I answered the question by first walking them through my approach conceptually before jumping into the SQL. I explained that to identify duplicates, I would group the records by the relevant columns and use a HAVING COUNT(*) > 1 condition to filter those groups. I also explained that depending on the business need, duplicates could be handled in different ways — for example, keeping the latest record based on a timestamp, removing all duplicates, or flagging them for review. I made sure to mention that in production, I’d also check with stakeholders to confirm which fields define uniqueness, as that can vary across datasets. The interviewer seemed more interested in my thought process and how I communicate technical steps clearly.