← 🗂️ Data Science

🗂️ Lineage Tracer

Selected node: Total Revenue
Hops to source: —
Tables in lineage: —
Upstream share: —
Graph: —
FPS: —
Self-check
running…
The science (graph theory)

Model. Lineage is a directed acyclic graph G = (V, E); an edge u→v means “v is computed from u”. The ancestors of a node (its provenance) are everything that can reach it, found by breadth-first search over the reversed edges; its descendants are its impact. The hop number of an ancestor is its shortest-path distance to the selected node, which is what the depth slider limits.

Cost. BFS visits each node and edge once: O(|V| + |E|). The share shown is (ancestors in view) / (|V| − 1).

Independent checks. On load the BFS result for every node is compared with a Floyd–Warshall transitive closure (O(|V|³)), hop numbers with shortest-path lengths, and acyclicity with a Kahn topological sort.

Simplifications. Table-level lineage over 12 illustrative nodes; no column-level lineage, schedules or freshness; every edge propagates with probability 1. Same algorithm as the 2D twin, which uses a different 18-node example graph.

Drag — rotate · Scroll — zoom · Click a node — trace lineage

🗂️ Data Lineage and Governance

An interactive 3D directed graph of raw tables, staging models, mart joins, and dashboard KPI tiles — click any node to trace its lineage backward through every join and transform to the source table it ultimately depends on.

🔬 What It Demonstrates

Backward lineage tracing (a reverse breadth-first walk over dependency edges) is how governance tools answer "where did this number come from?" — collecting every ancestor table, join, and transform a metric depends on.

🎮 How to Use

Pick a dashboard metric or click any node directly. Use trace depth to reveal the backward walk hop by hop, and watch animated particles show which way data actually flows — from raw sources toward the dashboard.

💡 Did You Know?

Regulations like BCBS 239 and GDPR increasingly require column-level lineage — not just "this table feeds that table," but exactly which upstream column produced each output field.