Non-specialists might have the feeling that building statistics on businesses is a very simple task: it seems one just has to "add facts". But for survey statisticians, business statistics are extremely complex: great heterogeneity of the universe, definition of the statistical units, difficulty of classifying businesses, quality of the register, variety of accounting standards, sample co-ordination, or reduction of survey burden. This paper argues that the issues raised by business survey methodology are conceptual and not only practical. It describes different aspects of statistical processing and tries to analyse the special features of business statistics.