Onboarded a junior with this dataset. Categorisation was granular enough to segment by sub-vertical without regex. Recommended.