10.16 Release Notes: May 4, 2026-May 6, 2026¶

Apply data quality checks by group (General availability)¶

You can now associate a data metric function (DMF) with a table or view by using a WITHIN GROUP clause, so the DMF is evaluated separately for each distinct combination of values in one or more grouping columns. Instead of getting a single scalar value across all rows, you get one result row per group, which makes it easier to identify quality issues within specific segments (for example, null counts per region or duplicate counts per product category).

This feature is generally available and includes the following capabilities:

• Use ALTER TABLE ... ADD DATA METRIC FUNCTION ... WITHIN GROUP (col [, col ...]) to create a grouped association. Optionally append GROUP LIMIT <integer> to cap the number of groups per evaluation (valid range 1 through 1000, default 1000). The WITHIN GROUP clause can be combined with the existing FILTER clause on the same association.

• Use the new optional WITHIN_GROUP_VALUES argument on SYSTEM$DATA_METRIC_SCAN and SYSTEM$EVALUATE_DATA_QUALITY_EXPECTATIONS to scan error rows or evaluate expectations on demand for a specific group.

• Define expectations on grouped associations. Each group is evaluated and recorded independently in the event table, and Snowflake fires at most one notification per evaluation, based on the worst-group (maximum) metric value.

Supported for most system DMFs in the SNOWFLAKE.CORE schema (including NULL_COUNT, DUPLICATE_COUNT, and ROW_COUNT) and for most custom DMFs. The following aren’t supported with WITHIN GROUP in this release: the FRESHNESS and REFERENTIAL_INTEGRITY_COUNT system DMFs; schema-level associations (ALTER SCHEMA ... ADD DATA METRIC FUNCTION); anomaly detection (the per-association ANOMALY_DETECTION property and the SNOWFLAKE.CORE anomaly DMFs); and custom DMFs that use CTEs, UNION, UNION ALL, JOIN, DISTINCT, or window functions.

For more information, see Apply data quality checks by group.

New columns in DMF results and references views¶

The following views and table functions now expose grouping information for DMF associations and their results:

• DATA_QUALITY_MONITORING_RESULTS view and DATA_QUALITY_MONITORING_EXPECTATION_STATUS view (and their Information Schema table functions) include a new GROUP_BY_INFO ARRAY column that identifies the group for each result row. The column is an empty array for non-grouped associations.

• The DATA_METRIC_FUNCTION_REFERENCES table function and the DATA_METRIC_FUNCTION_REFERENCES view view include a new PROPERTIES VARIANT column that consolidates association properties as a JSON object. Two keys are specific to WITHIN GROUP: properties:within_group (a JSON-encoded string of the grouping column references; apply PARSE_JSON() to query as an array) and properties:group_limit.

• SYSTEM$EVALUATE_DATA_QUALITY_EXPECTATIONS returns a new GROUP_BY_VALUES VARIANT output column. The column contains a JSON object that mirrors the WITHIN_GROUP_VALUES input format, identifying the group for each expectation result.

For more information, see the following topics:

• Apply data quality checks by group

• DATA_METRIC_FUNCTION_REFERENCES

• DATA_METRIC_FUNCTION_REFERENCES view

• DATA_QUALITY_MONITORING_RESULTS view

• DATA_QUALITY_MONITORING_EXPECTATION_STATUS view

• SYSTEM$DATA_METRIC_SCAN

• SYSTEM$EVALUATE_DATA_QUALITY_EXPECTATIONS