A cutover is planned for a Sunday night. What must be prepared beyond the technical steps, and what is the most commonly missing item?

What must be prepared

• A runbook executable by whoever is available, not by the person who wrote it. No implicit knowledge, no steps requiring one individual's access, and every command written out.
• A rollback that has been exercised, with its own runbook and its own decision criteria. A rollback documented and never tested is a hope.
• Explicit go/no-go criteria checked before starting, and abort criteria checked during. Both agreed in advance, because judgement under time pressure at 2am is worse than judgement made calmly.
• A named decision-maker with authority to abort, present for the duration.
• Verification steps with expected results, so "it worked" is a check rather than an impression.
• Communication prepared: what customers are told, what support is told, and what the internal status update says at each stage.

The most commonly missing item

The point of no return, identified explicitly. Every cutover has a moment after which rollback becomes impractical — usually when writes have accumulated in the new system, or an external party has been switched.

Knowing where that point is, and what the criteria are for crossing it, is the single most valuable piece of cutover planning and it is routinely absent. Teams discover mid-cutover that they cannot go back, which converts a controlled operation into an incident.

The second most missing

What happens to in-flight work. Requests submitted before the cutover and completing after it, scheduled jobs due during the window, external callbacks arriving mid-switch. These need a defined behaviour rather than being discovered.

The insurance-specific consideration

A policy or claim in progress across the cutover is subject to rules and data on both sides. The plan must say whether such records move mid-lifecycle or complete on the old system — and if the latter, the coexistence period extends to the longest lifecycle, which may be years.

That decision determines the shape of the whole migration and is frequently made implicitly, which is how an eighteen-month plan becomes a four-year one.