01
Structure
Do expected fields exist, and are types and formats valid?
Home / Quality & Lineage
Trust is built into the workflowMake data quality and lineage part of the design.
Data is easier to trust when teams can check important assumptions and trace how a result was produced.
A quality ladder
Test from basic shape to business meaning.
Not every table needs the same checks. Focus first on the failures that would change a decision or break a downstream workflow.
02
Completeness
Are required values present within the expected range?
03
Relationships
Do keys, joins, and references match the rules of the model?
04
Meaning
Do business definitions and metric assumptions still hold?
Trace the lineage
Follow a result back to its source.
Lineage shows how datasets and transformations depend on each other. It can help answer: where did a metric come from, what changed upstream, and which reports might be affected?
Lineage is most useful when metadata stays connected to the actual pipeline runs and model definitions. A diagram without updated ownership or definitions can create false confidence.
Explore architecture notes →Application events
Original records and schema versions.
Customer activity
Typed fields and reusable definitions.
Active users
Documented time window and rule.
Product report
Decision context and access owner.
Ownership makes quality actionable.
Data owner
Defines meaning
Maintains business definitions, sensitivity labels, and expected use.
Pipeline owner
Maintains delivery
Owns freshness, transformation logic, retries, and alert routing.
Data consumer
Uses with context
Knows how to interpret the data and report mismatches.