Can you define a clustering key directly on a VARIANT column?

• A. Yes
• B. No
• C. Only if the VARIANT is an object
• D. Only with a specific data type

Answer: (B)

Clustering keys rely on stable, directly comparable scalar values to guide micro-partition pruning. A VARIANT column stores semi-structured data, not a single primitive value, so it can’t be used directly as a clustering key. To cluster effectively on data inside a VARIANT, you extract the specific field into a separate column (or a generated/computed column) with a primitive data type, and then define the clustering key on that column. That’s why you can’t define a clustering key directly on the VARIANT column.