We present a summary of the National Compound Collection (NCC) pilot; which harvested chemical structure data from 746 publicly-available PhD theses to create an enhanced database of diverse and interesting (largely organic) molecular entities. The database comprised ∼75 000 structure entries, of which 70% were new to ChemSpider at the time of upload. The dataset was evaluated for structural uniqueness by twelve external drug discovery groups from the pharmaceutical, biotech, academic and not-for-profit sectors. These partners generated data reported here comparing the NCC pilot with their in-house compound collections. The proportion of NCC structures considered to be useful for drug discovery ranged from 5–80% depending on the strictness of the filters used; most interestingly from a drug discovery standpoint ∼13k NCC compounds (18% of the NCC) passed the filters and were of good diversity. These compounds are quite different from those that are already present in the screening collections but not so different that they are no longer considered to be drug-like. In general, the drug discovery teams would consider these compounds to be high value molecules for inclusion in their screening collections. This pilot addressed the potential value of unpublished data and explored the practicalities of large-scale data extraction, to inform both retrospective and prospective extraction of chemical data from theses.