How is data skew best described in the context of Snowflake data distribution?

• A. Skew occurs when data is uniformly distributed across keys.
• B. Skew is irrelevant to query performance.
• C. Skew improves parallelism across clusters.
• D. Skew occurs when a small subset of values accounts for a large portion of the data.

Answer: (D)

Data skew in Snowflake means the data isn’t evenly spread across the partitions, with a small subset of values containing a large portion of the rows. When a few values dominate the data, the partitions that hold those values become much heavier, creating hotspots and uneven work distribution for queries. This can slow down execution and reduce pruning efficiency, especially for filters on those skewed values, because some compute nodes end up handling far more data than others.

Uniform distribution would not exhibit skew, so that option doesn’t fit. Skew isn’t irrelevant to performance—quite the opposite, it can cause slower queries and uneven resource use. And skew does not improve parallelism; it tends to harm it by creating imbalanced workloads across compute resources.