Whole-row or column-based duplicates?
A whole-row match only removes rows where every cell is identical. That is the safest choice when rows were pasted or exported twice.
Column-based matching is what you need for contact lists, orders and CRM exports, where the same customer can appear with a different phone number or note. Pick the columns that identify a record, for example Email, or First name + Last name + Company, and later rows with the same values are removed.
Keep the first or the last copy
By default the first occurrence is kept. Choose Last occurrence to keep the most recent row instead, which is usually right when newer rows were appended to the bottom of an export. The rows that remain stay in their original order.
Near-duplicates hiding from exact matching
Exports often contain values that look identical but differ by a trailing space, a capital letter or an invisible character copied from the web. The quality check counts these as near-duplicates. Turn on Ignore capital letters and extra spaces, or add Trim spaces and Remove invisible characters from the Text section, and they are caught too.
Blank values are never treated as duplicates
When you match on a column such as Email, rows where that column is empty are kept. Two contacts without an email address are not the same person, so they are not merged. Use Remove rows with missing values if you want to drop them.