The Data Lineage Diagram Engineers Actually Trust

Incidents you can't trace upstream

A revenue dashboard misfires. Hours go into bisecting which upstream table, dbt model, or Spark job introduced the bad value. Without an accurate lineage visualization, root-cause analysis becomes archeology across Slack threads and Git history.

Downstream impact is invisible until production

Schema changes ripple silently through dbt models, ML features, and BI dashboards. Without downstream impact visibility, a renamed column or dropped field reaches finance, analytics, and ML pipelines before the engineer who shipped it even notices.

Audits without data lineage documentation

Compliance teams ask where regulated data originated, which transformations touched it, and who has read access along the path. Without trustworthy data lineage documentation, that answer takes days of manual reconstruction across SQL files, dbt projects, and BI semantic layers.

Column changes silently break reports

A renamed column. A dropped field. A type coercion. Without column level lineage, the first signal is a blank Looker report, a failing dbt test in production, or a stakeholder asking why yesterday's MRR figure has vanished.

Column level lineage, parsed automatically from your warehouse

DataHub parses SQL query history across 20+ dialects, including Snowflake, BigQuery, Redshift, Databricks, and Postgres, to derive column-to-column dependencies without any manual mapping. Joins, CTEs, window functions, subqueries, and CASE expressions are all resolved at parse time, so the lineage graph reflects what your warehouse is actually executing, not a hand-drawn diagram that drifted out of date six sprints ago.

• Automatic column mapping from SQL query logs, dbt manifests, and OpenLineage events
• Transformation logic surfaced inside each lineage node, so reviewers see the actual SQL, not just edges
• Python SDK and GraphQL API for custom column-level ingestion from proprietary pipelines

Quantify downstream impact before you merge the PR

Before renaming a column, deprecating a table, or changing a join key, DataHub surfaces every downstream asset that depends on it: dbt models, Airflow DAGs, Looker explores, Tableau workbooks, ML feature stores, and consuming services. Filter the blast radius by degree of separation, owner, or domain, so you can route a heads-up to the exact teams that need it before the change ships.

• Bidirectional traversal: search upstream sources and downstream consumers across the full graph
• Filter by entity type: datasets, charts, dashboards, dbt models, ML features, pipelines
• Degree-of-separation filtering to scope blast radius from a one-hop check to the full transitive closure

Data lineage documentation auditors and engineers both accept

Export any lineage diagram as a PNG, named after the entity, for incident postmortems, architecture review documents, SOC 2 evidence binders, and GDPR data-flow records. Where automated parsing cannot reach legacy systems or proprietary jobs, engineers can edit lineage edges manually with edit privileges, and every change is logged with author, timestamp, and source, producing data lineage documentation that survives an external audit.

• PNG export with entity name as filename, drop-in ready for runbooks and compliance evidence
• Add, remove, or correct lineage edges under fine-grained edit privilege controls
• Every manual change logged with user, timestamp, and rationale for full audit traceability

A lineage visualization built for exploration, not just inspection

The DataHub lineage visualization is interactive and built for real investigation work. Expand nodes one hop at a time, collapse branches you don't care about, switch between table-level and column level lineage in the same view, and follow a path from raw Kafka topic to executive dashboard without losing context. Every node deep-links to its full metadata profile: schema, owners, freshness, quality assertions, and recent incidents.

• Expand and collapse nodes to manage graph complexity on dense pipelines
• Filter by platform, owner, domain, tag, or data product to scope the view
• Click any node to open its full metadata profile inline: schema, owners, quality, incidents