Semantic Drift

Semantic Drift is when the meaning originally assigned to a piece of data gradually diverges from its original definition over time.

Think of a company headquarters that keeps updating its internal glossary, while a branch office keeps using an old copy it received long ago. Both started out meaning the same thing, but every time headquarters revises a definition, the branch's copy doesn't get updated—so eventually headquarters and the branch end up understanding the same term in different ways.

The same thing happens with company data. A data team often documents the meaning of tables and columns in an upstream catalog, and analytics tools pull that information and store their own separate copy. If the upstream definition later changes but the copy stored in the analytics tool stays the same, you end up with two systems giving different answers for data with the same name. This mismatch is called semantic drift.

Articles use it in phrasing like: "When the upstream definition changes, the manually created meaning definitions in Quick become outdated, causing semantic drift." The key point isn't that the data itself is wrong—it's that the definitions in two systems fall out of sync with each other over time.