When sources change shape.
WHAT CHANGES OCCUR
Columns added Columns removed Columns renamed Types changed Meanings changed without any structural change
WHY THAT LAST ONE IS WORST
Nothing detects it, and every downstream number becomes wrong.
WHAT TO DETECT AUTOMATICALLY
The schema differing from what is expected.
WHAT TO DO ON DETECTION
Alert, and decide deliberately.
WHAT NOT TO DO
Automatically adopt the new schema without review.
WHY
An accidental change in the source silently propagates.
WHAT ADDITIVE CHANGES ALLOW
Proceeding, with the new column ignored or adopted.
WHAT REMOVALS AND TYPE CHANGES REQUIRE
A decision, since downstream consumers break.
WHAT A CONTRACT PROVIDES
An agreement with the source owner about structure and meaning.
WHY THAT MATTERS ORGANISATIONALLY
Most schema surprises are a communication failure, not a technical one.
WHAT TO ESTABLISH
Who to tell before changing a source.
WHAT TO VERSION
Your schema expectations, in version control.
WHAT TO TEST
That the pipeline fails clearly on an unexpected schema, rather than producing wrong output.
