The development of modelling in aquatic ecology has focused on mechanistic biogeochemical models. However, such models have substantial data requirements for inputs and also for proper validation, which hinders their use for less studied systems. Another significant problem with complex models is their structural and computational difficulty, and thus often an associated absence of proper uncertainty analysis for the model results. This makes the use of the outputs for public policy making (e.g. in lake management) rather questionable. We see no compelling reason (other than lack of awareness of choices) why all lakes and all questions should necessarily be studied using the same high-profile models. Here we review two alternative statistical approaches, Linear Mixed Modelling and Structural Equation Modelling, and the different ways they have been used to extract maximum information from existing data. These methods offer promise for tackling the problems highlighted above, although our aim is not to promote any one method over the others. Rather, we want to stimulate debate about the remaining unknown factors in lake modelling as well as about the balance between data and models, and the still too uncritical way in which model outputs are interpreted and used for decision making.