Drug analysis and development with PET should fully exhaust the ability of this tomographic technique to quantify regional tracer concentrations in vivo. Data evaluation based on visual inspection or assessment of regional image contrast is not sufficient for this purpose since much of the information present in dynamically acquired data is not used by these approaches. Compartment modelling of dynamic PET data is generally the method of choice since it allows a quantitative assessment of the underlying pharmacokinetic parameters describing drug transport, metabolism and molecular interactions. We present here an overview of key issues of compartment modelling with specific attention to the assumptions underlying the various models and their limitations. We believe that a thorough understanding of the applicability of models is mandatory for the development, successful execution and analysis of quantitative PET studies. Otherwise, meaningful and interpretable results will often not be obtained.