Rare categories in original data can result in re-identification when regenerated in synthetic data, so privacy-protection mechanisms must be used during the synthesization process to ensure privacy. Example of rare categories are highly unique values in the dataset such as CEO as a job title.