Docs
Review Schema Drift
Schema drift is the difference between what a source looked like at the last discovery and what it looks like now. Sources change without telling you: a column is added, a type is widened, something is renamed. Drift events make that visible in the catalog instead of letting it surface later as a failed run or a wrong number.
Where to Find It
Per source, as a list of drift events. Each one records when it was detected, which environment it was seen in, and what changed. The app lists drift events for the development environment today.
How Changes Are Classified
| Classification | What it means for consumers |
|---|---|
| Patch | Cosmetic. Nothing reading the object needs to care. |
| Minor | Additive. Existing consumers keep working; there is now more available. |
| Major | Potentially breaking. Something that already reads this object may stop working. |
The classification is about impact on whatever consumes the object, not about how much work the change was on the source side. A single renamed column is a small change that is nonetheless major.
Expand an event to see the individual column-level changes it carries, each with its own classification. That is where you find out whether a “major” event is one broken column or twenty.
What to Do With a Major Change
- Check what it affects with Trace Lineage .
- Correct the contract so it describes the source as it now is (see Review the Data Contract ).
- Re-run discovery and confirm no new drift event appears.
The Boundary, Stated Plainly
Amfelt detects drift, classifies it, and shows you exactly what changed. Acting on it is your decision, and you make it through the contract, which means the decision is reviewed and versioned like every other configuration change.
Next
Last updated on October 28, 2026