Why not use big data to correct obvious glitches with the GPS recording from the devices?
Example I: I ran a path that many other Garmin users had ran before me. When I ran it, one or two GPS recordings suddenly jumped several hundred meters off course and then back again. This causing several of the meta data to be completely wrong (i.e. max speed, length of activity). Why not use the data from all the other Garmin users to correct the obvious errors in the recording?
Example II: Same as the other, just that here my Garmin device suddenly decided that I was running at 20 000 meters above the sea level. If the x thousand Garmin users before me have reported the path to be at, for argument sake, 255 meters above sea level, then I would be very surprised if that still was not the case.