Define a dependable identity key, review every repeated group, choose which row survives, and optionally fill its blanks from the duplicates.