The information produced by the Farm Accounting Data Network (FADN) is used to formulate common agricultural policy decisions. They, therefore, must be correct and present a clear and coherent picture of the primary sector. The representativeness of the sample supporting the survey, however, can be invalided by some important factors. The first of these is the presence in the sample of farms not subject of investigation and the second is the different distribution between the farms in the sample and the ones in the field of observation. The developed analyses showed that both factors influence the FADN sample. The first factor, more precisely, is present in the sample of all Member States and derives from the positioning of the entry threshold to the FADN at levels of economic size too low to allow only commercial farms to enter the survey. The second factor, analysed only for Italy, arises from the replacement of non-responding farms with farms that have economic size and/or type of farming different from those of the fallen farms. The presence of the first factor is believed to depend on a "precautionary" behaviour of the Member States. The presence of the second factor, instead, derives from the difficulty of the Italian FADN to manage the considerable lack of response resulting from the adoption of a random sample.