Natural and manufactured materials rely on complex hierarchical microstructures to deliver a suite of interesting properties. To predict and tailor their performance requires a joined-up knowledge of their multiphase microstructure, interfaces, chemistry and crystallography from the nanoscale to the macroscale. This Perspective reflects on how recent developments in correlative characterization can bring together multiple image modalities and maps of the local chemistry, structure and functionality to form rich multimodal and multiscale correlated datasets. The automated collection and digitization of multidimensional data is an essential part of the picture for developing multiscale modelling and 'big data'-driven machine learning approaches. These are needed to both improve our understanding of existing materials and exploit high-throughput combinatorial synthesis, processing and testing methods to develop materials with bespoke properties.