Cutover and Go Live
Cutover is the execution of the switch: a sequenced plan run against the clock, with checkpoints, a command structure and a defined point beyond which there is no going back. It is a logistics exercise more than a technical one.
If you change this
It implies 6 topics across 3 modules, including Data & Analytics and Integration which you might not expect.
See the path ahead →Why it matters
Most cutover failures are not technical. They are a task that took three times as long, a dependency nobody sequenced, a person who was unreachable, or a decision nobody had the authority to make at 3am on a Sunday.
- Sequence is the whole design: Data cannot load before configuration, interfaces cannot run before data, users cannot be released before validation.
- The window is fixed: Contingency has to be in the plan, not in hope.
- People are the constraint: A forty-hour weekend needs shifts, handovers and someone awake.
The workflow
- 01
Freeze
Changes to the legacy system stop; the cut-off is communicated and enforced.
- 02
Final extract
Data taken from source at a known point, with counts and control totals.
- 03
Load and reconcile
Data loaded to target and reconciled against those totals before proceeding.
- 04
Configuration and access
Production configuration confirmed and access provisioned from the position data.
- 05
Interface enablement
Feeds switched to the new system in dependency order.
- 06
Business validation
Scripted transactions run by the business, and signed.
- 07
Release
Users given access, legacy access removed, hypercare begins.
Set the point of no return explicitly, with a time and an owner. The worst position in a cutover is discovering a problem when rollback is no longer viable and nobody had decided that it had passed.
What you need in place
- A minute-level plan for the critical path: Everything else can be hourly.
- Control totals agreed before extract: Reconciliation needs a number both sides accept.
- A command centre and a contact tree: Including escalation to someone who can decide.
- Shift cover: Nobody makes good decisions at hour thirty.
- Legacy access removal: Dual running by accident is how two ledgers diverge.
- A communication plan: Users told what is unavailable, when, and what to do instead.
Questions to ask
01Critical path
What is on it, and what is the contingency?
If contingency is zero, the plan is a forecast.
02Reconciliation
What totals must agree before we proceed?
Agreed beforehand, or the argument happens at 2am.
03Point of no return
When is it, and who declares it?
It should be a scheduled decision, not a realisation.
04Authority
Who can make a call during cutover without convening anything?
Someone must, and everyone must know who.
05Legacy
When does legacy access stop?
Leaving it open guarantees transactions in the wrong system.
06Failure
What is the plan if a load fails halfway?
Restart points should be designed into the sequence.
Metrics and KPIs to track
Plan adherence
Tasks completed within their planned window.
Critical path slip
Cumulative variance against the window.
Reconciliation exceptions
Items not agreeing between source and target, and their value.
Defects at go-live
By severity, raised during the cutover window.
Time to first transaction
Release to first successful business transaction.
Alongside this