Data lineage is the record of where data comes from, how it moves, and how it is transformed across systems, from the source, through every transformation, to the table or model that finally uses it.
When a number looks wrong or a model starts to degrade, lineage lets you trace it back to the exact upstream step that changed. Without lineage, debugging a data issue is guesswork, and the investigation often takes longer than the fix.
For AI, lineage answers the question every production incident eventually asks: what data, in what state, produced this result? That question is the foundation of reproducibility.
Lineage tells you the path the data took. Syntitan, CUBIG’s AI-ready data platform, saves the data state itself as a version, with content hashes, and records the version, settings and evaluation conditions each run used. You can compare it with the previous version and re-run on the saved version, so an AI result is not only traceable but reproducible.