As said , you have to ensure that your raw data is as stable as possible...for me the 3D print was the turning point. I had tried to calculate the average across 10 readings (lihe @craigcurtin ) and you still see the function node for this ... but after using the 3d print it was not necessary at all.
good luck