Skip to content
Upstream

FAQ, read 10 Nov 2026

30 lineage questions, answered with names and numbers

Read the methodThe ranking

What is data lineage?
Data lineage is a map of where data comes from, how it changes, and what reads it. A good map runs from source tables through transformations to dashboards. Analytics engineers use it to trace a wrong number upstream and to see what a change will break downstream. Depth varies: table, column, BI field, or metric.
What is the difference between column-level and table-level lineage?
Table-level lineage says table B is built from table A. Column-level lineage says column revenue in B comes from amount and tax in A. All 16 tools here show table and column lineage per the evidence read on 10 Nov 2026, but they differ on transformation logic, joins and BI fields. Column detail is what makes impact checks precise.
What does BI lineage mean, and why does it matter?
BI lineage carries the trail past the warehouse into dashboards and reports. At its best it maps a single chart field to the warehouse column behind it. That is the last hop an analytics engineer needs when a stakeholder asks why a number moved. Tools that stop at the dashboard icon leave the final step to manual work.
Which lineage tools reach BI fields?
Per the evidence read on 10 Nov 2026, four of 16 tools reach BI fields with a full yes: DataHub, Euno, Foundational and Select Star. Eleven reach them partially, for example only for Tableau and Looker, or only for supported warehouse and BI pairs. OpenLineage and Marquez returned none found for BI-level lineage on that date.
Which tools offer metric-level lineage?
Metric-level lineage is rare. On 10 Nov 2026, nine of 16 tools returned none found: Alation, Atlan, Databricks, DataHub, Foundational, Manta, Monte Carlo, OpenLineage and SYNQ. Seven showed partial support: Coalesce Catalog, Collibra, dbt Labs, Euno, Secoda, Select Star and Snowflake. Euno's partial support is Tableau Pulse metrics, per Euno's docs.
Are there open-source data lineage tools?
Yes, two options in this ranking. DataHub is Apache 2.0 licensed, has 150+ integrations per its README, and scores 62.5. OpenLineage is an open standard and Marquez is its reference backend; both are Apache 2.0 and score 36.0 together. DataHub gives you a catalog and UI; OpenLineage gives you a format for collecting events.
What is the difference between OpenLineage and Marquez?
OpenLineage is a specification: it defines run, job and dataset events and facets that tools emit. Marquez is the open-source reference implementation that receives those events, stores them and serves a lineage API. Both are LF AI & Data Graduated projects. The two are scored as one entry on Upstream, at 36.0 out of 100.
What lineage does dbt provide natively?
dbt Labs offers dbt Catalog on the dbt platform, with column-level lineage built from SQL parsing of models. Per dbt's docs, column lineage reflects select statements only, not joins and filters, and Starter costs $100 per user per month. BI reach comes through exposures. dbt Labs scores 54.0, with its lowest marks on coverage of the wider stack.
What lineage does Snowflake provide natively?
Snowflake builds lineage from its own query history, shown in Snowsight and queryable with GET_LINEAGE and ACCESS_HISTORY (365 days). It requires Enterprise Edition or higher, and lineage before November 2024 for data movement is not available. Snowflake also accepts OpenLineage events. It scores 51.5, held back by stack coverage and agent access.
What lineage does Databricks provide natively?
Unity Catalog captures lineage automatically for queries on Databricks, down to the column level, and stores it in system tables kept for a rolling 395 days. It needs tables registered in Unity Catalog. Databricks documents no standalone lineage price, and it scores 47.5. Its transformation-level detail was none found on 10 Nov 2026.
How does Upstream score lineage tools?
Each tool gets a 0 to 10 mark on seven criteria, weighted and summed to a total on a 0 to 100 scale. Depth of lineage counts 25, Lineage that carries meaning 20, Served to agents 20, Coverage of the stack 15, Stays current without work 10, Safe to deploy 5 and Price transparency 5. Weights were fixed before any score was made. See /method/.
Why do vendor claims cap at 7 out of 10?
A claim that only the vendor makes cannot score above 7 on a criterion until a second source confirms it, such as the vendor's docs backed by a release note, a spec or a third-party read. This stops marketing copy from earning a top mark. A criterion with no evidence scores 0, and the page says none found on the date read.
What does none found mean on Upstream?
None found means we looked for evidence on a given date and did not find it, for example none found on 10 Nov 2026. It is not a claim that the feature is absent. A criterion marked none found scores 0, because a score needs a source. If you can point to a source, send it through /corrections/.
Who ranks first on Upstream, and why?
Euno ranks first with 67.0, then DataHub at 62.5 and Secoda at 61.5. Euno scored 7 on both of the heaviest criteria, Depth of lineage and Lineage that carries meaning. Euno is cloud SaaS only and publishes no dollar prices; a team that needs lineage inside its own network, or a full catalog around it, will pair it with another tool.
Why does DataHub lead on Served to agents?
DataHub scores 8 on Served to agents, ahead of Euno's 7, tied with Alation and Collibra. Its MCP server traces table and column lineage upstream or downstream, marks tools read-only, keeps mutation tools off by default, and uses each user's own token in shared deployments. The PyPI package reached 0.8.0 on 2026-10-21.
Which lineage tools publish prices?
Per the evidence read on 10 Nov 2026, dbt Labs lists Starter at $100 per user per month, and Coalesce lists Starter at $150 per user per month for Catalog and SYNQ (now Coalesce Quality). Monte Carlo publishes credit-based tiers. Atlan, Alation, Select Star and Collibra showed no price list. OpenLineage and Marquez are free under Apache 2.0.
Which lineage tools can I self-host?
DataHub open source and OpenLineage with Marquez are Apache 2.0 and self-hostable. Secoda offers single-tenant and self-hosted deployment on higher tiers. Monte Carlo lists self-hosted storage from its Scale tier. Euno describes itself as cloud-native SaaS, so no self-hosted option was found on 10 Nov 2026. Check each vendor for current terms.
Which lineage tools serve lineage to AI agents over MCP?
Hosted or published MCP servers with lineage tools appeared in the evidence for Atlan (41 tools), Alation, Collibra, Coalesce Catalog (16 tools), DataHub, dbt Labs, Euno, Foundational, Monte Carlo, Secoda, Select Star, Snowflake and SYNQ. They differ on scoping: Atlan scopes results to the user's persona, and DataHub marks tools read-only. Compare the Served to agents scores.
How often does Upstream re-read its scores?
Every fact carries the date it was read, and all current facts and scores were read on 10 Nov 2026. A weekly refresh adds dated news from primary sources, and a monthly refresh re-reads every price and claim older than 60 days. A score changes only when a re-read fact changed, and the old value, the new value and the fact are logged.
How do I report a correction?
Use the corrections page at /corrections/. Send the page, the claim you dispute and a source URL, ideally the vendor's docs or a release note. Corrections that cite a source are checked against it, and if the source holds up the fact, the score and the read date are updated. Vendors and readers are treated the same.
How were the scores made blind?
Scorers were given the rubric and an evidence packet in which every vendor name was replaced by a code, and URLs were removed. No vendor was identified as the subject of the work. A second scorer, also working from coded names, checked each score against the evidence. It changed 7 of 112 scores. Names were restored afterward.
Why do Snowflake and Databricks rank below catalog tools if their lineage is native?
Native lineage is strong inside its own platform, but the rubric also weighs BI fields, meaning, agents and stack coverage. Snowflake (51.5) scores 4 on coverage and Databricks (47.5) scores 4, because lineage stays within registered tables or the platform's own queries, with BI and external systems added in preview or by registration.
How do lineage tools collect lineage?
Four methods show up in the evidence. SQL parsing reads statement text, as DataHub and Foundational do. Query-log mining reads executed history, as Coalesce Catalog does over 30 days of SQL. Metadata artifacts such as the dbt manifest give project structure. Runtime events, such as OpenLineage, are emitted by jobs as they run.
Which tools run impact analysis before a change ships?
Foundational runs impact analysis in CI on pull requests and can block a merge, per its docs. Select Star offers a GitHub Actions app to prevent breakage, and Collibra traces downstream impact at table, column and report level. SYNQ exposes impact through a public API. Impact checks need accurate downstream links.
Which tools propagate tags or descriptions along lineage?
Atlan's Metadata Propagator pushes owners, certificates and tags, though tag propagation is off by default. DataHub propagates documentation and glossary terms, the latter in public beta on DataHub Cloud. Snowflake tags can inherit and propagate. Coalesce Catalog copies parent column descriptions. Euno says it carries Cyera classifications into BI fields.
Does a lineage tool need access to my table data?
Not necessarily. Euno states it reads metadata and query logs, not table contents. Foundational is metadata-first, and Secoda says it needs only read permissions on databases. Select Star says it will not access data values unless explicitly authorized. Check the setup grants for each vendor before you connect a warehouse.
How current is lineage in each tool?
It depends on the collection method. Euno syncs sources daily by default, SYNQ extracts lineage several times a day, and Monte Carlo says dbt context can take up to 24 hours. Databricks captures lineage automatically and keeps system-table history for 395 days. Snowflake began capturing in November 2024. Stale links are the usual failure.
Which lineage tools changed names or owners recently?
SYNQ is now Coalesce Quality after Coalesce acquired it on March 10, 2026. CastorDoc became Coalesce Catalog after March 19, 2025. Select Star's homepage says it is now part of Snowflake, and Secoda's site says Atlassian acquires Secoda. Manta now sits inside IBM watsonx.data intelligence. dbt Labs announced a merger with Fivetran.
Is data lineage the same as a data catalog?
No. Lineage is the map of how data flows. A catalog is a broader inventory of assets with search, owners and definitions, and many catalogs include lineage as one feature. Upstream ranks lineage only, so catalogs such as Atlan and Collibra are scored on their lineage, not on catalog breadth.
Which tools offer an API or SDK to query lineage?
SYNQ exposes lineage through LineageService and ImpactService in a public API. Alation's agent SDK has a lineage tool, and Snowflake offers GET_LINEAGE. Databricks offers a lineage API plus system tables. Euno publishes a Python SDK (v0.5), and Marquez serves a lineage API. Programmatic access matters most when agents consume lineage.

Where Euno is behind

Euno is cloud SaaS only and publishes no dollar prices; a team that needs lineage inside its own network, or a full catalog around it, will pair it with another tool.

Last reviewed 10 Nov 2026.