The Rollback Ratio: Turning Irreversibility into Managed Risk

As action becomes cheaper and faster — through automation, platforms, and everyday “frictionless” choices — many failures shift from single catastrophic errors to compounding sequences of small commitments that are expensive, slow, or politically hard to reverse. This paper proposes the Rollback Ratio (1:N) as a portable heuristic for both people and machines: for every unit of committed exposure, hold (or invest in) N units of rollback capacity sufficient to recover to “okay” within an explicitly accepted time-to-okay (TTO) horizon. The framework treats reversibility as an asset and a budget, not a moral preference, and reframes “undo” as a multi-layer recovery problem spanning technical reversion, operational repair, and compensatory remedies. We connect personal domains — such as lottery spending, impulse purchases, and habit-driven escalation — to sociotechnical domains where autonomy scales faster than repair capacity, including agentic automation, product and pricing changes, privacy incidents, and operational disruptions. We define the construct, show how corridor-scoped autonomy and rollback budgeting operationalize 1:N in systems, and propose measurement primitives that make the ratio enforceable: rollback SLOs (restore/compensate latencies), receipt replay success, appeal resolution latency, and repair throughput under burst conditions. We discuss limitations (false precision, gaming, exception drift) and outline empirical calibration directions, including mapping TTO to required N across domains, testing ratio-based throttling to reduce ship–freeze oscillations, and evaluating whether receipt-backed contestability measurably improves recovery outcomes under dispute.