Clean data without losing records

Draft learning material, awaiting human review.

Sort whole records together. Sorting only a salary column detaches salaries from names. Filtering hides nonmatching rows; it does not delete their records.

Define duplicates using a suitable key. Two people with the same surname are not necessarily duplicates. Keep a recoverable source copy before cleanup and document your matching rule.

Use validation lists for controlled categories. Check numbers stored as text, accidental whitespace, missing values, and blank rows. Reconcile row counts and totals after an import. A missing value is not automatically zero.

Two regions contain 15 and 18 records before a merge. A 33-row result is a useful first check, but still check unique keys and totals.