Data Quality Checks

Data Quality Checks validate the data behind a published Data Mart before you use it in reports and dashboards. Configure checks for the whole Data Mart, individual Output Schema fields, and relationships, then run them manually or on a schedule.

Data Quality Checks are supported for Google BigQuery, Legacy Google BigQuery, AWS Athena, Snowflake, AWS Redshift, and Databricks in Cloud and Self-Managed editions.

Before You Start

A Data Mart must have:

• Published status
• A saved Output Schema
• At least one applicable, enabled check

Open the Data Mart and select the Data Quality tab to configure checks and view the latest report.

Available Checks

Scope	Check	What it detects
Data Mart	Empty table	A Data Mart with no rows
Data Mart	Primary key uniqueness	Duplicate values across the Output Schema primary key
Data Mart	Duplicate rows	Rows duplicated across all supported materialized Output Schema fields
Field	Null rate	A percentage of NULL values above the configured threshold
Field	Column uniqueness	Repeated non-null values
Field	Constant column	A field containing only one distinct value
Field	Type mismatch	A stored column type that does not match the saved Output Schema
Field	Data freshness	A latest timestamp older than the configured threshold
Field	Negative values	Numeric values below zero
Relationship	Relationship integrity	Source join values missing from the target Data Mart
Relationship	Reverse relationship	Target join values that are not used by the source Data Mart

Checks are shown only where they apply. For example, Primary key uniqueness requires primary key fields, Negative values requires a numeric field, and relationship checks require a configured Data Mart relationship.

Data freshness is available only for timestamp types that unambiguously represent an instant:

• Google BigQuery and Legacy Google BigQuery: TIMESTAMP
• AWS Athena: TIMESTAMP WITH TIME ZONE
• AWS Redshift: TIMESTAMPTZ
• Databricks: TIMESTAMP

Data freshness is not available for Snowflake fields.

Configure Checks

1. Open a published Data Mart and select Data Quality.
2. Expand Checks configuration.
3. Enable the Data Mart checks you need.
4. Under Field checks, click Add checks, select a field, and choose a check.
5. Enable the required Relationship checks for existing Data Mart relationships.
6. Select a severity:
   • Error for critical issues
   • Warning for issues that need attention
   • Notice for informational findings
7. Configure a threshold where required:
   • Null rate uses a percentage threshold.
   • Data freshness uses a threshold in hours.
8. Click Save.

The system preset initially enables basic checks such as Empty table, primary key checks when a primary key exists, null-rate checks for key fields, and relationship integrity checks. You can replace the preset with your own saved configuration or disable every check.

Run Checks

You can start checks in several ways:

• On the Data Quality tab, click Run. If the configuration has unsaved changes, use Save & Run.
• On the Data Marts list or Models canvas, use Actions → Check Quality to queue checks for multiple Data Marts.
• On the Data Mart’s Triggers tab, create a Data Quality Run trigger to run the saved configuration on a schedule.

Only one Data Quality run can be active for a Data Mart at a time. A bulk action queues each eligible Data Mart independently, so one ineligible item does not prevent the others from starting.

Review Results

The summary at the top of the Data Quality tab shows the latest run state and counts by severity. The detailed report below it contains:

• The status of every enabled check
• The number of violations
• Up to three example values for a failed check
• The SQL used to execute and reproduce the check

Checks can finish as Passed, Failed, Not applicable, or Execution error. Findings are grouped by Error, Warning, and Notice severity. A completed run can report issues without being an execution failure.

Data Quality runs also appear in:

• The Data Mart’s Run History
• Project-level Run History
• The Models canvas, where status icons summarize the latest result

Run History keeps the configuration snapshot and check results from that specific run.

Consumption

In the Cloud edition, each successfully completed run counts as one Data Quality Process Run, regardless of how many checks it executes or whether it finds data quality issues. Manual, bulk, and scheduled runs follow the same rule.

Warehouse query-processing costs remain with the connected storage.