I avoid using the Google Mercator CRS (EPSG:900913, in QGIS Master) in QGIS projects because I know that it has problems. For example, users get into trouble trying to measuring distances using the tape measure. But people do use Google Mercator for projects and a recent post here suggest that they also do things like calculating the areas of polygons from shapefiles projected to EPSG:900913. This looks like a dangerous thing to do. Recently I created a test shapefile containg one object, a polygon, square and having sides 100 km in length. This corresponded with the extents of the British Ordnance Survey's 100 km grid square 'SU'. The shapefile was loaded into QGIS and the layer CRS and the project CRS were set to EPSG:27700. I then used the field calculator to calculate the area of the polygon, which it did correctly (10,000 sq km). I then reprojected the shapefile to EPSG:900913 and used the field calculator to calculate the area of the polygon. The area of the 100 km x 100 km square was now reported as being *25502 sq km*. One other oddity came to light too. As mentioned, I loaded the shapefile projected to EPSG:27700 into QGIS, set the layer CRS and the project CRS to EPSG:27700. The 'Identify Features' tool then correctly reported the area of the polygon correctly (10,000 sq km). Enabling OTF and checking again with the 'Identify Features' tool then produced an area of 9,980 sq km. But why? I have observed a similar problem with Google mercator and vector files. I have a google satellite map, and several point vector files all in google mercator transformation, but the relationship between the locations on the map and the locations of the points of the vector files are not stable. If I open print composer or change the zoom I find the points have moved relative to the map, so that my points are not longer in the correct places, although I don't know whether the points or the map have moved (or both?). The same thing happens with vector line files. This happens whether or not I have 'on the fly transformation' checked and if I check 'on the fly transformation' and then use vector files in another transformation (e.g. EPSG4326) the same problem occurs. I wonder if there is an issue with the stability of the transformation and if this is affecting both my point locations and the polygon areas referred to by Nick. Source.