"Understanding atomic disorder in functional materials via total scattering" - Crystallography allows materials scientists to understand and engineer atomic structure, but these tasks become complex when the long-range structures are incoherent (crystals with intrinsic disorder) or non-existent (liquids and glasses). In both cases, large-box models are often required to faithfully model materials, but they cannot be constructed using traditional least-squares refinements. I will show how data-driven reverse Monte Carlo fits to pair distribution function (PDF) data can effectively create atomic snapshots of materials. Multiple types of experimental data and chemical knowledge can be incorporated, with the unique advantage that these constraints (PDF, Bragg, bond valence) can be satisfied simultaneously. Careful statistical analysis of the resultant models is crucial to understanding the physical meaning of the atomic arrangements in the model. Specific examples will be presented from a range of structural problems including functional inorganic materials and crystal growth.