Align two snapshot comparisons by scope and observation basis — what should the operator record?
Two snapshots can have similar filenames and different scope. Before building a comparison, align their query revisions, selected objects and timestamp bases. The result may be a useful operational diff, but it should not be described as a consecutive-period trend or a full population change when the actual evidence only supports differences between returned sets.
Check window, identity grain and field definitions before describing changes between returned tables.
Record query_revision, requested_window, identity_grain, field_contract and receipt_time for each input. A mismatch in one dimension should remain visible rather than being hidden by renaming both tables baseline and current and assuming they are otherwise equivalent for the intended calculation.
Include elapsed observation interval and comparability exceptions with the output revision. Snapshots are not necessarily consecutive days, and a permitted comparison of observed values does not establish when the underlying source changed or that all relevant changes were captured between the receipts.Inventory comparison dimensions
Record query_revision, requested_window, identity_grain, field_contract and receipt_time for each input. A mismatch in one dimension should remain visible rather than being hidden by renaming both tables baseline and current and assuming they are otherwise equivalent for the intended calculation.
Choose the comparable projection
Select fields whose definitions and units agree, keeping source visibility changes separate from value changes. If membership differs, provide shared-key and exclusive-key views instead of performing field arithmetic across unrelated row positions or similarly spelled display names in the two artifacts.
Test a scope mismatch
Use labeled fixtures with identical keys but different metric definitions or requested windows. Confirm the checker blocks the unsupported comparison or labels it scope-incompatible even though the identifier join succeeds and both numeric columns contain plausible-looking values in every row.
Inspect actual differences
Classify added-to-returned-set, absent-from-returned-set, changed-observation and unavailable-for-comparison. Do not relabel these as new-to-platform, deleted or growth unless current authoritative evidence actually supports those broader source claims beyond the bounded local comparison performed here.
Release a bounded diff
Include elapsed observation interval and comparability exceptions with the output revision. Snapshots are not necessarily consecutive days, and a permitted comparison of observed values does not establish when the underlying source changed or that all relevant changes were captured between the receipts.
Keep the boundaries clear.
- Check window, identity grain and field definitions before describing changes between returned tables.
- Operational record: Inventory comparison dimensions. Context badges do not certify access.
Start with a read-only check: ask your agent to report actual source coverage before collecting records or exporting an audience.
Primary references
Documentation reviewed 2026-11-15. Product capabilities, client availability and plans can change. Third-party marks identify their products, not an affiliation.