What's new: A focused set of improvements to Alation Lineage covering sources we used to miss, accuracy that keeps up with change, and lineage you can query programmatically and conversationally.
What it solves: People lose trust in lineage graphs when even one source is invisible, one stale link persists, or one hop is silently stitched through a gap. These updates close those cracks across file-based sources, cross-database movement, OpenLineage ingestion, version-aware maintenance, and lineage APIs.
Who it's for: Data stewards, governance teams, platform engineers, and anyone building agents that need to trace an answer back to its source. Available now for Alation Cloud Service customers.
You can have a lineage graph with a thousand nodes, and the one node that's missing is the one someone asks about during an audit.
I've spent the first few months in my role as the PM for Lineage talking to customers and digging through support cases, and the pattern is consistent. Teams don't lose faith in lineage because of some dramatic failure. They lose faith quietly. A dashboard that traces back to nothing. A view that changed last month but still shows the old connection. A graph that stitches through a gap so cleanly that nobody realizes a deleted table is being treated as if it still exists. Each of these is small. Together, they're why a steward opens the lineage tab, sees something that doesn't look right, and decides it's faster to ask the person who built the pipeline.
That's the failure mode we set out to fix. Not with a single new capability, but with a set of focused improvements across three areas: seeing the whole picture, trusting what you see, and working with lineage at the scale modern data estates demand.
When one missing source erodes trust in everything
Lineage is a confidence system. It either shows you the full picture or it doesn't, and "mostly complete" does not inspire trust in people who are making a change that could break a downstream report.
The problem is structural. Data doesn't only live in warehouses and BI tools connected by SQL. It lives in spreadsheets feeding Tableau workbooks. It moves between relational databases through custom scripts that no parser ever sees. It flows through Spark and Trino jobs that report via OpenLineage but resolve to the wrong catalog object because the storage path doesn't match how the source was cataloged. Each of these is a place where the graph quietly goes blank.
In a July 2024 Gartner survey of 1,203 data management leaders, 63% of organizations either did not have the right data management practices for AI or were unsure whether they did.¹ Gartner's prescription starts with metadata: organizations with basic or manual metadata management practices will struggle to make their data AI-ready.¹ Lineage is that metadata in its most practical form: if you can't trace where a number came from, you can't certify whether the data feeding an agent or a report is trustworthy.
Regulators see the same gap. In its November 2023 progress report, the Basel Committee on Banking Supervision found that only two of the 31 global systemically important banks it assessed were fully compliant with its risk data principles, and it named incomplete data lineage among the top obstacles.²
These aren't exotic edge cases. They're the gaps customers and support cases surface again and again, and they're now closed.
Alation lineage now delivers coverage that matches your unique data estate
File-based sources now appear in the graph. A Tableau workbook built on a CSV or Excel file used to be a dead end in the lineage graph, as if the report had no upstream source at all. Now, the file appears as its own node, and when that source file changes or gets retired, every report and dashboard that depends on it is visible. For cataloged file systems, lineage extends down to individual file attributes through the public APIs. No configuration required: file lineage is produced automatically the next time BI lineage extraction runs for any server that has file connections.
Lineage between relational databases. Alation can now build table- and column-level lineage between two RDBMS databases, whether data moves via queries or custom scripts, by matching object names across environments. An admin declares the upstream source in the target's Lineage Settings, and the system resolves lineage that previously produced nothing. This matters most for organizations with data movement that happens outside a tool Alation parses natively. This feature is generally available for cloud tenants today.
OpenLineage resolution fixes. Events from tools like Spark and Trino now match to cataloged sources using the richer identity most producers already send (AWS Glue, Hive Metastore, Trino, Athena, Redshift), not just the raw storage path. Events that were accepted but silently produced no lineage because the path didn't match how the object was cataloged now resolve correctly, no configuration required.
The net effect is fewer places where the graph goes blank. Impact analysis starts from the real origin of the data, not from wherever automation happened to reach.
A graph that stays current as the estate changes
Yet coverage alone isn't enough if the graph reflects last quarter's logic. A data estate changes constantly: views get rewritten, dbt models get edited, sources get decommissioned. Lineage that doesn't move with those changes is a liability, because a stale link is worse than a missing one. A missing link is at least honest about the gap.
Version-aware lineage ensures that changing a database view or dbt model source instantly replaces the old connection with the new one at both the table and column level. Before this, every edit to a dbt model, even adding a comment, could create a second connection, and every subsequent touch added another. A single model could accumulate many stale, conflicting links. That failure mode is now permanently eliminated going forward.
Clearer handling of temp, deleted, and uncataloged objects. A node that isn't a live cataloged object used to be ambiguous: "deleted" and "temporary" were blurred together, and lineage silently stitched through gaps so a broken pipeline could look perfectly connected. Admins can now opt in to handling that distinguishes a true temp table from a deleted object from one that was never cataloged, each with its own badge. Stewards can tell a genuine gap from a summarized one at a glance.
Time-based lineage filtering lets admins scope query-based lineage to a recent time window, so users see recently active relationships instead of the full history. Useful for cutting noise on busy graphs where years of accumulated query history obscure the relationships that matter today.
Impact analysis now matches the graph exactly. Impact Analysis and Upstream Audit trace through temporary objects to the permanent tables behind them and clearly label which table each affected column belongs to, in both the app and the CSV export.
The common thread is that the graph now maintains itself as the estate changes, rather than accumulating artifacts from decisions that no longer apply.
Lineage where the work happens
A purely visual lineage graph isn't enough for teams that need to analyze, report on, and automate their data workflows at scale. And as agents increasingly consume lineage to make decisions, lineage needs to be available programmatically, not just in a browser tab.
Lineage in Alation Analytics is now generally available. You can run impact analysis, upstream audits, and custom reporting over the entire lineage graph using the same analytics tooling you already use for other Alation metadata. New lineage, updates, and deletes all stay in sync.
Enhanced lineage APIs let you filter by direction, exclude temporary and deleted objects for UI parity, and perform bulk cleanups by pattern-matching external IDs instead of pulling everything and filtering on your side. One change to note: the lineage API now excludes deleted objects by default. If your integration relied on them appearing, pass excludeGone=false to restore the previous behavior.
Lineage Co-Pilot, now in open beta, meets you where the question actually is. When a lineage graph has nothing to show, a Diagnose Lineage action asks the co-pilot to investigate why. It walks the ingestion pipeline, job status, and configuration, then reports what it finds in plain language. On any graph, you can also ask questions in context, from "what uses this table?" to "how does query-log ingestion work?" The co-pilot diagnoses and suggests next steps; it doesn't re-run jobs or change anything on its own. To try it, turn on the feature flag in your lineage settings (Alamigo must also be enabled).
And for teams building agents, Agent Studio includes two ready-made lineage agents:
The Lineage Agent lets any user trace upstream sources and downstream consumers, follow column-level transformations, and save lineage documentation to the catalog.
The Lineage Builder Agent helps admins create and delete manual lineage and diagnose why lineage is missing, and it asks for approval before making any change.
Both are built on lineage tools, including Get Lineage, so the agents you build can check upstream sources and downstream consumers before they act. An agent that can trace an answer back to its source is an agent you can put in front of users. That's why lineage is built into AIOS, Alation's open, governed, self-improving intelligence operating system. Learn more here.
Common questions
Do I need to turn these features on?
Some are on by default, including lineage in Alation Analytics and file-to-BI lineage. Others are opt-in so you can adopt them on your own schedule. The Community post lists the setup steps and any prerequisites for each.
What about lineage I already have? Does this fix existing issues?
Most of these updates apply going forward. Version-aware lineage, for example, keeps new changes clean from here on. If you have historical duplicate links from dbt model edits or stale view connections, your Alation account team can help work through a cleanup.
How does this relate to Critical Lineage?
Everything in this post is part of core Alation Lineage and applies across your connected sources. If your challenge is regulatory, tracing a specific metric in a filing back to its source with an accountable chain of ownership, Critical Lineage in Critical Data Manager builds on the core lineage graph and adds the governance layers regulators expect.
Can Alation build lineage between databases without a parsed ETL tool?
Yes. RDBMS-to-RDBMS lineage matches objects by name across two systems once an admin declares the upstream source. The important caveat: getting the upstream setting right matters, since it determines where lineage resolves. QLI-based lineage requires the same database type on both sides; script-based lineage requires matching table names.
Available now
These improvements are available for Alation Cloud Service customers on the cloud. To see what's enabled in your instance, check your lineage settings or reach out to your Alation account team. For the full feature-by-feature detail, including setup steps and known limitations, head to the Community post What’s New in Data Lineage.
And keep the feedback coming. Much of this work started with it.
Sources & notes
Every external claim in this post is independently verifiable.
63% of organizations either do not have or are unsure whether they have the right data management practices for AI; survey of 1,203 data management leaders, July 2024. Organizations with basic or manual metadata management practices will face challenges making their data AI-ready. — Gartner, "Lack of AI-Ready Data Puts AI Projects at Risk," 26 February 2025 https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk
Of 31 global systemically important banks assessed, two were fully compliant with all BCBS 239 Principles; several banks still lack complete data lineage. Assessment data as of June 2022. — Basel Committee on Banking Supervision, "Progress in adopting the Principles for effective risk data aggregation and risk reporting," 28 November 2023
https://www.bis.org/bcbs/publ/d559.pdf
- AI
- Alation News
- Data Quality
- Data Intelligence
- Data Governance
- Modern Data Stack
Madhur Garg is a dynamic product leader with over 8 years of experience. Handling all things Lineage at Alation. Contact him at madhur.garg@alation.com.
