Schema Drift

Schema drift is the gradual, often unintended change of a data's structure over time: a property's type changes, a new value appears or a name shifts, without anyone noticing.

Schema drift is more dangerous than an obvious error, because it happens silently: it raises no error, it just gradually becomes wrong. Its source is usually a lack of coordination — a new app version sends amount as a string instead of a number, or a team adds a new value outside the allowed set, or an event's name changes in code. Each is small on its own but together they pull the schema away from what the reports assume. The key difference from a deliberate change is that drift is unplanned and invisible; so its only antidote is continuous validation and change monitoring, not a good initial definition. Whoever reads the data should know: when a metric's trend breaks for no reason, the first suspect is schema drift, not a real change in user behavior.