The Mental Model
Lineage is the map of how data flows. It helps you debug wrong numbers, assess change impact, and explain how a metric was produced.
Lineage is like a family tree for data. If one parent changes, you can see which children may be affected.
Dataset reference: ecommerce tables and grain assumptions.
Trace a unit change to consumers
A source changes amount from dollars to cents. Trace the dependency before editing a downstream dashboard.
raw_orders.amount_cents
-> stg_orders.amount (divide by 100)
-> fct_orders.net_amount
-> net_revenue metric
-> finance dashboard
-> revenue APIUse the dependency graph in dbt artifacts to identify model edges, then add consumers that live outside dbt. A graph can show that one model depends on another without proving the units or every field-level transformation. Review SQL, model descriptions, exposures, and consumer ownership together.
Acceptance check: the fixture value 1250 becomes 12.50 exactly once, every affected consumer has an owner, and the migration plan covers both old and new source versions.
Interactive Check
Question: raw_orders.amount changes from dollars to cents. Which downstream objects might be impacted?
Reveal the answer
Any staging model using amount, any fact table deriving revenue, any revenue metric, and all dashboards or AI tools consuming that metric.
Practice: Trace the Blast Radius
Follow one changed source column through models, metrics, and consumers.
Use the guided lab below to record your result, assumptions, and the check that would catch an incorrect result.