A complex material, like concrete, can be considered a system rather than a material.Looking at concrete as a system helps us to understand its complexity and its response to external loads and long-term performance and it opens the way for modifying traditional concrete mixtures and designing new ones with predefined properties.In all this, models play a crucial role.The actual relevance of understanding and modelling cementitious materials is presented in this paper.By referring to an inventory of long-term performance of concrete structures it is illustrated how important it is to know the consequence of adopting changes in mix design.Different types of models, i.e.models acting on different length scales, are briefly presented and outstanding fundamental questions are listed.Then three examples are presented where experimental studies and numerical modelling have revealed new insight in the performance of cementitious materials.First the use of a micro-scale model is discussed for mixture optimization of blended cements.The second example shows that simplified models for the evolution of transport properties in cement-based materials in view of service life predictions exhibit shortcomings, which can only be understood and tackled by performing in-depth experimental investigations.Finally an example is shown of modelling self-healing of microcracks.That study revealed that self-healing of cracks is not only a matter of filling a crack with reaction products, but also of densification of the microstructure adjacent to the crack.