Modelling is a pertinent approach: (1) to better understand and to predict pathogen spread in host populations according to the biological system characteristics under various management scenarios; and (2) to evaluate the epidemiological and the economic effectiveness of control strategies. To end with useful models, a back-and-forth between models and biological data is needed. First, building epidemiological models consists in proposing from all of the up-to-date available knowledge an integrated conceptual view of the system. This highlights which processes are well known vs. which are still of the biological assumption type. Second, observed data can be used to estimate observable parameters (such as disease-related mortality rates and production losses), whereas epidemiological models can be used to estimate unobservable parameters (such as transmission rates). Sensitivity analysis is a powerful tool to identify parameters with a major influence on model outputs, these parameters need to be precisely informed. Third, data can be used to evaluate / validate models, which in turn can help to identify potential control points of the biological system, to compare scenarios and test biological assumptions, and even (when the model has been evaluated) to predict future states of the system according to past (known) states. We illustrate such interactions between observations and models in the context of livestock infectious disease spread and control, with examples as Q fever, paratuberculosis, bovine viral diarrhea in cattle, and Salmonella carriage and the PRRS in pigs. Focus is made on the multi-scale modeling (from the within-host immune response to the infection dynamics at a regional scale), and the coupling of epidemiological and economic models to account for farmer decisions in evaluating collective control options.