GPS distance totally wrong on Charge 3

[BertilDator]

Was doing cross country skiing, and set activity to "run" (as there is very limited activity list). Had phone connected, and got perfect GPS route on Fitbit activity. However, all the splits are way off. Total distance in Fitbit shows 4.8km while real distance was 7.5 km (I know the route, and even double checked via ontegomap.com).

How can Fitbit calculate distance so wrongly even though the GPS route is perfect on the activity? I repeat, the tracked route is 100% correct on the Fitbit activity in question, so there is no disconnections, or trees, or anything between the phone and GPS satellite, nor between phone and Charge 3.

Is Fitbit using GPS route for distance calculation when assisted GPS is enabled during "Run" activity, or does it STILL use stride length?

Should I use "Biking" activity for XC skiing to get accurate distance based on the GPS route?

[leftyx]

I did another test run today, this time recorded the run with Strava in parallel. 

FitBit - 3.95km    -    Strava - 4.61km

(Note: Strava even matched the FitBit run and its own recording together ;-))

When I load the FitBit .tcx in GPS track Editor it calculates the distance as 4.54km (the ~2% difference with Strava is probably acceptable), unfortunately I can't load the .TCX into strava as it detects the activity already exists (from FitBit and its own recording). 

There is a 15% difference between FitBit and Strava, granted the apps may calculate distance differently but this is too much of a difference. Also, based on the *same* FitBit data, there is a 13% difference between FitBit and GPS track Editor distance.

This all indicates the FitBit distance calculation is incorrect (too low). 

[larsll]

I did another test today; tracking with Fitbit and Strava at the same time. 3.44km vs. 3.48km.

What I was careful about today was to start the walk BEFORE leaving the house; I live in an area with poor mobile data connectivity, and I have a theory about Fitbit not liking not being connected at the beginning of an activity.

The runs did not get merged in Strava; I got to Strava activities; one saying that I had a run with myself...

[SgtPepper]

Another report to add to the thread. This started happening for me just 3 or 4 days ago. As with most people, my Fitbit is under reporting distance

Step count is accurate, the route is accurate, but the distance reported is about 15%-20% less than I would expect (or that my partner's fitbit Charge 2 reported on the same walk) 

[leftyx]

I compared some FitBit data from previous runs, note this is the *same* person running the *same* route using the *same* FitBit device, so I would expect the results to be quite similar.

Date	distance	time	pace	steps	steps/5km
15.04	5.07	27	5'21	4421	4360
16.04	5.05	26	5'07	4218	4176
20.04	5.03	27	5'20	4349	4323
22.04	5.06	27	5'14	4199	4149
23.04	5.02	29	5'52	4785	4766

As there is a slight variation in the distance, I recalculated the steps per 5km. The runs from 16.04 & 22.04 and 15.04 & 20.04 are similar, and between these 4 runs there is a +/- 2% difference to the average which might be acceptable. 

However the run from 23.04 seems totally off, the difference compared to the average of the other 4 runs is 11%.

As this is a noticeable difference I started investigating, including posting here and raising the issue with FitBit support. So far, FitBit support has been sending me generic information on GPS tracking that doesn't explain this significant difference. 

[SgtPepper]

(seems that my previous attempt at this post was deleted so I'm deleting the helpful hyperlinks I'd added)

I've done a bit more digging to see if I could spot any patterns by exporting the last week's exercise as tcx files and comparing with the distance calculated by GPS Track Editor.  Full details in the table below. There is a small but acceptable delta (+/- 4%) between the two calculated distances until April 22. From April 24 onwards, Fitbit Distances are 38% below the more accurate distance calculated by GPS Track Editor. 

I also looked for the update history of the Android fitbit app. It looks like there was a fairly significant update on 23rd April to v3.19.1. Coincidentally, this was the same update which broke the Exercise tile. 

So, it looks like the update on April 23rd could possibly have caused the distance calculation problems. There has been another update to version 3.19.2 today (April 27) to fix the exercise tile crashing bug. I'll see if this also fixes the GPS distance bug (seems unlikely, given that Fitbit hasn't acknowledged the issue on this thread).

If the same magnitude of error shows up in my exercise again today, I'll try rolling back to v3.18 of the Android app and see if this fixes the issue, which would then be conclusive proof it's an app bug. 

Date	Fitbit Distance (miles)	GPS Track Editor (miles)	Difference (miles)	% delta
19/04/2020	4.14	4.3	0.16	3.720930233
20/04/2020	1.95	1.88	-0.07	-3.723404255
22/04/2020	4.2	4.02	-0.18	-4.47761194
24/04/2020	2	4.73	2.73	57.7167019
25/04/2020	2.54	5.39	2.85	52.87569573
26/04/2020	3.05	4.96	1.91	38.50806452

[Povilas]

My experience is that FitBit app tracks GPS close enough to other apps (50-100m difference from 5km). However, if GPS is used through the device, it ignores GPS data. But it is recording it! You can validate it because the map afterwards contains the correct path.

The device instead uses step counter. It does basic multiplication of step count by your stride length. In my case it's 101cm while in reality it's 121cm. Hence up 150-200 meters wrong per each kilometre i run. Moreover, stride length depends on your pace and is not the right way to measure distance.

Image where GPS says 6.2 but steps overrides with 4.83 

This doesn't happen constantly, most of the time GPS full distance is not even shown, but seeing that it happens even sometimes, confirms that GPS data are being ignored, but are collected correctly.

[tom43614]

Povilas,

I do not use the step counter because I am on a bicycle (GPS through Samsung Galaxy S10+).

The map generated is correct, but the mileage is off by .10 per mile.

[leftyx]

As @SgtPepper indicates, there seems to be a difference in FitBit distance calculation after 22.04, which seems to align to the FitBit version update 3.18 > 3.19.1. It definitely seems something has changed in the FitBit distance calculation. 

Based on previous runs (*same* route):

> before 22.04 difference is 1% - 7% (with FitBit actually reporting a larger distance)

> since 23.04 difference is 15%-17% (with FitBit reporting a smaller distance)

Date	distance	distance	FitBit <> GPS
15.04	5.07	4.85	-4.5
16.04	5.05	4.73	-6.8
20.04	5.03	4.95	-1.6
22.04	5.06	4.72	-7.2
23.04	5.02	6.07	17.3
27.04	5.07	6	15.5

@SgtPepper did you have a look at the "remove local inconsistencies" option of GPS track editor? It seems there is a much larger number of local inconsistencies in the GPS data since 23.04 (number of inconsistencies jumped from 5 - 49 to 244 - 326).

I updated FitBit app to 3.19.2 after my last run (27.04) will have to go for a run tonight to check if that makes any difference. 

@Povilas I'm not seeing a distance between distance for the route (indicated at the top of the screen) and the distance indicated on the map itself (Image where GPS says 6.2 but steps overrides with 4.83 ), as this doesn't seem to happen all of the time, any idea when this occurs? 

[SgtPepper]

Thanks @leftyx - you're right, when I created my table I did not remove local inconsistencies.

I've recalculated the table (below) and I'm still seeing the major change from 24th April, aligning with 3.19.1 update, although now ranging from 17%-25% inaccuracies (still very significant). I've also added my walk from last night using 3.19.2, and I'm still seeing the same range of error.

I'm much happier with the updated numbers - and I've gone back to check previous Walks recorded by Fitbit along the same routes, and the GPS Track Editor calculations (with inconsistencies removed) match the actual distances. 

@leftyx I think you are onto something with the thought about increasing 'local inconsistencies' since 3.19.1. It certainly seems that GPS Track Editor detects and removes a significantly higher number of local inconsistencies since 3.19.1 - sometimes way over 50% of the points are being removed.

I cracked open two of the TCX files and it would appear that the 3.19.1 Android client is writing a similar number of points as 3.18, but a high proportion of them are ones that GPS Track Editor considers to be inconsistent. 

In the TCX files created by the FitBit app, every TrackPoint element contains a DistanceMeters element. Looking at the last TrackPoint element in the routes, the value in meters matches the distance that is displayed in the Fitbit app. Therefore, my conclusion is also that something has changed about the way the fitbit app calculates distance traveled and writes it into the log.

Date	Fitbit Distance (miles)	GPS Track Editor (miles)	GPS Track Editor (inconsistencies removed) - miles	Difference (miles)	% delta	s/w version
19/04/2020	4.14	4.3	3.86	-0.28	-7.253886	3.18
20/04/2020	1.95	1.88	1.88	-0.07	-3.7234043	3.18
22/04/2020	4.2	4.02	3.98	-0.22	-5.5276382	3.18
24/04/2020	2	4.73	2.64	0.64	24.2424242	3.19.1
25/04/2020	2.54	5.39	3.39	0.85	25.0737463	3.19.1
26/04/2020	3.05	4.96	3.71	0.66	17.7897574	3.19.1
27/04/2020	1.21	2.25	1.52	0.31	20.3947368	3.19.2

[Hodgo93]

So glad I've found this forum! Me and my partner both had charge 2s back in January 2019 when we first starting running and had no issues at all. We stopped in the summer and started again about Nov 2019. My partner upgraded to a charge 3 in January as her battery wasn't lasting more than a day. I had a second hand Fitbit 2 with a cracked screen. All of a sudden, bear in my we have the same phone and we're with the same network, my distances started dropping. I just assumed it was because my charge 2 was probably about 3 and a half years old at this point and had a cracked screen so the other week I upgraded to the charge 3 as well. No change. My distance for the same run (if anything I run further) is half a mile less than my partner's. We know hers is accurate as it's the same distance on Google maps. After reading all of these comments on here I think I've lost faith in Fitbit and will be returning it and going to Garmin!

[Procter101]

If I could get fitbit to actually respect UK law I would be overwhelmed
with surprise!!!

[leftyx]

Did some further research and loaded the .tcx files into Excel to calculate the distance between the latitude/longitude points measured by FitBit manually, using formula:

=ACOS(COS(RADIANS(90-Lat1)) *COS(RADIANS(90-Lat2)) +SIN(RADIANS(90-Lat1)) *SIN(RADIANS(90-Lat2)) *CO...

Adding up the values between each trackpoint matches the distance calculation in GPS Track Editor exactly, which validates the distance calculation (based on FitBit data). 

To compare the data in a graph, I plotted DistanceMeters straight from FitBit data, along with a calculated delta between each trackpoint - shown in blue. To compare against the GPS data added distance and delta - shown in yellow. 

Image: https://ibb.co/DbHcwZj

For the activity on 22.04 (FitBit app version 3.18)

> similar result in distance calculation.

> some variation in the calculation for both FitBit distance and GPS distance (both lines are not exactly straight).

> FitBit trackpoints "follow" the GPS trackpoint trend, which is consistent with FitBit dynamically calculating stride length. 

For the activity on 27.04 (FitBit app version 3.19.1)  

> large difference in distance calculation.

> GPS delta distance between trackpoints very variable, GPS distance calculation shows jagged line.

> FitBit distance calculation shows a very straight line.

For the activity on 28.04 (FitBit app version 3.19.2)  

> similar result in distance calculation.

> GPS distance calculation shows some variation.

> FitBit distance calculation shows a very straight line.

> FitBit trackpoints "follow" the GPS trackpoint trend, which is consistent with FitBit dynamically calculating stride length. 

Based on the above, there seems to be an issue with variability in trackpoints logged by FitBit with app version 3.19.1, but I don't see the same trend as @SgtPepper with app version 3.19.2...   

In any case, this indicates FitBit uses stride length as the base for distance calculation, rather than the actual GPS data, which explains the differences in distance calculation. This is *not* what's indicated at https://help.fitbit.com/articles/en_US/Help_article/1141, and produces a less accurate result. Stride length should be used when GPS data is not available, instead of the primary measurement (even if that is automatically calculated). 

[leftyx]

(previous post was deleted, reporting without external links)

Did some further research and loaded the .tcx files into Excel to calculate the distance between the latitude/longitude points measured by FitBit manually, using formula:

=ACOS(COS(RADIANS(90-Lat1)) *COS(RADIANS(90-Lat2)) +SIN(RADIANS(90-Lat1)) *SIN(RADIANS(90-Lat2)) *COS(RADIANS(Long1-Long2))) *6371

Adding up the values between each trackpoint matches the distance calculation in GPS Track Editor exactly, which validates the distance calculation (based on FitBit data). 

To compare the data in a graph, I plotted DistanceMeters straight from FitBit data, along with a calculated delta between each trackpoint - shown in blue. To compare against the GPS data added distance and delta - shown in yellow. 

 

For the activity on 22.04 (FitBit app version 3.18)

> similar result in distance calculation.

> some variation in the calculation for both FitBit distance and GPS distance (both lines are not exactly straight).

> FitBit trackpoints "follow" the GPS trackpoint trend, which is consistent with FitBit dynamically calculating stride length. 

For the activity on 27.04 (FitBit app version 3.19.1)  

> large difference in distance calculation.

> GPS delta distance between trackpoints very variable, GPS distance calculation shows jagged line.

> FitBit distance calculation shows a very straight line.

For the activity on 28.04 (FitBit app version 3.19.2)  

> similar result in distance calculation.

> GPS distance calculation shows some variation.

> FitBit distance calculation shows a very straight line.

> FitBit trackpoints "follow" the GPS trackpoint trend, which is consistent with FitBit dynamically calculating stride length. 

Based on the above, there seems to be an issue with variability in trackpoints logged by FitBit with app version 3.19.1, but I don't see the same trend as @SgtPepper with app version 3.19.2...   

This indicates FitBit uses stride length as the base for distance calculation, rather than the actual GPS data, which explains the differences in distance calculation. This is *not* what's indicated at https://help.fitbit.com/articles/en_US/Help_article/1141. Stride length should (only) be used when GPS data is not available, instead of the primary measurement (even if its automatically calculated). 

[SgtPepper]

Awesome work @leftyx! It's possible that there was user error in my walk last night (e.g. perhaps the app hadn't properly restarted after I updated from the Play Store?) so I will try again on 3.19.2 to check again (if it stops raining!)

Any app which uses GPS for distance calculations will end up needing to use some sort of filtering/ smoothing algorithm similar to GPS Track Editor: inaccuracies are inherent in GPS readings and therefore if the sample rate is high enough you will end up with a long distance being travelled over a very jagged line. Is it possible that FitBit have been tweaking their smoothing algorithm to cause these differences?

[SgtPepper]

Another update from me after going for a walk this evening. I'm much more confident that I was using 3.19.2 after a reboot. I'm still seeing FitBit underreporting the distance walked, but this time by 7.6% rather than the ~20% I had seen on 3.19.1.

So, it's looking better but I don't think it's conclusive and I'll keep tracking.

Date	Fitbit Distance (miles)	GPS Track Editor (miles)	GPS Track Editor (inconsistencies removed) - miles	Difference (miles)	% delta	s/w version
19/04/2020	4.14	4.3	3.86	-0.28	-7.25	3.18
20/04/2020	1.95	1.88	1.88	-0.07	-3.72	3.18
22/04/2020	4.2	4.02	3.98	-0.22	-5.53	3.18
24/04/2020	2	4.73	2.64	0.64	24.24	3.19.1
25/04/2020	2.54	5.39	3.39	0.85	25.07	3.19.1
26/04/2020	3.05	4.96	3.71	0.66	17.79	3.19.1
27/04/2020	1.21	2.25	1.52	0.31	20.39	3.19.2(?)
28/04/2020	1.95	2.88	2.11	0.16	7.58	3.19.2

[Povilas]

>@Povilas I'm not seeing a distance between distance for the route (indicated at the top of the screen) >and the distance indicated on the map itself (Image where GPS says 6.2 but steps overrides with 4.83 ), >as this doesn't seem to happen all of the time, any idea when this occurs? 

@leftyx You see that map picture ends with 4.83km? There's no 5th or 6th kilometer! Howerver, if you look at top left corner, it says 6.2km - which is the correct distance.

The issue that all the performance data and final length is considered to be 4.83, making my runs slow and of too short distance.

[leftyx]

@Povilas thanks for the clarification. I can see the difference in your screenshot, but I don't see a difference in the distance for the route (top left) and as indicated on the screen for any of my activities, so so it seems this issue doesn't occur for me. This is also why I asked if you have any idea when this occurs, so I can look at past runs a bit more specifically? 

@SgtPepper thanks for all your responses, I'm think its still a inconclusive and there are questions about the differences between the distance measurements with the different FitBit app versions. I'll keep tracking as well.. 

[Povilas]

@leftyx 

>This is also why I asked if you have any idea when this occurs, so I can look at past runs a bit more specifically? 

I tracked my runs in total probably only 5 times. Because I just end up deleting them because they mess up Cardio fitness and are simply wrong. Now, I track without GPS, so I get my heart data, and distance doesn't mess up anything. And I use another app to keep track of my pace.
This situation happened only 2 times where correct distance is visible. Sadly, I don't know how to reproduce it, or why it happened.
-------------------------------------
Today I got a reply from FitBit support. They said nothing useful, except how sorry they are. Didn't reply to my questions, didn't fulfil my request of them replying to this thread.

One more weird thing they said was that my phone is not supported (Samsung A8). Moreover, they gave me a link to check the list of supported phones, so, I've sent them back a screenshot where I marked my phone with red in their list 😄

They are totally unhelpful.

[leftyx]

@Povilas I don't see the difference in the map screen for any of my runs.

I've had a bit of back-and-forth with FitBit support, but unfortunately no solution as of yet. 

Another two days of data to add to the stats (@SgtPepper I noticed you've updated the stats as well) and I now also see a large (!) variation with the latest app version. In between the runs on 29.04 and 30.04 I have reinstalled the app and unpaired/paired my FitBit device, as suggested by support. However this didn't seem to have an effect (if anything, it seemed to have made things worse.. 

FitBit		GPS track editor	- local inconsistencies
Date	distance	distance	# of incons.	distance	FitBit <> GPS	FitBit <> GPS
15.04	5.07	4.85	12	4.85	-4.5	-4.5
16.04	5.05	4.73	49	4.67	-6.8	-8.1
20.04	5.03	4.95	6	4.95	-1.6	-1.6
22.04	5.06	4.72	5	4.72	-7.2	-7.2
23.04	5.02	6.07	244	5.8	17.3	13.4
27.04	5.07	6	326	5.61	15.5	9.6
28.04	5.14	5	4	5	-2.8	-2.8
29.04	5.11	4.7	2	4.7	-8.7	-8.7
30.04	5.02	6.49	383	6	22.7	16.3

What's interesting is that usually FitBit reports a slightly larger distance than GPS Track Editor calculates, and there are only a few (2-49) inconsistencies in the GPS data. However in some cases (23.04, 27.04 and 30.04) FitBit reports a much lower distance than GPS Track Editor, and there are many (244-383) inconsistencies in the GPS data. 

Not sure what causes this, but I'm trying to be as consistent as possible - doing the same run and starting the FitBit activity from the activity tab on the pebble itself. The difference with yesterday (29.04) is that at the point (=petrolstation) where yesterday's run finished, today (30.04) FitBit only recorded 4.10km, almost a km more to go!

[Procter101]

Completely agree just get meaningless platitudinous replies.
Never again