Numbers may give precision to all kinds of analyses, and also to the analyses of political phenomena. However, political numbers, that is the counting of political phenomena, must be subordinated to well chosen criteria. Numbers alone are never enough. They must be combined with criteria of relevance. In this article, the author makes two examples. The first one is the well-known, but still very important case of the classification of party systems. As Sartori has shown, the simple counting of the number of parties is likely to be totally misleading. The other example refers to the analysis of authoritarian and totalitarian regimes. The definition of authoritarian regimes by Linz is based not just on the number of groups (limited pluralism), but on the fact that authoritarian pluralism is non-competitive and non-responsible. In the same vein, one cannot say that the existence of several groups and associations in the Soviet Union suggested the appearance of pluralism. Associations licensed by the Communist Party were never autonomous. Hence, just counting their number meant fully to misunderstand the evolution of the Soviet Union and to confuse the process of differentiation through modernization with political decompression leading to an opening of the regime (that, in fact, never took place). The overall lesson is that theorizing must always precede counting (and quantification).