Sense built in GPS is rubbish - Dynamic GPS like Charge 4 is needed

[SunsetRunner]

I was always a bit suspect of the built in GPS on the Sense. Firstly, routes that I had ran for months and recognised as 5km by Garmin and Strava were now coming up shorter and my splits were all slower.

So I ran a route with an Inspire 2 connected to an iPhone, the Strava app on a Samsung S20 and the Sense on my left wrist using built in GPS.

Inspire 2 - 5.17km

Strava App - 5.22km

Fitbit Sense - 4.96km

I then drew the route really quickly on the Strava browser app and it came to 5.2km

The sense is at least 200 metres wrong on a 5k run which would easily be 1k over a long run. Add in the weird heart rate issues etc and I am definitely ditching this device.

Forget their smartwatch range for now outside the Versa 2 and stick to the trackers

[t.parker]

@SunsetRunner export TCX file and upload it http://mygpsfiles/app . You will be probably surprised that the distance of the GPS route and the distance reported by Fitbit is different. The reason is that the Fitbit algorithm uses steps and stride length also for the GPS-based runs. It's not bad GPS, it's Fitbit implementation of the distance calculation (it's been like that for years). Because the stride may change during the run, sometimes you will get a better result, sometimes worse, and sometimes horrible. Connected GPS won't solve the issue (my old Charge 2 suffers the same problem, then Ionic with built-in GPS, and now Sense, too). This won't get fixed as it's not broken, it's how it has been designed to work.

[SunsetRunner]

Hi, I will try that is it makes sense. Only thing though is that the Inspire 2 would also use the same algorithm but that seemed to come in fine 😞

[N8teGee]

I can agree with everything @t.parker has posted above. I would add that sometimes the distance is close, but it's more down to you having a consistent stride length for that particular period/activity. You can put it to the test yourself. Simply get into your car and start a GPS exercise, now drive for a short period. You'll notice the route is tracked on the map, but your distance will be close to zero. You'll also notice that bike tracking uses only GPS for distance as the step data isn't usable in this situation (so it can be done properly).

The last Fitbit tracker to use a traditional method of tracking GPS distance was the Surge, that's right a device from 6 years ago.

Many of us have been pushing Fitbit for an explanation as to why distance is calculated this way and if they intend to fix this. They even denied this was even the case until a couple of months ago. 

[t.parker]

@N8teGee will it surprise you that Garmin uses a similar (but not the same) approach? 🙂 They also use steps and stride. I did more data analysis and tests and came to the conclusion that Garmin's accuracy depends on refining the stride length and distance in real-time when we track the exercise. It also takes into account any moments we lose connection or the accuracy of GPS drops and uses elevation data (like Strava, when computes pace). So the stride length is being refined all the time during a run so the GPS and distance will get very close. They are not 100% equal to the GPS route distance but, as expected, are very close (and if we are unhappy, we can edit any run parameter later). This is supposed to compensate for GPS drops and inaccuracies and this method, when done correctly, actually works (comparing to the Stryd sensor stride-step-GPS method seems to be working quite well). My educated guess tells me that Fitbit does stride length adjustment AFTER the run in the app (or maybe server-side) when the GPS route is ready. It means that one run may go horribly wrong but the next run, if you make it "a twin" of the previous run will be very accurate (I did such a test several times to confirm it, I could deliberately make a good run or trash it). Lack of real-time refining will bring errors when the stride length changes during a single run. Garmin takes it into account. Looking at the Sense GPS output, the only-GPS metrics would be a nice thing. I find Sense GPS working surprisingly well (and sometimes it beats Garmin Fenix 6 Pro). Fitbit should implement options, give users a choice with specifying downsides of each method. That would reduce the number of complaints.

[N8teGee]

I believe this is the way Fitbit is supposed to do it too, at least that is what we were told at some point (I've seen/heard so many different things now it's difficult to remember). As you say though, something is definitely wrong in their implementation. You could be right about the processing being done post work out, you seem to have done some extensive investigating (compared to my lack of). So I'm happy to accept your knowledge of this will be greater than mine 😁

It's just a shame that the same few people are the only ones bringing this up, perhaps this is the reason we aren't getting anywhere with Fitbit (we are easy to ignore). 

[SunsetRunner]

I will redo the trial tomorrow or tonight and see if changing my running stride manually on both the Sense and the Inspire 2 makes a difference.

This image from the site posted above shows that the Sense GPS does lose track quite a bit. Blue line is the Inspire 2 connected to an iPhone and the Orange is the Sense which is very wayward and covers areas I did not even touch

[CBV2]

This is all very interesting, in an academic sense. I am not particularly concerned with absolute measures of distance, as you other posters here seem to be, and only activated the GPS stuff to get a map of my walking routes in the park.  Which has then lead me to look at ways of extending my walks in the park to increase the number of steps I might do. You know, nothing better to do with my time.  But having said all that I have been a little intrigued by the results you quote, IconDaCarver, and t.parker and N8teGee, and have thought to look at mine.

Firstly, though, let me say that a cursory scan through various sites dealing with GPS do make interesting points about its accuracy. Which essentially say that at the domestic level, i.e. non military or aviation or the like, GPS is not very accurate. Fellrnr.com has a very good paper on his site GPS Accuracy of Garmin, Polar, and other Running Watches which is worth a read.  As to my own experiences I recently changed from a Versa2 to a Sense. I started looking at the GPS results 9 months or so ago with the Versa2, linked to my iPhone, and my usual walk typically came in at 3.6km, plus or minus 0.2km.  This morning, using the Sense GPS, I used a slightly longer route which normally comes in at 3.61km, once again, plus or minus up to 200m .  However, in view of this discussion I resurrected MapMyWalk on my iPhone, which I have not used since pre Fitbit days and it gave a distance of 3.53km.  Measuring the distance on Google Maps gave 3.54km, closer to the MapMyWalk distance than the Sense distance.   

My conclusion is that the iPhone/MapMyWalk combination is probably roughly equivalent to the Versa2/iPhoneGPS combination and both are more accurate than the Sense.  But the differences are not enough to get me fussed. Interesting issue, though.

And reading through the posts again, t.parker you say " sometimes you will get a better result, sometimes worse, and sometimes horrible".  True, true. While I said above that my results were usually plus or minus 200m, pretty consistently, looking back over the months I see I had one result at 4km, I have no idea why.  

[t.parker]

@CBV2 the accuracy of the distance is very often a major reason why people blame the GPS, hence the focus on it as it's been shown here many times that distance of GPS route and distance given by Fitbit are two different things. For me, when it comes to the map, the accuracy is good enough. There will be always some errors but I can see where I was running. The distance however is important because of the metrics. Running parameters are all connected (pace/speed/cadence/distance) and if one is incorrect then it suggests that others will be, too (well, there is no speed and cadence on Fitbit but pace and distance are related).

One thing I am missing on every watch is a feature of snapping the GPS route. Something like I can do using Google Maps when I drove without navigation but the Google Maps can figure out the way and snap my drive to it making the timeline more accurate. I haven't seen such a feature on any watch and that would be incredibly useful as it would allow correcting the map after the activity (and possibly recalculate parameters). I know there is such a feature working during the route creation phase on some watches (and Strava, and Komoot, etc.) but I'd rather have my run tracked, press "snap to known routes" and all inaccuracies, drifts, etc. would be cleared. Well, I will buy the flagship model from the first company introducing such a feature (even if I don't like the watch :D).

[CBV2]

t.parker, going back to the earlier posts regarding the way Fitbit and others "correct" trip (walking in my case, running in yours) results, presumably because of the presumed errors in GPS, and my comment about the "sometimes horrible" results I have had, I think I can now see where they are coming from.  As an example, if I am walking from A to B, a distance of 4km, and I usually cover this distance in say 5,300 steps, with my usual (quite short) stride length, then Fitbit arrives at a distance result of e.g. 4.05km, or maybe 3.95km, whatever.  But if I am walking in an abstracted fashion, with a shorter stride length than usual, resulting in more steps than usual, then Fitbit calculates the distance using my normal stride length and the actual steps, arriving at a distance of e.g. 4.2km.  This seems to be what has happened in the two horrible result cases that I have, where the number of steps is noticeable more than what I normally have for the walk.  As you say, it would be a lot cleaner if Fitbit simply calculated the trip using GPS metrics, or at least gave us the option of using that approach.

And, going back again to your earlier post, and to your reference to http://mygpsfiles/app , I have extracted and downloaded a TCX file for one walk and Fitbit gave a distance of 3.8km while mygpsfilesapp give a distance of 3.6km.  However, I have walked that route before and the Fitbit distance for one such walk was 3.49km.  How interesting. Thanks for that reference. 

[M4M]

I didn't know that. Thanks N8teGee. I wanted to update my Surge and tried a Charge 5. It was always losing GPS, and showed me running through peoples yard. Also, it didn't calculate my pace and distance correctly. I promptly returned it. I just don't understand why Google (now that it owns Fitbit) would sell a product like that. I'll keep my Surge for now, it's consistently accurate.