I'm working with a SQL Server table that contains some duplicate records, and I need to clean them up without accidentally removing legitimate rows.

The duplicates are identified by a combination of several columns rather than by the primary key. I was thinking about using ROW_NUMBER() with a CTE to identify duplicate records and then removing everything except the first row.

Is this the safest approach for a production database? Are there any additional checks I should perform before running the DELETE statement, especially when the table contains a large amount of data? Mapquest Driving Directions

I'd also like to know how others usually verify the affected rows before actually removing them.