A data engineering team stores clickstream data in an S3 data lake queried by Athena. Analysts almost always filter queries by a specific date range and, less frequently, by AWS Region. Currently all files are written to a single flat prefix, and query costs and scan times are high. The team wants to design a partitioning scheme that minimizes data scanned for the most common query patterns while avoiding an excessive number of tiny partitions. Which partitioning approach should they choose?