I will be buying an Apple Watch, and I'm really optimistic about its health implications. I've totally bought into the idea of little computers that can passively track my movement.
I've been wearing a Fitbit One every day for almost exactly one year. I've also had an iPhone 6, complete with its M8 motion coprocessor, since November 2014. I guess because I had the Fitbit first, I just kept using it for step counting. Even though the iPhone 6 has been persistently counting much of the same information in the background, I never paid much attention to it because the iOS 8 Health app is just not that great for visualizing data. However, I recently came across two apps that do a great job of showing walking data: Fitport and Pedometer++. I recommend both.
But as the Apple Watch ship date nears, I'm really looking forward to seeing the Watch's health app UI and its other features in action. As much as I've loved using my Fitbit, I know that if an Apple Watch and iPhone 6 can do just as good of a job at counting steps and floors, I'll probably end up shelving the Fitbit.
So. I figured now was as good a time as any to really try to understand how my Fitbit One and iPhone 6 count steps and floors differently, if at all.
Overview of comparisons
I know from casual checking that both the Fitbit and iPhone 6 show extremely similar results when compared over small time frames. For example, both show precisely the same floor counts and very similar step counts for typical daily movement within my home.
But when comparing over a longer time frame, like the 90-day comparison you'll read about below, the results vary much more. To go even deeper, I analyzed the results of the Fitbit and iPhone 6 during two different walks in (very hilly) Greenville, South Carolina, where I live.
I also compared my Fitbit and iPhone 6 in a more "industrial" setting and even looked at how they worked when they weren't supposed to be counting: elevators and car rides.
The point of this exercise wasn't to be super scientific but rather to get a feel for how the numbers vary for my own personal use.
But First: How the Fitbit and iPhone 6 work
Both the Fitbit One and iPhone 6 have an accelerometer and barometer that they use to measure steps taken and floors climbed. Beyond knowing the general "hardware" used by Fitbit and the iPhone 6, however, they're really a black box on the software side. All we know is that they each use some sort of algorithm to identify motion patterns indicative of steps. One specific that we do know about Fitbit: one floor equates to a 10-foot rise in elevation. Apple does not publish anything specific about the elevation required for a floor.
Results over a 90-day period
I think it's worth comparing the Fitbit and iPhone 6 over a longer time scale. Fitport lets me look at my iPhone 6's HealthKit data over several different time frames. The longest time frame other than "all data" is 90 days, so I decided that would be long enough for the sake of comparing to Fitbit.
I just want to say that from a fitness ego perspective, it's hard to pick a worse segment of time than the 90 days leading up to this March blog post because I really suck at walking in the winter time and rarely even sniff the usual 10,000 daily step goal. But anyway, here's how the iPhone 6 and Fitbit compare from December 23, 2014 through March 23, 2015:
- Fitbit: 509,482 steps and 1009 floors
- iPhone 6: 506,909 steps and 735 floors
I would never expect an equal step count over such a long period because there are too many variables. For example, there are plenty of Saturday mornings when my iPhone 6 isn't in my pocket while I'm rolling around on the floor with my kids, but my Fitbit One is usually clipped on. And in the year that I've owned a Fitbit, there have been several multi-day stretches when the Fitbit battery died without me knowing it—meaning that it wasn't recording any data at all.
All that considered, the step counts seem remarkably similar.
The floor count difference is much more interesting. At first, I was really stumped considering how both devices always seem to register floors climbed in my home the same when observed over small time periods.
Then it occurred to me that I live in a fairly hilly part of the country. Greenville, SC is in the foothills of the Smokey Mountains, and it's hard to walk more than 20 feet in any direction without going up or down.
Given that I regularly walk through the trails, parks, and sidewalks that weave in and around downtown Greenville, I thought it made sense to pay closer attention to what the Fitbit and iPhone 6 were counting there.
Two walks into Greenville
I have several walking routes that I take into the downtown area on a regular basis. I picked two different routes to "study" how my Fitbit and iPhone 6 counted steps and floors across a variety of elevations and walking surfaces.
- Walk 1: A 1-mile walk covering some fairly steep elevation changes
- Walk 2: A 1.5-mile walk with more subtle elevation changes
I used MapMyWalk to measure the elevation profile and exact distance of each route. MapMyWalk uses my iPhone 6's GPS system to measure distance, and it uses USGS data to determine elevation. I thought these would both be interesting "checks" on the Fitbit and iPhone 6's step and floor algorithms.
Walk 1: Steep Terrain (1 mile long)
The first walk is probably my favorite of all. It's a truly spectacular walk, which starts at a relatively high elevation on the outskirts of downtown, then descends rapidly into the Reedy River basin in Falls Park, weaving through little gardens, water falls, and bridges. The lowest elevation occurs roughly midway through the walk, near the base of Reedy Falls. It then rises up sharply past the falls and into the heart of downtown along Main Street.
I've made this walk countless times, but until I actually graphed the elevation profile, I never realized how U-shaped it was.
Walk 1 Elevation Profile
As you can see, the iPhone 6 counted 70 more steps than my Fitbit and 4 fewer floors:
Walk 1: Steps Counted
The step count difference is somewhat surprising. iPhone 6 came in roughly 4% higher. 4% of one mile is a little over 200 feet. To me, this seems like a really big difference. Based on other data, I'm fairly sure my personal "steps per mile" metric is roughly 1755 steps per mile. I'm 6'4" and have a fairly long stride, so I'm well under the standard 2000 steps per mile that you often see quoted.
It's likely, however, that I take a different number of steps when walking up or down steep terrain. So it's hard to know if the iPhone 6 is truly "wrong" here. It's also possible that steeper terrain changes my natural step motion, introducing opportunity for measurement error.
Walk 1: Floors Counted
I was paying close attention, and neither the Fitbit or iPhone 6 recorded any floors at all on the first half, which is what I would expect since it was entirely a descent. The entire 4-floor difference occurred in the second half.
Roughly midway along the second half, I took a fairly tall staircase to get up to street level. I don't know the precise height of the staircase, but I would estimate that it's much closer to two standard stair flights than one. Indeed, this was the one and only point along the one-mile walk where my iPhone 6 counted any floors at all (i.e. the 2 floors that it counted total).
It seems that the Fitbit is much more sensitive to elevation changes than the iPhone 6. So which one is "right?"
I guess it depends on your semantic persuasion. In a strict sense, the iPhone 6 was more accurate in counting literal floors (the only true staircase). In a more "feel good" sense, however, the Fitbit's floor metric seems like a better measure of effort spent going "up." In terms of heart rate, ascending 81 feet over a roughly half-mile stretch is quite different than walking the same distance on flat terrain. It's nice to get some credit for that.
It's entirely possible that the iPhone 6's floor counting algorithm is tuned to look for more of a stepping motion, as you would make going up real steps, while the Fitbit is satisfied as long as its barometer indicates upward movement while you're making any kind of step motion.
Walk 2: Less Steep Terrain (1.45 miles)
The second walk is longer than the first, but isn't nearly as scenic. It does, however, offer an interesting elevation contrast with the first walk in that it follows a more gradual decline along a road that leads to downtown Greenville.
Walk 2 Elevation Profile
Walk 2: Steps Counted
I suppose it's not surprising that two different pedometers would yield closer results over flatter terrain, but I think it's pretty impressive that they are so close. Both devices also show a pace of about 1755 steps per mile, which as I mentioned earlier, is consistent with the typical number of steps I cover per mile while walking casually.
Walk 2: Floors Counted
Coincidentally, Walk 2 also featured exactly 2 real flights of stairs, which the iPhone 6 counted precisely. Once again, the iPhone 6 counted only true floors, while the Fitbit's floor count is more indicative of walking up hills in general.
I have to say, I'm a bit surprised that the Fitbit counted so many more floors given my perception of the elevation changes that I encountered, but I think this just further underscores the Fitbit's sensitivity to elevation changes while your feet are moving.
Counting real floors
Since my two walks show just how differently the Fitbit and iPhone 6 count floors over varied walking surfaces, I thought it would be interesting to see how they performed in a more controlled, "industrial" setting. So I decided to walk up to the sixth floor of a parking deck in downtown.
According to at least one source, the standard height of a parking deck story is 10 feet. The parking deck I chose, to me, looks like most any parking deck, but it's worth noting the first flight was slightly shorter (maybe 25% shorter) than the others because the first level of the deck is actually a little below street level.
At the 6th floor, the Fitbit was dead-on at 6 floors, while the iPhone 6 recorded only 5. It's quite possible that the shorter first floor caused the iPhone 6 to fall just a little short, but that's impossible to know. Unfortunately this parking deck only had 6 floors, so I couldn't go a little higher to test that theory.
It would be interesting, if tiring, to test this in a really tall high-rise to see just how much, if any, the two devices diverge in such a continuous, controlled setting. If you do this, let me know how it goes.
But as I noted earlier, both the Fitbit and iPhone 6 consistently measure floors climbed in my home with equal and exact precision.
Effect of non-walking motion
First World humans have many ways of moving up, down, and sideways that don't require moving feet. Obviously, just being able to detect small changes in barometric pressure isn't enough to detect steps. Otherwise, we could really game step counts by taking elevators and escalators.
Both the Fitbit and iPhone 6 performed equally well (zero activity) in an elevator in my testing. In other words, both devices are really smart about the motion pattern that defines a human step going up stairs.
Driving in a car, however, is a tougher test than an elevator because of all the little bumps and undulations in a typical car ride. I was curious what effect, if any, driving had on the Fitbit and iPhone 6.
I tested this by looking at my Fitbit and iPhone 6's step counts before and after driving across town. It was a roughly 15-minute drive that covered everything from residential roads to a 6-lane interstate.
The Fitbit correctly recorded zero steps, but the iPhone 6 logged 18. This surprised me a lot.
First, I was impressed that the Fitbit was good enough to know all the little ups and downs were not the result of taking steps.
Second, it's funny that the iPhone 6 got beat considering it has the potential to gather so much more contextual information. Compared to the Fitbit, the iPhone 6 is a super computer. My iPhone 6 is totally aware of my current speed, which it could use as a check on the step count. For example, if my body is moving at 65 miles per hour across the earth's surface along a coordinate path that matches a known interstate route, I'm probably not walking.
As with any arbitrary measurement system, it's more meaningful to look at trends than fret over individual data points. Fitbit and iPhone 6 are both terrific at measuring steps, and both inform you about your movement over time. If you walk from A to B, both will credit you steps for that, even if it's a slightly different number of steps.
When it comes to measuring floors, the Fitbit is clearly more sensitive to elevation changes than the iPhone 6. However, the iPhone 6 is a networked, location-aware computer. If elevation really matters, it's probably better to use something like MapMyWalk to more accurately measure vertical distances anyway. A "floor" is just an arbitrary unit of height after all. It really only makes sense for people who mostly walk indoors.
Rather than get hung up on data accuracy, I think it makes sense to focus on the main goal: move more. I'm absolutely fascinated with the fact that small computers can constantly measure my motion and give me incentive to move more by constantly informing me about my movement patterns. I fully expect the Apple Watch and its future descendants to take this to an entirely new level.
I'm no anthropologist, but I believe the version of the human body we inherited evolved to move around a lot—certainly way more than we move in modern environments. If the first generation of computers made us sit down, hopefully the next generation will put us back on our feet.
* * *
3/28/2015 Update: Elliptical Test
I decided to see how the Fitbit and iPhone 6 differed on a NordicTrac elliptical. I was on the elliptical for 20 minutes straight, covering a (supposed) distance and elevation of 1.7 miles and 895 feet, respectively. There were a variety of resistances and inclines.
As I expected, neither device counted any floors since there was no actual change in elevation. Interesting that the Fitbit counted so many more steps, though.