Name the record you are trying to keep
One customer can have several jobs; one product can have several variants. A repeated customer name or product family is not enough to identify redundant rows. Write a rule such as one row per supplier product code and pack size, then identify the fields that implement that rule. Confirm how missing key fields should be handled.
Classify the matches before taking action
Divide candidates into exact duplicates, repeated keys with compatible information, and repeated keys with conflicting information. The second category may need an agreed merge rule. The third needs a decision about the correct source or separate valid records. Do not choose the first row simply because it appears first in the file.
- Exact duplicate: the agreed identity and relevant values match.
- Possible merge: the key matches but one record fills a missing field.
- Conflict: the key matches and an important value differs.
- Incomplete key: there is not enough information to decide.
Understand what the tool will remove
Excel's Remove Duplicates uses the selected columns to identify matches and removes entire duplicate rows. A value in an unselected column can therefore disappear with its row. Microsoft's guidance also distinguishes filtering unique values from deleting duplicates. Previewing candidates is a useful separate step before a removal decision.
Source: Microsoft Support — Filter for unique values or remove duplicate values.
Keep a decision log and source references
Retain the unchanged source, stable row references and a list of proposed actions. State the approved key, comparison normalization and survivor rule. Trimming accidental spaces may be appropriate for comparison, but a normalization that erases meaningful punctuation or product suffixes can create false matches. Keep original values accessible for review.
Reconcile the before and after counts
Record the starting rows, approved removals, merged groups and unresolved candidates. Compare the output with that action log. Check representative survivors and every material conflict under the agreed review scope. A smaller file is not by itself evidence that cleanup was correct; each removed record should have a defensible relationship to what remains.
Make recurring cleanup repeatable
Save the rule with the output, including the fields that must be present. A future export with different columns or a changed meaning of product code needs a fresh check. Stop the automated or manual rule when its assumptions no longer hold. For a service quote, describe the row unit, candidate keys, source count and known conflicts without sending the full list through the website form.
A few useful answers
Is a repeated email address enough to remove a row?
Not necessarily. Decide what each row represents; the same contact may legitimately appear on several projects or transactions.
Should I always keep the newest row?
Only when the timestamp has an agreed meaning and the newer record is authoritative for this task. Modified time alone may reflect copying rather than a data revision.
Can matching rows be deleted automatically?
Only under a defined, authorized rule with an unchanged source and adequate checks. Ambiguous keys and conflicting values need separate review.