System data metric functions¶

This topic is a reference for the system data metric functions (DMFs) that Snowflake provides to all accounts. DMFs are the building block of data quality checks.

About system DMFs¶

Snowflake provides system DMFs in the CORE schema of the shared SNOWFLAKE database. System DMFs are maintained by Snowflake; you cannot change the name or functionality of any system DMF.

Each system DMF enables you to measure a different data quality attribute. You can assign more than one system DMF to a table or view to allow for a more comprehensive data quality measurement to address your governance and compliance needs.

The Statistics DMFs return a measurement rather than a pass or fail signal. You define the expectation that determines whether a given value indicates a problem. The Outliers DMFs apply a built-in threshold, such as Tukey fences or a Z-score cutoff, and return the number or percentage of values that fall outside it.

The Machine Learning DMFs compare a column of predicted values with a column of actual values, for example, to monitor the output of a machine learning model. Specify the predicted column first when you associate one of these DMFs with a table or view.

Many count and percentage DMFs measure undesirable values, so lower values are better. Other DMFs, such as NOT_NULL_COUNT, measure values that meet an expectation. The ROWS_IN_PAST DMFs in the Volume category count values in a trailing window that ends at the scheduled evaluation time; a value of 0 means that no data arrived.

System DMFs¶

Currently, Snowflake supports these system DMFs to measure common metrics without having to define them:

Some system DMFs measure closely related conditions:

• NULL_COUNT and NOT_NULL_COUNT are complements: together they count every row. The same is true of NULL_PERCENT and NOT_NULL_PERCENT.
• BLANK_COUNT counts values that are empty or contain only spaces, such as '' and ' '. EMPTY_STRING_COUNT counts only the empty string ''. Neither function counts NULL values.
• FUTURE_TIMESTAMP_COUNT counts values that are later than the scheduled evaluation time, and NOT_IN_FUTURE_COUNT counts values that are at or before it. Neither function counts NULL values, so together they count every non-NULL value.
• DUPLICATE_PERCENT returns the percentage of non-NULL rows whose value occurs more than once, and UNIQUE_PERCENT returns the percentage whose value occurs exactly once. Both exclude NULL values, so together they add up to 100 when the column has a non-NULL value.
• ALL_SPACE_COUNT counts non-empty values that contain only whitespace, including tabs, line breaks, and no-break spaces. Unlike BLANK_COUNT, it doesn’t count the empty string ''.
• CONTAINS_SPACES_COUNT counts values that contain a space, tab, or no-break space anywhere, and UNTRIMMED_STRING_COUNT counts only values with leading or trailing whitespace.
• INVALID_ISO8601_FORMAT_COUNT checks only the layout of a value, and INVALID_TIMESTAMP_STRING_COUNT checks whether the value can be parsed as a timestamp. For example, 2026-13-45 has a valid ISO 8601 layout but can’t be parsed.
• INVALID_INTEGER_COUNT counts values that aren’t written as a integer, and INVALID_NUMERIC_TYPE_CAST_COUNT counts values that can’t be cast to a number. For example, 3.7 is counted by INVALID_INTEGER_COUNT but not by INVALID_NUMERIC_TYPE_CAST_COUNT.
• TRUE_PERCENT, FALSE_PERCENT, and NULL_PERCENT add up to 100 for a Boolean column that has at least one row.
• SCHEMA_CHANGE_ADD_COUNT, SCHEMA_CHANGE_DROP_COUNT, and SCHEMA_CHANGE_TYPE_COUNT each count one kind of change that SCHEMA_CHANGE_COUNT also counts. Renamed columns are counted only by SCHEMA_CHANGE_COUNT.
• FRESHNESS measures how long ago a single table was last updated, and SOURCE_TARGET_LAG measures how far the latest timestamp in a target table trails the latest timestamp in a source table.

Related functions¶

DATA_METRIC_SCHEDULED_TIME isn’t a metric and you can’t associate it with a table or view. Call it in the body of a custom DMF to retrieve the scheduled evaluation time, for example to define a custom freshness metric.