From APO to IBP: Transition Design, Parallel Planning Runs and Planning-Result Testing

Keep this: job aid

Parallel-run comparison with tolerances

• Agree tolerances per key figure and level before the first run; write them into the test plan.
• Freeze one input snapshot (history, stock, orders, master data) and record its timestamp.
• Reconcile loaded record counts in both systems before comparing any plan.
• Compare top down: network total, plant, product family, top products.
• Classify every exception: input difference, design change, engine difference or defect.
• Log each planner decision with reason and evidence; never widen tolerances after seeing results.
• Only one system feeds the active PIR version; the other releases to a comparison version.
• Exit per stage: two signed cycles, PAT accepted, no severity 1 or 2 defects, cutover rehearsed.
• After switch-off: CIF queues empty, no active models to the old system, one PIR set reconciled.