Spreadsheet duplicate finder

How matching works: after normalising the ticked columns, two rows are a duplicate pair only when every key column reaches the threshold. Blank key cells score zero, so a row with a blank key never matches anything. Candidate pairs are found by blocking on each column's first character and length band, which keeps the run fast but means two rows whose key text starts with a different character in every key column are never compared. The cleaned file re-joins the kept rows with standard CSV quoting, so cell values survive but number formats and styles from an Excel file do not. Nothing is removed until you download, nothing is uploaded, and the cluster report lists every cluster with its scores.