Structure determinationof molecular solids through NMR crystallographyrelies on the generation of a comprehensive set of candidate crystalstructures and on the comparison of chemical shifts computed for thosecandidates with experimental values. Exploring the polymorph landscapeof molecular solids requires extensive computational power, whichleads to a significant bottleneck in the generation of the set ofcandidate crystals by crystal structure prediction (CSP) protocols.Here, we use a database of crystal structures with associated chemicalshifts to construct three-dimensional interaction maps in molecularcrystals directly derived from a molecular structure and its associatedset of experimentally measured chemical shifts. We show how the mapsobtained can be used to identify structural constraints for acceleratingCSP protocols and to evaluate the likelihood of candidate crystalstructures without requiring DFT-level chemical shift computations.