Halide perovskites are enabling a variety of high-performance devices, yet these materials exhibit heterogeneity in their chemical, structural, morphological and optoelectronic properties on different length scales. A global understanding of the relationship between these properties, and how they evolve under operation, is essential for commercialization. This perspective covers efforts to attain such relationships through multimodal and correlative microscopy. A variety of ex-situ, in-situ and operando measurements have led to key conclusions that the length scales that are extremely important – yet least understood – are on the grain, sub-grain and atomic scales. The multi-faceted disorder of these materials and ionic nature makes complete characterization challenging, but exploitation of the latest experimental probes is stretching our knowledge, allowing rational approaches to improve performance and stability. Future adaptions will employ machine-learning driven algorithms to process large hyperspectral data sets and operando measurements on the manufacturing line, in turn refining processing and device architectures.