Step Counting Accuracy

fyoung1111

I don't expect my activity monitors to be perfect and think all of the posts about counting steps while brushing hair are just plain silly. All other things being equal, if you sit on your butt all day you will earn less than 2,000 steps and if you run or walk five miles you will get your 10,000 . . . MAYBE

I have discovered that my Fitbit Force has a very unusual blind spot in its step counting algorithm. While it does just fine at a lazy 3 mph walking pace and at nearly any running pace, it cheats me out of at least a third of my total steps if I walk at a healthy 4 mph.

Yesterday, I walked 3 miles at an average pace of exactly 3 miles per hour while using four different step counting devices getting the following results:

Fitbit Force 5,965
Fitbit One 6,216
Polar Loop 6,073
Runtastic Pedo 5,981

Then I upped my pace to an average of 4.01 mph over the same three mile course with the following results:

Fitbit Force 3,518
Fitbit One 5,416
Polar Loop 5,367
Runtastic Pedo 5,576

The Force number for the faster 3 miles HAS to be wrong. Forgetting that it is in marked disagreement with all of the other numbers, it would also indicate a step length of 54 inches. Not very likely for a 5' 9" guy with a 30" inseam .

This is just the most recent in a long series of experiments I have conducted to confirm this illogical phenomenon. The results are quite consistent. When I first reported this to Fitbit support, I got a lot of stupid responses like "wear your Force on your non-dominant wrist" or "set your stride length". For the last few days, they have been promising to call but haven't quite gotten that together yet.

I would love to find out if any of you Forces or Flexes out there get similar results. Keep it simple. Walk slowly while counting 100 steps in your head. Pause and see what your tracker reports. Then do it again at a brisk walking pace. I'll be happy to pass these results along to Fitbit. I'm pretty sure this is something they can fix as soon as they are willing to acknowledge that the problem actually exists. Notwithstanding my recent disappointing customer support experiences, I'm a big fan of the Company. None of these devices are perfect. I've used four different ones so far and UPS is bringing me a Basis later today.

If I ever do actually hear from Fitbit, I'll post an update.

Kimsied

Whether you reach 5 miles in 10,000 steps will depend on your stride length and whether you were walking or running. At 5' 1" my stride tested at 30.1 inches walking 4 mph, and it takes me considerably more than 10,000 steps to reach 5 miles walking. Running, more distance is covered in fewer steps so maybe running. Fitbit somehow "decides" whether to categorize our steps as walking or running and decides which stride length setting to apply. I mention that because I believe sometimes it will apply my running to some of my brisk walking. My running stride tested at 31.6 inches using a short strided jog that probably wasn't much faster than the 4mph walking. I am not sure the increments it uses, but I think it is at least minute by minute as this is somewhat connected to which "activity level" the fitbit credits us for and that seems to be minute by minute. In my case,I believe that Fitbit is crediting me with my running stride when it is giving me "very active minutes" which kind of makes sense because "very active" is meant to be vigorous and walking is usually considered "moderately active". But it isn't just the steps per minute, speed it factors in, it is also using something from the movement data--I believe either impact or up and down movement. I've had workouts involving jumping (jumping rope, plyometrics, etc) where it automatically credited me with a higher activity level even though the steps and "average pace" were lower than in some of my walks where it mostly credited me with moderately active. This is fair enough, jumping rope is more vigorous than walking for me. But I was really surprised to see the speed/pace wasn't as high. So there is a little more to it. I bring this up, because it sounds like your Force is not as sensitive to your movements as your One. I had read somewhere that torso worn accelerometers tend to be more accurate than wrist worn for a lot of activities--though the opinion was that the ideal would be devices you could wear in different places depending on the specific activity (and could somehow be claibrated to the changing locations). But where and how these types of devices are worn does effect what they can track. I have no experience with the Force or Flex. From othger's posts they seem so variable--like they work well for some but will either over or under estimate for others.

I think the customer service may have misunderstood since your Force was picking up fewer steps than your other devices. What distance did your Force and One give for the 3 mile walk? Did you use the activity timer to get that information? I think the Fitbit staff may have misunderstood that your force seems less sensitive not more sensitive. Asking you to wear it on your non-dominant hand should make it a little less sensitive since that arm whould move less throughout the day. To make it more sensitive, my understanding is that you would wear it on your dominant hand but set it for your non-dominant hand. Which calibration you use (dominant or nondominant) can apparently make a difference. The on lable use it to set it for the hand you wear it, but sometimes people have tweeked the sensitivity a little by playing with the setting and which arm they wear it on (often wearing it differently than they programmed it).

Are your Fitbit One and Fitbit Force stride lengths set the same? I think this setting should mainly effect the distance estimate and average "pace". but also it will depend whether walking, running or a mix of both were applied. So the settings for both are important. Sometimes it can be good to calibrate the stride settings for a more leisurely walk (whatever pace is natural for you) for "walking" and your brisk "in a hurry" walking pace for running. Of course, you would only want to do this if you rarely run or are using a different method to track your running. The different devices might sense the same movement different and class it differently.

fyoung1111

Wow. Finally someone more verbose than me . Seriously, thanks for your observations.

I did not know either Fitbit device had an activity timer. I did not record the distances provided by the trackers. I used Runkeeper and an iPhone 5s to log distance and speed. The Runtastic pedometer app was also running on that phone and linked to the M7 motion coprocessor.

The One was borrowed from a friend I had gifted it to in August. He is about my size but I doubt he ever changed the default for his stride. I have now set up a second Fitbit account and linked the One to me. Swapped my Polar Loop for his One. I'll make sure the stride lengths are the same in both before I conduct any further experiments. Having said that, my understanding was that the stride settings did nothing but convert steps into distance. I always guessed that the trackers differentiated between walking and running based on the frequency of the steps (steps per minute).

As a part of this obsession, I have read several scientific papers on the subject of normal human gait. As we walk faster, most of us accomplish it by increasing both the frequency AND length of our steps. Averaging the non-Force trackers for my little experiment yields step length and frequency for the 3 mph walk at 31.21" and 101.5 steps per minute respectively compared to 34.86" and 121.48 for the 4 mph segment. All of this casts doubt on the accuracy of one stride for walking and another for running assumption unless, of course, you always do one or the other at the same speed. While I am on this subject anyway, on a slow jog I seem to do a step length of about 37" at a frequency of 157 so there is a pretty big difference between walking fast and running slow. Most people switch gaits at about 5 mph. Anyway, I don't care about distance. If I did, I would use the GPS in my phone or Garmin Forerunner for that. I do, however, want credit for my steps!

When I started my conversation with Customer Service, I was not yet bench marking the Force against other activity monitors. It was simple matter of telling them I had walked around a 1/16 mile indoor track a billion times and counted 105 steps while my Force counted 60 or 70 or 80. Then I repeated these Force only experiments at various speeds on a treadmill to enforce the "rigor" and consistency of my little experiments. Believe it or not, they told me "Fitbit trackers do not accurately count the steps simulated from a treadmill, or other stationary exercise equipment. Your tracker uses changes in the algorithms of step count, therefore requires that you physically move forward in order to properly record steps." The Polar Loop had no such problem. Neither did the Force at speeds of 3 and 6 mph.

Still waiting to hear from the nice folks at Fitbit. They really are very polite and pretty responsive as long as you stick to email. I've been waiting two days now for the promised phone call :mad:

pinkraynedropjacki

I would also think that not everyone swings their arms at all while walking, I know for me the flex or force would be useless cause I get most of my steps marching on the spot here at the laptop & my arms are not moving at the time, same as when I'm walking to work or home, I usually have my phone in my hand and am not swinging my arms at all, even on my way home I have stuff on my shoulders that would make swinging my arms at all impossible without everything falling off the shoulders. Thank goodness for my One.

Of course it's not accurate exactly down to the last step, cause not everyone is the same when they walk.

fyoung1111

Fitbit says "Force will count your steps when you are pushing a stroller or a shopping cart because we do want to give you credit for this activity. That said, because your hands are not moving, your step count may be a bit lower during this activity."

While I have conducted no serious experiments, I have found this generally to be true. Same goes for walking with your hands in your pockets.

I have been very careful to swing my arms normally and consistently in my 4 mph blind spot experiments.:happy:

The One is incredible but they are awfully easy to lose. I never did it but a lot of people also drown theirs in the clothes washer. I lost my first One half a dozen times (it kept popping out of its holder) but found it every time but the last. Fitbit replaced it for free (one of the reasons I love them). I gave the replacement to my girlfriend when I got my Force and she lost it within two weeks That and the goofy wrist band you have to sleep in are the ONLY problems I have with the One.

pinkraynedropjacki

Not that easy to lose at all in fact. We have 3 people here all with a One each...not once have any of us lost them, or had them fall out or off. This is over a year of usage so far. If work correctly , like on you underwear at your hip so your pants/shorts can hold it in place.... then it never falls out or off & counts accurate. People who lose theirs are just not taking care of it & where to put it. I know I've also never lost it when putting it on my bra.

fyoung1111

I guess I need to start wearing a bra

UPDATE: Fitbit finally called. They don't seem to understand the weirdness associated with the fast walk only accuracy problems. They did (rightly or wrongly) determine that my Force was defective and agreed to replace it.

Unless I get a tidal wave of corroboration of the fast walk error from other Force or Flex users in the interim, I will be testing the replacement Force to see if it has the same problem or not and will report back here.

I hope Fitbit is right. Did I mention that I love the company?

luckydays27

I know a lot of people mention losing or washing the FB One. I have not been in that position yet but my husband has. He, of course, does not need or want to wear a bra and after losing his FB twice and washing it once, he decided that he would research the devices out there.

he concluded that the FB One was the best choice even with the rubber clipped case design so what he decided to do is find another case for it and put it on the lanyard. He came up with this one:



The FB One fits snuggly inside and is not going to fall out. There is even enough room to have the FB in the rubber case, clipped to the case logic case so he is doubly sure it will not lost.

He took of the keyring and strung a piece of cord through it, then burnt the ends so the line was one solid piece. He has found this to be a great option for him. The FB One sits at his chest level and he gets all the benefits that the One has without the headache or hassle of thinking its going to get lost.

I researched the FB Flex months ago and compared it to the One and found that it was not as accurate as the One. I bought my husband the Force when it came out (after the first FB was lost) and he found the accuracy just was not right as compared to the One.

Indiri13

I'm a fast walker most of the time (5'10" and 34" inseam if it matters) and I've noticed that out of 100 steps the Force usually counts between 60-75 of them. For me I think it's close enough, only because I want a trend compared to my weight loss, meaning the number at the end of the day doesn't matter as long as it is *consistently* off by about the same amount (it is).

fitphoenix

I would also think that not everyone swings their arms at all while walking, I know for me the flex or force would be useless cause I get most of my steps marching on the spot here at the laptop & my arms are not moving at the time, same as when I'm walking to work or home, I usually have my phone in my hand and am not swinging my arms at all, even on my way home I have stuff on my shoulders that would make swinging my arms at all impossible without everything falling off the shoulders. Thank goodness for my One.

Of course it's not accurate exactly down to the last step, cause not everyone is the same when they walk.

I don't have other devices to compare against, so I can't beat the OP in terms of hard data, but thought I'd share that in my month of use of the Force, I have walked swinging my arms, walked without swinging my arms, walked while holding my phone to my ear with the hand with the Force on it, walked outside while clutching a travel mug (or shoving my hands as deep into my pockets as they will go) and not moving my arms at all, and the Force has been pretty darn consistent with tracking my steps accurately. (As you say, not perfect, but I don't think anyone reasonably expects that.) No idea if I magically fit into their exact programming model or what, but I thought I'd share my anecdotal success.

fyoung1111

I'm a fast walker most of the time (5'10" and 34" inseam if it matters) and I've noticed that out of 100 steps the Force usually counts between 60-75 of them. For me I think it's close enough, only because I want a trend compared to my weight loss, meaning the number at the end of the day doesn't matter as long as it is *consistently* off by about the same amount (it is).

Thanks for helping me believe I haven't lost my mind. Philosophically, I agree with you and could care less if it throws a few extra steps my way while I brush my teeth or drive my car. I did object to a 30-40% under-count when I was actually walking though. I just want them to get the message and fix their damn algorithm. This is about having the thing perform its most basic function after all!

My Basis 2014 arrived today. I'm curious to see how it performs. All three trackers get a clean start at midnight. I know I need to get a life but I've gotten obsessed with this :grumble:

fyoung1111

Did you use the activity timer to get that information?

Kimsied,

Thanks so much for that head's up. There is a lot of detail on this subject at https://help.fitbit.com/customer/portal/articles/413311-how-do-i-log-or-record-an-activity-

In a nutshell, you carve out a period of "activity" the same you you indicate a period of sleep. If you are not sleeping, you can apparently go back and tell Fitbit what you were doing during this period that was not walking or running. Would have come in handy when I spent 5 1/2 hours circumnavigating Alameda island in a kayak last weekend. I let Runkeeper handle that because it was outdoors but, if I had been on a rowing machine in the gym, this would have worked much better. Of course, if I had spent 330 minutes on a rowing machine in the gym, I should be committed. Anyway, I hope you get the idea.

Thanks again Kimsied. I missed this capability altogether. I bet a lot of other people did too.

Frank

fyoung1111

UPDATE

Its been three days since my "order conformation" for my warranty replacement Force and nothing has shipped yet. Don't understand the delay. The Force shortage in the marketplace seems to have ended. Maybe they have suspended shipments while they get to the bottom of the wrist rash problem?

There is a new dog in the step counting accuracy contest. Shockingly, its my shiny new $200 Basis B1 2014 Carbon Steel. Here is the data from a pair of measured three mile walks. The only difference between the first and second was a slight speed difference and whether I had my hands in my front hoodie pocket or not.

Walk One--3 miles--Average Speed 3.65 mph--Hands in Pocket
Steps
Basis 3,291
Force 5,529
One 5,597
RTP 5,550

Walk Two--3 miles--Average Speed 3.59 mph--Hands Swinging at Side
Steps
Basis 4,417
Force 5,555
One 5,764
RTP 5,727

RTP is the free version of the Runtastic Pedometer which allegedly feeds off the M7 coprocessor in my iPhone 5S.

Three takeaways:

1. The Force performs admirably at speeds just UNDER 4 MPH.

2. The Force works equally well with hands swinging or pocketed but pocketed hands befuddle the Basis.

3. The Basis does an overall poor job of counting steps. From inspecting the data, it thought I had stopped walking altogether several times during both walks which were, in fact, uninterrupted.

The weather has gone completely to Hell here in the Bay area so further experiments are suspended until Monday

Kimsied

There have been some shipping delays throughout the country. I don't know about the Fitbit shipment but someone on the opposite coast to me send a package Fedex (I believe overnight--they shipped it Wednesday and it was scheduled to arrive Thursday). Well it hasn't, they keep changing the delivery date and currently they have no date estimate for delivery. I wonder if there is an issue with whatever delivery service they use. Fedex says it is severely disrupted throughout the country, for example. On one hand, I would expect them to investigate the rasgh issue--but it sounds like it doesn't affect everyone. I think they would announce if they were doing a product recall and would offer a different device. If you are wanting a Force after having a defective one it is unlikely the rash issue is affecting you. I suspect it may be related to the weather even if fine in your area.

fyoung1111

Thanks Kimsied. But, according to their order status widget, they haven't shipped it yet. Also, Fitbit and I are are in adjacent counties so it doesn't have to go so far.

You are right. No wrist rash for me. That whole thing is really very odd. Never seems to have been a problem with the Flex and it looks like the same material to me. Maybe they "improved" it to make it more durable since the Force band is not user replaceable.

MountainMaggie

You can play around with the stride length. Mine wasn't accurate for the stride length measured so I adjusted it a couple of times till it became more accurate. If it's underestimating, shorten stride length, if overestimating, longer stride length (I think that's right).

I have an unusual job that involves knowing how to pace out certain distances, and I've learned that my pace when working is a lot different than the pace I measure for the sake of measuring, and I suppose this is true for actually working out versus measuring as well.

Hope this helps if it's not already resolved.

TArnold2012

Anyone have problems with step count yesterday? Mine didn't count, I tried updating but nothing helped.

fyoung1111

You can play around with the stride length. Mine wasn't accurate for the stride length measured so I adjusted it a couple of times till it became more accurate. If it's underestimating, shorten stride length, if overestimating, longer stride length (I think that's right).

I have an unusual job that involves knowing how to pace out certain distances, and I've learned that my pace when working is a lot different than the pace I measure for the sake of measuring, and I suppose this is true for actually working out versus measuring as well.

Hope this helps if it's not already resolved.

Maggie,

It's the number of steps counted that I have a problem with and nothing I have read or been told leads me to believe that correctly setting the stride length will improve basic step counting accuracy. It will make your distance estimates more accurate though presuming your tracker is counting all of your steps.

As someone who has become a little obsessed with this subject, I have read some "scholarly research" on the subject. One of the things I learned from that is that, in most of us, stride length (scientifically defined as two steps) varies dramatically with walking speed. The thing is we speed up by combining faster steps with longer steps. The upshot of that is, unless you are always walking at the same speed, your stride length will always be wrong

I got on a treadmill yesterday and did another of my inane experiments to corroborate the "scholarly" data and see what my own values were at a variety of speeds. This is what I got:

SPEED (MPH) STEP SIZE (Inches)
1 17.79887006
2 25.24867118
3 29.70058191
4 34.91068058
5 33.77429365
6 39.43703704
7 44.17774617

The reason step size decreased when I moved from 4-5 mph is that I transitioned from walking to running (frequency or # of steps per minute went from 121 to 156.) Since you can't actually measure stride length on a treadmill, I recorded steps per minute at the various speeds and backed into the step length with simple algebra. The values here are averages from five or more observations at each speed.

I know. I'm sick :noway:

fyoung1111

Anyone have problems with step count yesterday? Mine didn't count, I tried updating but nothing helped.

I was running a One and a Force yesterday (Sunday) and other than an occasional "try again later" kind of message from Fitbit, I had no problems. I get those messages pretty frequently.

Is this what you mean or were you unable to get it to sync with your phone? For reasons I don't understand, Fitbit won't synch wit your phone (or computer for that matter) unless it is connected to the Internet when you try to synch. Seems like it should just hold the data for later but it doesn't work that way.

Kimsied

It's the number of steps counted that I have a problem with and nothing I have read or been told leads me to believe that correctly setting the stride length will improve basic step counting accuracy. It will make your distance estimates more accurate though presuming your tracker is counting all of your steps.

As someone who has become a little obsessed with this subject, I have read some "scholarly research" on the subject. One of the things I learned from that is that, in most of us, stride length (scientifically defined as two steps) varies dramatically with walking speed. The thing is we speed up by combining faster steps with longer steps. The upshot of that is, unless you are always walking at the same speed, your stride length will always be wrong

Mostly agree. Stride length settings shouldn't effect whether the fitbit is actually counting your steps. It can improve the accuracy of the pace and distance stats--even if fitbit is missing some of your steps (assuming there is some consistency in the pattern of how it misses them). If you use the Fitbit counted steps in your stride calculation than that helps as it really just needs to know what distance to provide per fitbit counted step. Whether this helps or not, kind of depends on your goals. Most people (I think) want a fitbit to track their activity throughout the day and give them an educated calorie burn based on their actual acitivity. For this goal, the distance isn't really that important the important thing is whether the fitbit is picking up enough of their activity (but not too much). If your goal is to have a way to estimate distance when walking, then calibrating the stride can help even if the Fitbit misses some of your steps.

I agree the stride varies by speed, and route and what we are wearing, etc. I guess I would just calibrate for the "average" in your life. I walk briskly outside most of the time, and calibrated for a walkign pace that felt natural outsde. I am sure my fitbit distance for steps inside is usually overstated as my inside stride is likely shorter. I don't worry too much about the accuracy, but it is usually pretty accurate when walking outside since I calibrated both my stride settings. We really have more than two stride variations so I think if it is usualy correct or close than that is good. Serios distance athletics enthusiasts likely have other ways to track their workout distances and there is always the option to log the activty.

At one point I tested my strides on a treadmill in part because I find ti easier to run on a treadmill. The result ended up underestimating my distances. I think my stride was shorter on a treadmill in part because the machine controls it and maybe I was taking smaller steps to avoid falling off the thing. I am sure it is different for regular treadmill users, but am not sure how well the treadmill translates to outside activity--it didn't translate well for me anyway.

fyoung1111

[/quote]
I think my stride was shorter on a treadmill in part because the machine controls it and maybe I was taking smaller steps to avoid falling off the thing.
[/quote]

You might be right but it was the only way I could create a constant speed environment. My 4 mph step length on the treadmill (34.91") dovetails nicely with values of 34.72" and 34.75" I got from the Basis and One respectively on a 3.96 mph walk home from the gym this morning. The Force, in this impromptu experiment gave an a totally ABSURD 54.28" step length. Translation: the Force under-counted my steps by a whopping 36%.

I'm 99.9% sure this is a problem with the Fitbit algorithm. It ONLY happens at walking speed in the very near vicinity of 4 mph. I want them to fix it. It's a pretty HUGE error. Credit for discovering the glitch would make me happy too

If they would ship me my damn replacement Force, I would be able to prove my theory in a day but there seems to be a problem with that. Five days after a promised replacement, it still hasn't shipped.

FINALLY, and almost certainly most important. I realize that I have created a tempest in a teacup over this issue. Is any activity monitor really accurate? Absolutely not. I know because I have bench-marked the five most popular against each other and they all have serious drawbacks. Should you wear one anyway. Absolutely! Knowing that little spy is riding on my waist or my wrist has absolutely changed the way I live my life. I wish parking lots were bigger so I could park further from the store. In fact, my car rarely goes to the store anymore. Just three miles each way. Opportunity to think and meet some nice dogs along the way!