Majority of queries scan under 100 GB indicating engine mismatch

When 80%+ of read queries scan under 100 GB (especially on Small or larger warehouses), this indicates a workload placement problem rather than a pure optimization problem. Public analyses show 99.9th percentile queries scan ~300 GB, and ~99% scan under 100 GB. These filtered aggregations, small joins, and reporting rollups don't need distributed warehouse compute but get billed at Snowflake's credit rates.

Run the provided audit query to measure what percentage of queries scan under 100 GB. If most read workload concentrates in smaller buckets and queries return in seconds, consider multi-engine routing to cheaper execution paths (e.g., DuckDB) for read-heavy BI and analytics while keeping large scans on Snowflake.