Collective adaptive systems (CAS) consist of a large number of possibly heterogeneous entities evolving according to local interactions that may operate across multiple scales in time and space. The adaptation to changes in the environment, as well as the highly dispersed decision-making process, often leads to emergent behaviour that cannot be understood by simply analysing the objectives, properties, and dynamics of the individual entities in isolation. As with most complex systems, modelling is a phase of crucial importance for the design of new CAS or the understanding of existing ones. Elsewhere in this volume the typical workflow of formal modelling, analysis, and evaluation of a CAS has been illustrated in detail. In this chapter we treat the problem of efficiently analysing large-scale CAS for quantitative properties. We review algorithms to automatically reduce the dimensionality of a CAS model preserving modeller-defined state variables, with focus on descriptions based on systems of ordinary differential equations. We illustrate the theory in a tutorial fashion, with running examples and a number of more substantial case studies ranging from crowd dynamics, epidemiology and biological systems.