Before
Analysts compared spreadsheet cells, recalculated weekly changes, searched release notes, and rebuilt analytics views to explain movements. They then assembled the evidence into a readout or manually handed unresolved discrepancies to the data owner.
How it works
- Read the weekly KPI register and recent release notes, then validate each record and calculate its signed change.
- Compare each metric with its own tolerance and investigate out-of-range values using relevant analytics segments and supporting evidence.
- Reuse or create a verified weekly brief, or route ambiguous and unresolved comparisons to the data-owner review queue.
Workflow details
- Batch intake: A manually started run reads the configured weekly KPI register and selects valid records containing current and prior values, reporting periods, tolerances, and native units.
- Evidence gathering: The workflow retrieves recent release notes and creates shared context for the batch, then processes each KPI independently using its signed change and tolerance.
- Automated investigation: Metrics outside tolerance are investigated with a read-only analytics report for the relevant weeks, while invalid inputs, failed lookups, ambiguous matches, or uncertain actions are marked for human review.
- Brief management: The workflow searches for an existing readout using the exact metric and period, reuses a unique match or creates a dedicated weekly document, records the baseline, current value, change, release evidence, and next check, and reads the document back for verification.
- Controlled handoff: Confirmed documents are returned as completed artifacts; unresolved comparisons are deduplicated against the review queue and opened as a data-owner decision card when needed.
What changed
Each weekly metric now receives an automatic tolerance check and evidence-based disposition. Confirmed results are written and verified in a dedicated brief, while uncertain cases are routed to the data owner with the source reference and decision context instead of requiring a separate manual handoff.
Work removed
Recalculating KPI differences, checking each tolerance, searching release notes, investigating relevant analytics segments, locating prior readouts, copying evidence into documents, verifying saved briefs, and manually creating duplicate-free review handoffs.
Expected outcome
For the two-metric sample batch, the workflow can remove at least two manual difference calculations and one manual evidence-assembly pass, while preserving human review for invalid, ambiguous, or unresolved cases.

