Import only directory.en.csv and training.en.csv. Data cleaning and analysis → Join and compare (original tables): directory left, training right, staff_id key, Full join. Run Join by key and download CSV. Inspect source-side headers and verify R01–T01, R02–blank, blank–T02; all four source rows have a destination. In a separate run change the key to name: three rows remain, but links become R01–T02, R02–T02 and blank–T01. Reject this result, restore staff_id and rerun. The join trims surrounding key whitespace; these inputs contain none, and that normalization must fit your identifier rules. Two roster identities are known; complete headcount is unknown. No match does not prove nonattendance.
