When using a column with very high cardinality as a clustering key, Snowflake recommends:

• A. Define the key directly on the column
• B. Use a hashed surrogate key
• C. Define the key as an expression on the column
• D. Do not cluster on that column

Answer: (C)

When a clustering key has very high cardinality, the real goal is to improve data pruning by grouping rows into a smaller, meaningful set of partitions. Clustering by the raw high-cardinality column tends to create many distinct values across micro-partitions, which offers little pruning benefit and can add maintenance overhead. Defining the clustering key as an expression on that column lets you bucket or transform the values into a smaller, fixed set of clusters, which Snowflake can prune much more effectively during queries. For example, clustering by a hashed expression or by truncating a timestamp to a specific granularity (like month or quarter) reduces the number of distinct cluster values while preserving the ability to filter on the original column. This approach avoids adding new columns while giving you practical pruning improvements. Clustering directly on the raw column or avoiding clustering on it misses these efficiency benefits, and using a hashed surrogate column introduces extra complexity and storage without the same flexibility.