Simini, Gonzalez, Maritan et Barabasi recently proposed a "universal model for mo- biliy and migration patterns": the radiation model. They compare the radiation model to the gravity model and conclude that the latter suffers important theoretical inconsistencies and systematic predictive discrepancies. By confronting both models, they address crucial issues about formalisms and usages of spatial interaction models: theoretical and methodological is- sues about the quality of a predictive model, but also issues about the success of a modelling approach, that is, its diffusion and stability. This paper aims to highlight the link between both aspects: the quality of the model, which at first sight relies on its formalism, and the succes of the model, which relies on its usage.