Dippr builds lineage by watching the queries that run on your engines, not by reading pipeline code or asking you to declare it. Every SELECT, INSERT, CREATE TABLE AS and similar statement that an engine executes is analyzed for which tables and columns it read and which it wrote. Those observed reads and writes are folded into a graph of tables and columns pointing at each other. Because it comes from execution, lineage is a lower bound. A table written by something that never ran through one of your engines, a warehouse job that queried the source system directly, or a dbt run that executed outside Dippr, leaves no trace. Dippr never hides that: every table and edge in the graph carries an honesty note saying whether it was actually observed, and how long ago.

Viewing lineage

Open a table or view from the catalog browser and switch to the Lineage tab. Dippr walks the graph from that table outward in both directions, up to a fixed depth, and draws what it finds. Lineage is only drawn per table or per column. Opening a catalog or a schema (rather than a specific table) shows an explainer instead of a graph, since there is no single object to anchor on. Opening a column shows the column-level graph for that column alone. An empty graph is a normal answer. It means one of two things, and Dippr does not distinguish between them on screen: nothing was ever observed reading or writing this table, or something was observed but you don’t have access to see it. Either way, data written by a system outside Dippr leaves no trace here.

Impact analysis

Below the graph, an Impact analysis table lists every downstream asset in this table’s observed closure: the tables and views that would be affected if this one changed, nearest first. Each row shows the number of hops from the table you opened, how many queries have read that downstream asset, how many distinct people have read it, and when it was last touched. If your access hides part of the true downstream closure, the panel says so.

What you don’t see

Lineage is filtered to what you’re allowed to query. A table you don’t have access to never appears as a node, and any edge that would have pointed to or from it is dropped along with it, so you can’t infer that it exists from a dangling connection alone. If access filtering removed anything from the view, the page tells you the view is filtered, without revealing how much was removed or what it was.

Exporting

From a table’s Lineage tab, Export OpenLineage downloads the graph you can see around that table as an OpenLineage-formatted file (NDJSON). The export applies the same access filtering as the graph on screen: it is your visible slice, never the whole system’s lineage. If the export had to stop early because there was too much to include, or because it filtered out data you can’t see, the page tells you before you rely on the file as complete.

What’s not captured

Dippr’s lineage is built entirely from queries it observed running on your engines. It does not capture:
  • Queries run directly against a warehouse or database, bypassing your Dippr engines
  • dbt runs, unless dbt executed its models through a Dippr engine
  • Anything that happened before your organization started using Dippr, or during a gap in monitoring
If a table shows no upstream lineage, that can mean it’s a genuine source table, or it can mean Dippr has never seen it written. The graph tells you which one applies for that table.

Browse the catalog

Where to open a table’s Lineage tab from.

Query history

The queries a table’s lineage is built from, with a detail page for each.