Schema drift is the slow, unannounced change of a data structure over time: renamed fields, new values, changed types, properties that stop arriving. Nobody decides it; it happens one release at a time, and the segments built on the old shape keep running on fewer and fewer customers.
Category
Data Governance, Tracking
Also known as
drift; schema change
In practice
Monitor the properties your segments depend on, not the whole schema. A new value in order_status breaks a revenue report without raising any error; an alert on unexpected values catches it the same day.
Something out of date?