Before
Define the file, its volume and the expected destination.

Prepare one bounded CSV dataset, understand its columns and produce a checked target export. Each useful transformation should be explainable and reviewable.
An older list may mix formats, empty fields or apparent duplicates. Automatically removing similar-looking records can discard legitimate information. The work therefore starts with mapping rules and exceptions to review.
The supplied source file is preserved as a reference. A trial on a copy reveals conversions and anomalies before the agreed export is produced. You approve the processing rules; ambiguous cases are isolated rather than guessed at.
Define the file, its volume and the expected destination.
Test the rules on a copy and record discrepancies.
Check meaningful totals and hand over the remaining exceptions.
The original file stays separate from the working file.
Agreed normalisations are explicit, not silent.
Uncertain corrections remain subject to your approval.
Provide an authorised export and retain a reference backup.
Approve the mappings using a trial copy.
Review exceptions before using the target export.
Not without reliable rules. Similar records are reviewed against agreed criteria; uncertain cases are presented for your decision.
The proposed starting scope covers files and a trial copy. A live import requires separate agreement, authorised access and suitable precautions.