Data Quality Assertions Facet

The facet records the results of tests (e.g. data quality assertions) run against a specific dataset or its columns. Each entry captures what was asserted, whether it passed, which column it targets, and optionally rich metadata such as the observed vs. expected values and the assertion body.

Use this facet to attach test outcomes directly to the dataset they validate. For run-level test reporting that is independent of any particular dataset, see the Test Run Facet.

Fields per assertion entry:

Field	Required	Description
assertion	yes	Test classification, e.g. not_null, unique, row_count, custom_sql. Equivalent to type in TestRunFacet.
success	yes	Whether the assertion passed (true) or failed (false). Equivalent to status: "pass"/"fail" in TestRunFacet. See success vs severity paragraph below.
column	no	Column being tested; omit for table-level assertions
name	no	Identifier for the test (e.g. assert_no_orphans).
severity	no	Configured severity: error (failure blocks pipeline) or warn (warning only). See success vs severity paragraph below.
description	no	Human-readable description of what the assertion checks
expected	no	Expected value or threshold, serialized as a string
actual	no	Actual observed value, serialized as a string
content	no	Assertion body (e.g. a SQL query)
contentType	no	Format of content, e.g. sql, json, expression
params	no	Arbitrary key-value pairs for assertion-specific inputs

success vs severity

These two fields are independent and serve different purposes:

• success reflects whether the test found issues: true means no issues were found, false means issues were found (regardless of whether the final pipeline execution was blocked).
• severity reflects the configured consequence of a failure: error means a failure blocks pipeline execution, warn means a failure produces a warning only and does not block.

A test can have success: false and severity: "warn" — meaning the test detected a violation, but execution continued. The overall run still succeeds. This combination is important to preserve in lineage metadata because it lets consumers distinguish between tests that are enforced hard constraints and tests that are advisory checks.

Relationship to TestRunFacet

DataQualityAssertionsDatasetFacet and TestRunFacet serve similar purposes with few differences:

	DataQualityAssertionsDatasetFacet	TestRunFacet
Attached to	A specific dataset	The run as a whole
Test type field	assertion	type
Result field	success (boolean)	status (string: pass/fail/skip)
Column scope	column field (dataset- or column-level)	Not applicable
Use when	Asserting quality of a particular dataset/column	Recording test executions for a run, independently of datasets

Example

{
    ...
    "inputs": [{
        "facets": {
            "dataQualityAssertions": {
                "_producer": "https://some.producer.com/version/1.0",
                "_schemaURL": "https://openlineage.io/spec/facets/1-0-3/DataQualityAssertionsDatasetFacet.json",
                "assertions": [
                    {
                        "assertion": "not_null",
                        "success": true,
                        "column": "order_id",
                        "severity": "error"
                    },
                    {
                        "assertion": "row_count",
                        "name": "assert_order_row_count",
                        "success": false,
                        "severity": "warn",
                        "description": "Expected at least 1000 rows",
                        "expected": "1000",
                        "actual": "999",
                        "content": "SELECT COUNT(*) FROM orders",
                        "contentType": "sql",
                        "params": {"threshold": "0.01"}
                    }
                ]
            }
        }
    }]
    ...
}

The facet specification can be found here.