Predict a proposed change's cost impact, pick the lowest-error source for its change type, and show the result with its value-ratio effect (R4). The ledger is cold-start by design — no synthetic back-test is seeded — so every prediction here is an honest estimate until realized changes accumulate.