How do you design a rollback/contingency plan for a cutover?

A rollback plan defines exactly how to revert to the old system if the cutover fails critically partway through — and critically, it must be decided and designed before the cutover begins, since improvising a rollback approach under the pressure of a failing live cutover is a recipe for making the situation worse.

flowchart TD
    A[Define go/no-go decision points in advance] --> B{Go/No-Go checkpoint reached}
    B -->|Go| C[Proceed to next cutover step]
    B -->|No-Go| D[Execute pre-defined rollback procedure]
    D --> E[Old system restored to known-good state]

Key elements include: clearly defined go/no-go checkpoints throughout the cutover timeline (not just one decision at the very end), a pre-tested technical rollback procedure (restoring the old system's data/state to a known-good point), explicit authority for who can invoke a rollback decision, and clear communication protocols for notifying stakeholders if a rollback is triggered — all of this designed and, ideally, rehearsed well before the actual cutover window, not invented in the moment.