re the kind of person that sees 8000 steps on your wrist and walks around the airport or the office to get to 10,000, a fitbit or similar would probably be of utility to you. There is nothing magical about 10,000 steps aside from “it’s more than 8,000, and a quantifiable goal,” but being able to set and track a simple, accomplishable activity goal like that is useful for many people. 2in case you’re curious: 10,000 is a totally arbitrary number that marketers of the earliest Japanese pedometers picked out of thin air as “probably more than most people get,” so that more people would buy their proto-Fitbits.
As an example, I have a close friend who credits the purchasing of a step tracker — in conjunction with Pokemon Go! and its Harry Potter themed cousin, Harry Potter: Wizards Unite — as a major factor in her personal goal to increase her total activity by walking more. Quantifying her steps was a tool to help her gradually and consistently increase activity, and finding new positively-associated, active hobbies made it easy and fun.
Which is to say, there are many, many paths to becoming more active — which is a generally positive thing for your health and well being even if you never intend to become a “gym person”— if counting your steps helps you be more active, then I would consider getting a tool to help count.
Wearables are exciting and cool-seeming because with the right information (and the right parsing of that information), we could do a lot of interesting things. But I think we’re just now approaching a technological realm where this is even close to feasible.
So, if you don’t already own one, is it really worth the money? I don’t know. It again depends on if you would benefit from specifically seeing that data. It’s certainly not the only way to quantifiably keep on top of your activity — you could do something as simple as using a timer and going for a 30 minute walk after work every day (or going on three brisk ten minute walks throughout the workday, or walking a known route every day). Wearables have the benefit of tracking your additional passive activity, but unless you’re also making strides3 pun intended to change your passive activity habits, you’re not likely to see that much change day-to-day.
Beyond steps, these devices sometimes claim to do other, often more advanced things, like tracking sleep, heart rate, heart rate variability (HRV), body temperature, galvanic skin response4which is a proxy for parasympathetic tone, or other “stress” or “recovery” metrics. With a few notable exceptions, I think most of these devices and their “advanced” metrics are useless.
Case in point: I personally wear this guy, which has activity and heart rate tracking, but mostly like it for its ‘dumb-ish smartwatch’ functionality, and as an occasional tool for actively keeping track of my heart rate when I’m trying to do heart rate zone based cardio stuff, not as a core tool in understanding my behavior.
The heart rate information it collects in its “passive” (not working out) mode is so noisy and non-granular that I can’t do anything with it. In my experience, it’s marginally better than Fitbit’s heart rate offering, and it’s still just not good enough to give me a long-term picture of e.g. resting heart rate over time, or HRV, or anything that would be useful in any context except “what is my heart rate right now.”
Most wearable sleep trackers are equally dismal. They normally operate by detecting motion and turning it into “sleep depth” via a not-very accurate heuristic. In practice, they are just accurate enough to tell me if I slept terribly or slept abnormally well, which I don’t really need a device for — I can just tell that by how I feel when I wake up in the morning.
To put a fine point on it: just like steps, heart rate and sleep information that is not granular or accurate enough to be useful or actionable is just expensive noise. It’s comforting to collect this data (I’m keeping track!) but if you’re not using it for anything, it’s a distraction, not a tool.5n.b.: Many businesses could use this advice as well, but that’s a rant for a different day
If it were granular enough to be actionable, that would be a whole different story. If either sleep data or heart rate data (or both!) were just one level more accurate, you could them to develop tight, meaningful feedback loops about sleep, stress, activity, and how they’re all interrelated.
To that, there are only two wearables on the market that I believe might even get close to this promise: Oura and Whoop. I haven’t tried either of them yet, so I can’t fully endorse them, but their data collection seems sufficiently accurate and granular that they could provide even someone like me — who doesn’t really need a step count to stay active anymore — with actionable insight.
I actually just bought a whoop after many years of being pessimistic on wearables generally, and intend to get the other one to play with soon enough. Expect an update on both fronts on the blog in the coming months.
Is a Fitbit’s calorie math, specifically, any good?
There’s a pretty-recent study about that!
It had a reasonable sample size (60), and compared an un-named FDA-Approved “gold standard” 6for energy expenditure, this was a gas exchange indirect calorimetry device, for heart rate, the study describes it as an “ECG device,” which, there are a bunch of these, but probably some sort of not-commercially-sold multi-lead guy with the sticky pads and 12 bazillion wires for both heart rate and calorie estimation against several different popular devices (the Apple Watch, Basis Peak, Fitbit Surge, Samsung Gear S2, PulseOn, and Mio Alpha 2).
The lowest “overall” error (which was calculated with some fancy statistics) for both was given to the Apple Watch, which actually surprised me, but maybe shouldn’t, given its heart rate monitoring abilities were recently cleared by the FDA for certain medical uses.
When you dig in a little further, all of the devices tested are pretty accurate (within 5%) when it comes to active heart rate, For calories, less so: an average median error rate of over 20%, with the Fitbit Surge doing the worst (around 30% error), and the PulseOn (20%) doing the best.
Calorie math is weird, so let’s put that 30% error stat in context: a 30% underestimation in calorie intake for me (a 200-pound, relatively active man) would net an average weight gain of around 1.5 pounds per week, assuming I was trying to eat a weight-maintenance calorie intake. I personally don’t love calorie tracking generally (which I’ll address in the third part of this question) but even if you were to choose that route, that kind of error rate is …not really going to cut it.
The baseline calorie intake it recommends before it’s “adjusted for activity” is probably pretty correct — it’s most likely just doing that as an estimation of your Basal Metabolic Rate, which, if you’re metabolically healthy and of relatively average body composition7 not actually talking about bodyfatness here; carrying a ton of muscle is what actually messes up the equations , can be accurately determined by just your weight, height, and age.
That means that the error is coming when it tries to estimate your activity, and factor that into total calories burned. It might be accurate, it might not, and without secondhand verification, it’s extremely hard to say.
There is a silver lining, though. These are population-based “average” error rates. If you look at the data in detail, you’ll see that they’re not just randomly wrong. They seem to be pretty consistently over- or underestimating, depending on the person.
This makes intuitive sense: the problem here is not that these devices are so bad at measuring activity that they can’t estimate your activity reasonably well, it’s that the model it uses is based on a hypothetical “average” person with “average” activity patterns and an “average” resting heart rate, and that in the wide diversity of human bodies, virtually no one is actually that average person.
If you remember your high school science classes, the activity estimations are generally precise but they’re not particularly accurate. For a small fraction of people, these estimates are very accurate just by chance, and for most, they are consistently under or over in a consistent direction and magnitude.
This means that if you’re diligently tracking your calorie intake, you can actually find out how accurate these estimates are over the course of a few weeks by attempting to exactly match your calorie intake to the device’s “calories burned” estimate. If you gain or lose weight8 Which, as a reminder, is not a good proxy for health or even body composition. doing that (based on a rolling average), the number of pounds gained or lost divided by the number of days you were tracking divided by 3500 is the daily average error rate for you, personally.
Should I use this calorie math (or some other calorie math) as a guide for how much to eat?
This brings me to the third question: even if it is accurate, either inherently or because you figured out your own personal offset to make it accurate, should you use that (or some other guide) as a yardstick for how much you eat in a given day?
I’d be in the “no” camp on this one, even if we had tools to perfectly measure daily calorie expenditure.
My general thought is that calorie intake and daily activity should be considered, within reason, unrelated. I just don’t think it’s that useful or psychologically beneficial to play the daily “how many calories do I get today?” horse trading game based on steps or hours on the treadmill or anything else.
For one thing, it assumes a level of detail and obsession in tracking and estimating your food intake, activity, and calorie burn that is very difficult to get right in the best of circumstances — we just went over how easy it is to misestimate activity when you’re using a device that is purpose built to do that — and is not really worth the effort for anyone who isn’t a competitive bodybuilder, fitness model, or weight-class athlete.
More importantly, I’ve personally seen this kind of tracking lead quickly into a weird, dangerous reward/punishment cycle where you overeat and then run for hours, or use a slight uptick in steps for a day as an excuse to go absolutely off the rails on your nutritional habits.
Even if you want to count calories and macros (which I don’t think is strictly necessary, especially for travelers —a point I discuss at length in the book — but is certainly one effective strategy for managing food intake), I think a much more useful approach is to treat it as a process that is reasonably divorced from how much activity you do on any specific day. Given you have some sort of consistency to how much you move in a given week or month, the day to day variation in intake estimation and activity level both even out in the wash.
Use an online calculator to find a daily calorie estimate that seems like a reasonable starting point, eat that much for a while, and see if things are changing (or staying the same) in the way you expect via scale, body composition or whatever else you’re using to measure and define “success.” If the numbers are not moving in the direction you want, change your intake number or activity level until you see the outcomes you want.
Not only does this encourage productive, consistent food habits, which “20 minutes on the treadmill equals dessert!!” does not, it’s just easier to deal with.
Overall, I find this kind of detailed quantification — of steps, calories, or anything else — a very Silicon Valley solution to the problem. Detailed numbers make you feel like you’re in more control or have greater understanding of yourself, but it’s actually just a productized salve for much simpler, but harder to address problems: how do I change my habits? How do I live a healthy, active life? Why do I want to? Buying new running shoes is easier and more fun than changing your life such that you use them four times a week. Buying an app that makes it so you can’t access certain other apps during work hours is easier and more fun than actually interrogating and changing your relationship to technology and media.9 to be clear, these aren’t “better than thou” examples; they’re ones from my own life..
Buying something that purports to solve a problem you have feels like progress, but it’s not. Knowing your heart rate or the number of steps you’ve taken in a day feels like progress, but it’s not. It feels like introspection and action, but it’s not. It feels like something that will help you know yourself better and live a better, more active, more introspective life, but: unless you do something with the information it provides, it’s not. It’s just a distraction from those goals.