I have measured a few track distances with gps on an Android app and with a measuring wheel for comparison. The GPS tracks were kalman filtered and then the error between KF data and MW compared. The gps track usually shows an error not larger than 5% but sometimes, i have outlier tracks with 15-18% distance error. I tried to analyze the raw gpx track compared to the KF one but can't find anything suspicious. Went through elevation, which was smoothed down a lot compared to the mess of the raw data (but for all of it). What else could the source be and how could i determine which track is errorous without comparing to the MW? I'm working with python dataframes etc.