Few mathematical methods attracted theoretical and applied researches, both in the scientific and humanist fields, as the Proper Orthogonal Decomposition (POD) made throughout the last century. However, most of these fields often developed POD in autonomous ways and with different names, discovering more and more times what other scholars already knew in different sectors. This situation originated a broad band of methods and applications, whose collation requires working out a comprehensive viewpoint on the representation problem for random quantities. Based on these premises, this paper provides and discusses the theoretical foundations of POD in a homogeneous framework, emphasising the link between its general position and formulation and its prevalent use in wind engineering. Referring to this framework, some applications recently developed at the University of Genoa are shown and revised. General remarks and some prospects are finally drawn.