Cutover and go-live: the night and what precedes it

What happens on go-live night is set out in go-live night — what actually happens. This station is about running it: the plan, the rehearsal, and the decision.

The principle governing all three: the night is run twice. Once as a rehearsal to the full clock, and once for real. Run it once and you discover your real timings while inside them.

The cutover plan

It is not a task list. It is a schedule by the hour, in four columns.

Time	Task	Owner	Measured duration
18:00	Stop entry in the legacy system	Operations manager	Immediate
18:30	Extract closing balances	Accounting	45 minutes
19:15	Load opening balances	Data team	3 hours 20 minutes
22:35	Reconcile the totals	Finance director	40 minutes
23:15	Point of no return	Sponsor	A decision

The fourth column is measured from the rehearsal rather than estimated. That is the entire difference between a plan that works and a plan of wishes.

Every task has a named owner, present that night, with their phone number in the plan itself.

The rehearsal

A cutover rehearsal is a full execution of the night’s plan on a separate environment, to the clock and with the same team.

Three things come out of it that come out of nothing else.

• Real timings. A load that “takes an hour” took an extra twenty-three minutes, which changes the whole night.
• Forgotten steps. Every rehearsal exposes three to five steps that were not in the plan.
• Who knows what. The one person who knows how to run a given step is identified in the rehearsal, and their work documented before the real night.

The rehearsal runs at least a fortnight before go-live, so there is time to act on what it exposes. Running it three days before is a ritual rather than a test.

The point of no return

A declared moment in the plan, by the hour, after which there is no going back.

It usually falls at the approval of the financial reconciliation: when the finance director signs that the opening balances are correct. After that, real transactions begin and going back means destroying real work.

An hour before it, a ten-minute decision meeting with three people: the sponsor, the finance director and the project manager. The question: do we proceed or roll back?

That meeting is not run on feelings. It is run against criteria written a fortnight earlier.

Rollback criteria

Written and approved before the night, because nobody writes an objective criterion at three in the morning after six hours of work.

The shape that works:

We roll back if any of the following occurs: a financial reconciliation difference unexplained within an hour, a schedule overrun of more than three hours before the point of no return, or a grade-one defect appearing in the purchase or sales cycle.

Three conditions make rollback actually possible.

• The legacy system stays operable even after reconciliation.
• The rollback plan is written and rehearsed, rather than “we restart it”.
• Whoever decides is named, and does not need a three-a.m. phone call for permission.

Rolling back is not failure. Failure is proceeding on balances that do not reconcile because nobody was willing to say stop.

The freeze before go-live

A fortnight before the night, a freeze begins — an administrative decision rather than a technical one.

Three things are frozen, each for a different reason.

• Configuration freeze. No setup change without the sponsor’s approval. Because every change invalidates a test that already passed, and there is no time to repeat it.
• Legacy master data freeze. No new items or customers in the old system after a published date, or your final load ships without whatever was added after it.
• Request freeze. Every new request goes to the post-go-live list without discussion. This is the freeze that always gets breached, because it is the only one people push back on.

The second is the one that gets forgotten and costs the night itself. A warehouse creates five items on Thursday, and the final load ran on Wednesday — so life in the new system begins with items missing, discovered on the first sales order.

Write all three freeze dates into the cutover plan itself, and announce them to everybody who enters data, not only to the project team.

The first week

Go-live does not end with the night. Three things are prepared before it.

One: presence on the floor. The team among people at their desks rather than in a meeting room. A question asked at the next desk is answered in a minute; the same question as a support ticket takes a day.

Two: a short morning meeting. Fifteen minutes every morning for ten days: what broke yesterday, what is the priority today.

Three: slack in approvals. Approvals pile up in week one because people are slow on a new screen. Reduce approval levels temporarily where possible, or assign somebody to chase them by the hour.

What you actually do in this station

• Write the night’s plan by the hour with owners, at durations measured in the rehearsal.
• Run a full rehearsal a fortnight before and record the timings and the missing steps.
• Fix the point of no return by the hour in the plan itself.
• Write and get the rollback criteria approved by the sponsor and the finance director in advance.
• Prepare the first-week plan: where from, when, and how escalation works.

The three commonest mistakes

One: a plan with estimated timings. The night runs three hours late and decisions get taken under exhaustion.

Two: cancelling the rehearsal for lack of time. It is the first thing cut, and the most regretted.

Three: rollback criteria held in people’s heads. Nobody ever rolls back, because declaring failure at three in the morning is a personal decision rather than an institutional one.

The acceptance test for this station

1. Show the night’s plan in its four columns with measured timings.
2. State what the rehearsal exposed in missing steps and timing gaps.
3. Fix the point of no return and explain why it sits there.
4. Show the rollback criteria, who approved them and when.
5. Show the first-week plan with presence, meeting and escalation.

What comes next

The system is live. The last station covers the two months that decide whether the project actually succeeded: hypercare and handover.