Dear Marcel,
thanks for your question!
Obviously, the generation of gridded data from satellite observations is not at all trivial and there are multiple aspects related to this.
One is the satellite retrieval itself, meaning the estimation of a certain parameter (e.g., precipitation, solar radiation) from the satellite observation. This retrieval is typically done on the spatial and temporal resolution of the satellite input data and, hence, depends on the satellite (e.g., polar orbiting or geostationary) and the satellite instrument. Based on the, what we call, instantaneous retrieval results, the gridded data are then generated in the second step.
The CM SAF gridded data include temporal as well as spatial averaging of the instantaneous retrieval results. The number of satellite observations that are used for every data point of the gridded data depends on the temporal averaging (e.g., daily, monthly) and the spatial resolution of the satellite instrument and the selected grid.
The way the gridded data is derived from the instantaneous satellite data is (at least should be) described in the corresponding ATBD; for most data sets, the number of observations used to generate the gridded data is available in the data file.
Hope this helps to answer your question. Please let me know if you have any particular data set in mind for which you would like to have more specific information?
Thanks, kind regards,
Jörg