Snowflake Search Optimization Service consumes credits to build and maintain an access path for selective point lookups and other supported predicates.
The service builds an index-like structure over selected columns and maintains it as table data changes. Initial build work can be substantial for a large table. Ongoing maintenance depends on the volume and frequency of changes.
Search optimization can reduce the data scanned by selective queries. It is most useful when those queries are frequent enough, and selective enough, to offset the service's build and maintenance credits.
Snowflake provides SYSTEM$ESTIMATE_SEARCH_OPTIMIZATION_COSTS to estimate the credits required for selected columns before enabling the service. Account usage views can be used to compare maintenance credits with query execution time and bytes scanned.
A test on a small table may not represent the cost of building the structure on a production table. Measure initial build credits separately from ongoing maintenance and query savings.
Snowflake documents the Search Optimization Service and its cost-estimation function.