Dress Rehearsal
A dress rehearsal is a full run of the cutover runbook, executed end to end at production-like volumes by the people who will run the real one. Its output is measured timings, a list of defects and corrections to the runbook. It is used to establish whether the transition fits the available window before the window is committed.
The word rehearsal is accurate. The point is not to prove the technology works — that is what testing does — but to prove the sequence, the timings, the handovers and the people work together under the conditions of the real thing. A rehearsal that skips the awkward parts because they take too long has not tested the thing most likely to fail.
What it produces
| Output | What it gives |
|---|---|
| Measured durations | Actual elapsed time per step, replacing the estimates in the runbook |
| Critical path and contingency | Total elapsed time against the available window, and how much slack is genuinely left |
| Runbook corrections | Missing steps, wrong ordering, unstated prerequisites, steps at the wrong grain |
| Defect list | What failed, with severity and a fix owner, tracked to closure before the next rehearsal |
| Reconciliation results | Record counts and control totals against tolerance, using full-volume data |
| Role confirmation | Whether the named owners can actually perform their steps, and where cover is missing |
| Evidence | Screenshots, logs and sign-offs used later as readiness evidence at the go/no-go |
| Decision on repeat | Whether another rehearsal is needed, and what has to change first |
Most programmes run more than one. An early rehearsal establishes feasibility with known gaps; a final one is executed under the real constraints — same window length, same rota, same freeze — and produces the timings used in the plan. The Cutover Runbook & Hypercare Pack records planned and actual times per step, which is what makes rehearsal output usable.
How it is used
The rehearsal is scheduled far enough before the transition that its defects can be fixed and, where the fixes are material, re-rehearsed. Working backwards from that, it usually sits before the final go/no-go and after the last significant build.
Its results are also evidence. Readiness criteria that reference a completed rehearsal — sequence executed end to end, timings inside the window, reconciliation within tolerance — are answerable with fact rather than opinion. Data migration rehearsals in particular are covered in the data migration planning guide, where full-volume dry runs and reconciliation tolerances are the substance of the exercise.
Where it goes wrong
Rehearsing with a data subset is the most common compromise, and it removes the only thing that mattered — how long the load takes at real volume. The second is rehearsing with a different team, usually because the people who will run the live cutover cannot give up a weekend twice.
The third is not recording actual times, which reduces a rehearsal to a pass or fail. The fourth is treating the rehearsal defect list as informal: items are noted, the rehearsal is declared successful, and half the fixes are still open at go-live because they never entered a tracked log.
The fifth is a rehearsal held too late to act on. If the only full run happens the week before the transition, the choice on finding a serious defect is between a delayed go-live and a waiver, and both decisions are taken under pressure. Scheduling the final rehearsal with room for a fix and a partial repeat is what keeps the result usable rather than merely informative.
Related terms
See also cutover runbook, smoke test and go/no-go decision.
Questions
How many dress rehearsals are needed?
It depends on what the first one shows. A common pattern is an early rehearsal to establish feasibility and a final one under real constraints, with fixes tracked between them.
What is the difference between a dress rehearsal and a dry run?
The terms overlap. A dry run often refers to one activity — commonly the data migration — while a dress rehearsal means the whole cutover sequence executed end to end.
Does a rehearsal need production volumes?
Timings taken from a subset do not transfer. Full-volume runs are what establish whether the sequence fits the window, which is the main question a rehearsal answers.
What is recorded during a rehearsal?
Actual start and finish times per step, defects with owners, reconciliation results, and corrections to the runbook itself.