The hours between stopping the old system and starting the new one — planned to the task, and rehearsed at least twice.
Cutover is the transition from legacy to live: final balances extracted, data loaded, systems switched, users let in. It is a sequenced plan with owners and timings for every task, a go/no-go decision point, and a rollback position. The work is not technically difficult; it is unforgiving of anything that was not rehearsed.
Cutover concentrates all the project's risk into a short window, usually a weekend, with the business unable to transact while it runs. Everything that has been approximate during the project — a load that takes longer than expected, a reconciliation that does not tie — becomes a problem with a clock attached. This is why rehearsal, not planning, is the control.
Legacy transactions stop at a declared point. Outstanding documents are completed or carried deliberately.
Closing balances, open orders, open payables and receivables, stock positions. Timed from rehearsal, not estimated.
Loaded balances tied back to legacy before anyone is let in. This is the step that most often overruns.
Against criteria agreed in advance, by named people. Deciding under pressure without criteria produces optimism.
Users in, hypercare team visible, and a daily issue triage for the first weeks.
A business schedules cutover over a long weekend: freeze Friday evening, migrate Saturday, verify Sunday, open Monday morning. The runbook has 140 tasks with owners and durations.
The dress rehearsal two weeks earlier produces the useful finding. The sequence works, but the total elapsed time is 41 hours against a 58-hour window — leaving far less contingency than the plan assumed, because three tasks took roughly twice their estimate. More importantly, the final reconciliation could not start until a load that finished at 2am, meaning the verification team would be checking financial control totals in the middle of the night.
Two changes follow, and neither would have been visible without rehearsing. The load sequence is reordered so reconciliation can begin on completed objects while later ones are still loading. And a formal go/no-go is added at the point where remaining work exceeds remaining window, with the rollback decision made there rather than at 6am Monday with nobody senior awake. Rehearsal's value is less in proving the plan works than in finding where it does not.
| Term | What it means |
|---|---|
| Runbook | The hour-by-hour task list for cutover, with owners, durations and dependencies. |
| Go/no-go | A decision point with defined criteria, taken before committing to the next stage. |
| Rollback | The plan for returning to the legacy system, which must be tested rather than assumed. |
| Freeze | The period when legacy changes stop, so the final extract is stable. |
| Dress rehearsal | A full timed practice of the runbook, in a realistic environment. |
A partial rehearsal validates the parts you were already confident about.
If the decision to abandon is made emotionally at 3am, it will be made badly. Define the triggers beforehand.
If balances do not tie, you need time to investigate — not fifteen minutes before opening.
Week one generates more questions than any other period of the project.
| Metric | What it tells you | What distorts it |
|---|---|---|
| Rehearsal elapsed vs window | Contingency, measured rather than assumed. Below about 30% is uncomfortable. | |
| Tasks exceeding estimate | Where the plan is optimistic, identified before it matters. | |
| Reconciliation completion time | Whether verification can finish before the business needs the system. | |
| Rollback tested | Yes or no. An untested rollback plan is a document rather than an option. |
We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.