Keep the original file and write down the expected record count and key columns. Work on a duplicate. Do not overwrite the source while experimenting.
1. Profile the source
Check the delimiter, character encoding, header names and the number of fields in each record. Identify identifiers that must remain text, such as postal codes or customer IDs. Use the CSV Data Profiler and mixed-type audit before changing anything.
2. Define the rules before cleaning
Specify whether whitespace is meaningful, whether duplicated IDs represent duplicates or legitimate revisions, and whether an empty field means unknown, not applicable or an error. For imports, capture the target system's date and required-column rules.
3. Make one transformation at a time
Start with the CSV Cleaner or Duplicate Row Remover. For spreadsheet files, inspect the worksheet structure before exporting cached cell values. Download each intermediate result and keep a simple change log.
4. Check for accidental data loss
Compare row counts and unique-key counts, then review records from the top, middle and bottom of the file. Include empty, quoted, Unicode and unusually long values. Check whether a deduplication rule removed intentional repetitions.
5. Test the destination import
Import a small sample into the real target application before importing the full dataset. Confirm column matching, field types, leading zeros, date interpretation and any CSV formula-safety measures. Revert to your untouched source if the result does not match your intended meaning.
Redaction and pattern detection are heuristic. Do not assume that masking a few email or phone patterns makes data anonymous or legally compliant. Review every export before sharing it.
Related resources
Diagnose common CSV errors · Understand XLSX limitations · See parsing methodology