A familiar pattern
Consider a hypothetical programme that closes with every milestone complete. Some months later the new service is live, the delivery team has moved on, and the legacy system is still running because reports and an overnight file transfer depend on it. The organisation now pays for two estates.
This happens when closure criteria test the delivery of the new service and say nothing about retiring the old one. If acceptance does not require dependencies to be resolved, unresolved dependencies pass through closure unchallenged, and switching off the old system ends up with no owner and no budget once the programme stops.
Make decommissioning a deliverable
Map the dependencies of the system being replaced early, including reports, data extracts and manual workarounds that never appear on an architecture diagram. Agree the criteria for switching it off, give that work an owner and a budget, and keep it on the plan until the last server has gone. Until then, the planned savings remain unrealised while the legacy estate continues to incur hosting, licence and support costs.
Fund the running costs before the build finishes
Programme funding is usually time-limited, while hosting, licences, patching, support and upgrades continue for as long as the service exists. Estimate those costs during design, agree which budget will carry them and who will do the work, and test the support arrangement while the delivery team is still available.
Transfer knowledge throughout
Knowledge leaves with people, and contracted teams often leave together. Pair the organisation's own engineers with the delivery team from the start, record significant decisions as they are made, and rehearse operational procedures together. A handover week at the end of a long programme can transfer documents, while real understanding takes months of shared work.
Measure the improvement people notice
Milestones measure activity. Choose one or two outcomes that users or staff would recognise, such as how long an application takes to process or how quickly a fix reaches the live service, measure them before the work begins and after it ends, and report both. That gives the closure report evidence that will still mean something a year later.
Close only when the service owner is ready
Name the service owner at the start, involve them in every acceptance decision, and include the retirement of replaced systems in the closure criteria. The Government Service Standard asks teams to operate a reliable service, and that obligation continues long after the programme board has stopped meeting.
Working document: a decommissioning checklist
Start this checklist when the replacement is designed, so that dependencies are found while the people who understand them are still available.
| Field | What to record |
|---|---|
| System and accountable owner | The system being retired and the person accountable for its retirement. |
| Dependencies | Every report, extract, interface and manual workaround, each with an owner and its replacement. |
| Data | Retention requirements, what will be archived, where, and when the remainder will be deleted. |
| Contracts and licences | Notice periods and end dates for hosting, support and software. |
| Switch-off criteria | The evidence required before the system is made read-only and before it is removed. |
| Read-only period | How long the system remains available for reference and who can use it. |
| Final steps | Access removed, infrastructure deleted and costs confirmed as stopped. |
| Sign-off | Service owner approval and date, recorded in the programme closure. |
Use the supplier transition checklist, read about migration and modernisation, or discuss a service transition.
Further reading
Government Service Standard: operate a reliable service
