- Cuts query runtime and compute cost by choosing the most efficient execution plan
- Reduces data scanned through index use, partition pruning, and predicate pushdown
- Improves concurrency by freeing resources sooner per query
- Makes interactive analytics feasible on large datasets