Structure-based drug design is an integral part of modern day drug discovery and requires detailed structural characterization of protein-ligand interactions, which is most commonly performed by X-ray crystallography. However, the success rate of generating these costructures is often variable, in particular when working with dynamic proteins or weakly binding ligands. As a result, structural information is not routinely obtained in these scenarios, and ligand optimization is challenging or not pursued at all, representing a substantial limitation in chemical scaffolds and diversity. To overcome this impediment, we have developed a robust NMR restraint guided docking protocol to generate high-quality models of protein-ligand complexes. By combining the use of highly methyl-labeled protein with experimentally determined intermolecular distances, a comprehensive set of protein-ligand distances is generated which then drives the docking process and enables the determination of the correct ligand conformation in the bound state. For the first time, the utility and performance of such a method is fully demonstrated by employing the generated models for the successful, prospective optimization of crystallographically intractable fragment hits into more potent binders.