The existing schemes to express the evolution of radio sources are either too simple to fit all the data (multi-frequency counts, identifications, spectra etc.) or have too many parameters and little predictive value. They do not also fit the identification statistics at different flux-densities (Swarup, Subrahmanya & Kapahi 1982, in ‘Astrophysical Cosmology’, Proc. Vatican Study Week) and the counts from recent deep surveys (van der Laan 1982, in ‘Astrophysical Cosmology’). Here, we report a preliminary attempt to derive a model that retains the simplicity of a small number of parameters and ease of refinement when new data are added.