When defining a multi-column clustering key for a table, the order of columns in CLUSTER BY should be:

• A. From highest cardinality to lowest cardinality
• B. From lowest cardinality to highest cardinality
• C. Alphabetical
• D. Random

Answer: (B)

The order of columns in a CLUSTER BY key determines how Snowflake prunes micro-partitions during a query. Placing the column with the fewest distinct values first creates a small, broad set of partitions to start from, and then adding a second column with more distinct values narrows the search further inside those partitions. This approach tends to maximize pruning efficiency, reducing I/O and speeding up queries. If you start with a highly distinct column, you can end up scanning many partitions before the filters narrow things down, which is less efficient. Alphabetical or random ordering doesn’t reflect data distribution or typical query patterns, so it usually won’t improve pruning.