A retail analytics team stores 8 years of immutable transaction history (about 60 TB) in Parquet files on Amazon S3, partitioned by year and month. Analysts query the most recent 3 months daily for dashboards, but the older data is queried only a few times per quarter for ad hoc audits. The team wants to minimize storage costs for the cold data while keeping fast, frequent queries on the recent data performant, all within a single query interface. Which approach best meets these requirements?