Trigger and source data
- Start from a scheduled query performance pull or page indexing change
- Capture a realistic sample payload before designing transformations
- Decide which fields are required, optional, or safe to ignore
- Store raw inputs long enough to debug early runs