Wherever possible, aggregations should be done in the source system. This is likely to reduce the size of the import file meaning faster imports
