The present level of biodiversity depletion and loss, and the diffusion of new geotechnologies create the outlines for a new paradigm, where spatial data is extremely valuable. Quality datasets may be used to support decision-making processes in public policies related to biodiversity conservation. Specimen datasets used by the scientific community for their analysis starts to become available to the public and private sectors. During the last two decades several important datasets like Species 2000, The Plant List, JABOT among others, and access platforms like GBIF or SpeciesLink have become available. On the other hand, widespread access to technologies has made geoprocessing and spatial analyses easier. However poor data quality is still critical and limits the usefulness of these datasets. Thereby, data quality assessments are important to ensure a responsible use of those datasets. The Brazilian National Centre for Flora Conservation was created in 2008, at the Rio de Janeiro Botanical Garden, with the main objective of assessing the extinction risk of plant species and to plan conservation actions. In this context, a dataset was created after the compilation of occurrence records (248,837) of 4,711 threatened species, obtained from 70 herbaria. The present study aims to assess quality of the dataset and records, and to test quality improvement after data cleaning efforts. We have used the five-component scheme for assessing dataset quality. Significance of the differences between expected and observed proportions were tested using the degree of confidence between proportions by the software R. The Mann-Whitney test was used to compare errors between the original dataset and the cleaned out one. Results indicate poor quality, not only for dataset (p<0.10) but also for records (p<0,10). Only 54,306 records (22.30%) were considered of good quality. Logical inconsistencies in the dataset were present in 8,237 records (3.37%). Historical collections of Brazilian herbaria are composed by different datasets, which were incorporated gradually over time without proper metadata. And for being able to use herbarium datasets for supporting the decision-making processes on biodiversity it is important to keep all metadata and appropriate documents that proves veracity of data.