Snowflake adds CoCo DDL workflow to Dynamic Tables Insights

Snowflake has made Dynamic Tables Insights generally available, pairing automated diagnostics for Dynamic Table SQL and refresh behavior with a remediation path through Snowflake CoCo. The feature evaluates each table’s declarative SQL definition and refresh execution metrics over time, surfacing recommendations when Snowflake detects inefficient incremental-processing patterns or runtime conditions.

In Snowsight, users can filter the Dynamic Tables list for tables with active recommendations, then open a table’s detail or overview page. An Insights panel appears beside the refresh graph and execution metrics. From there, selecting Analyze with CoCo starts a workflow that evaluates upstream and downstream dependencies, explains the flagged SQL structure and generates ready-to-use DDL for the user to review before applying a proposed change.

The diagnostics cover both query structure and refresh-execution characteristics. Snowflake cites nested QUALIFY RANK() or ROW_NUMBER() deduplication logic as one structural pattern that can prevent a Dynamic Table from using the fastest incremental path. The recommended remedy is to move the clause to the outermost SELECT. Runtime recommendations can also address refresh mode or warehouse sizing.

Teams do not have to inspect tables one by one. Snowflake’s documentation provides an INFORMATION_SCHEMA.DYNAMIC_TABLES() query that filters for nonempty recommendation arrays, creating a SQL-based inventory of tables with insights. Snowflake says the CoCo integration is also available through CoCo CLI and CoCo Desktop workflows, in addition to Snowsight.

Snowflake positions the feature as a way to reduce refresh execution time, refresh latency and compute or credit consumption. Those remain expected benefits, not measured results: the announcement and documentation provide no quantified savings benchmark or customer measurement.

Snowflake added the general-availability entry to its completed 10.33 release notes on September 18 and published the engineering announcement on September 21. The opened materials do not specify a separate exact availability timestamp or account, edition, region, privilege or rollout prerequisites.