Import a copy of the data and use facets to locate unexpected categories or missing values. Review clustering suggestions before merging names, since similar strings may identify different entities. Export the cleaned table and transformation history together, and retain original identifiers so corrections can be traced back to their source.
This is best suited to cleaning inconsistent tabular datasets. Before committing a larger project, account for these trade-offs: not designed for live multiuser editing; external reconciliation services have separate availability.