Define “same record” first
Two rows can share a city, status, or date and still represent different transactions. A duplicate key should reflect the business meaning of one record. Write that rule in a sentence before selecting columns: “Rows are duplicates when the same SKU appears in the same region,” for example.
Choose the key pattern
One stable identifier
Use one column when an order ID, SKU, or authorized email address is the record identity. Check that the field is actually unique and consistently formatted. A repeated display name is usually not a safe identifier.
A combination of fields
Use several columns when no single field is unique. Date plus SKU, or customer ID plus invoice number, may define the record. Include every field required to distinguish legitimate variations.
All relevant columns
Choose all columns when two rows count as duplicates only if their displayed values match throughout the selected sheet. This is conservative about deleting rows, but small differences such as a timestamp or note will keep both rows.
Worked example
Consider three rows:
| Row | SKU | Region | Quantity |
|---|---|---|---|
| 2 | A-17 | North | 4 |
| 3 | A-17 | North | 4 |
| 4 | A-17 | South | 2 |
With SKU plus Region as the key, row 3 duplicates row 2 and row 4 remains. With SKU alone, row 4 joins the same group, which may be wrong if region changes the meaning of the record.
Preview before removal
Open the duplicate-group preview and inspect the retained row, later matches, blanks, capitalization, leading zeros, and text-versus-number values. Try the keep-first strategy only when row order represents preference. If the newest or most complete row should remain, do not silently keep the first row.
When to stop
- The key would combine personal fields without an authorized purpose.
- The workbook contains formulas, macros, styles, charts, or links that the browser export does not preserve.
- The preview groups together rows that are legitimately repeated.
- You cannot explain why one row should be kept.
Safe export checklist
- Keep an untouched source copy.
- Record the selected worksheet and key columns.
- Compare row counts before and after.
- Open the exported file in the destination application.
- Keep the output private and use only authorized data.
For a complete workflow, read how to find duplicates before deleting. For blanks and capitalization, see the cleanup checks or open the Excel guide.