Choose a duplicate key you can defend
Two records with the same name are not automatically the same customer. A useful duplicate rule starts with the identity of the record.
Decide what one record represents
For a customer list, a customer_id may be sufficient. For order lines, the same customer can appear repeatedly, so customer_id alone would remove valid purchases. A combination such as order_id and line_id may be the right key. The tool can compare several columns as one tuple without concatenating their values into a potentially ambiguous string.
Choose retention independently
First and last mean position in this input, not oldest and newest timestamp. With id,name records 001,Avery; 001,Avery Updated; and 002,Morgan, keeping last retains the second 001 record. If a trusted date should determine retention, sort and verify the source before using a positional rule. Rowlane does not silently select the row with more complete fields.
Decide how spaces and case affect identity
By default, AB1 and ab1 are different, as are AB1 and a value with surrounding spaces. You can ignore case or trim key whitespace for matching. Those options change the grouping decision, not the values in the kept file. Write down the choice so another person can reproduce your result.
Review missing keys and removed records
A blank key is not an identity. Rowlane keeps records with a blank in any selected key field and adds a manual-review finding. Download removed records as well as kept records; their separation makes reversal straightforward. Before using the result, compare input count with kept count plus removed count and spot-check the duplicate groups you care about.