The handover folder is complete. Architecture diagrams, release instructions and a recovery plan all have their own subfolder. Then somebody asks the receiving team to deploy a change without the supplier's help. The instructions refer to a role it cannot assume.
In this hypothetical example, document delivery passed. Operational acceptance has not yet been demonstrated. The distinction is easy to lose when a programme is closing and everyone would like the remaining work to fit inside a signature box.
Agree what is being accepted
Start with a service boundary and an identified revision. Name the application version, configuration and infrastructure in scope. State the environments, dependencies and procedures the receiving team will take over. A statement that 'the platform is accepted' is too broad if nobody can say whether the integration, production access or overnight processing is included.
Name the person who can accept the service and the people who must advise that decision. An engineer can confirm a test result. A service owner may accept the operating arrangement. A security authority may need to decide a specific residual risk. These are different acts, even if one person holds several roles.
The Government Service Standard's reliability point connects reliable operation to testing, monitoring and the ability to respond. The practical engineering task is to show how those arrangements work for this service.
Convert assurances into demonstrations
For each significant assertion, ask what the receiving team should be able to do. 'We have monitoring' becomes 'the right person receives a useful alert and can identify the affected service'. 'Rollback is supported' becomes 'the previous application can be restored under the stated data conditions'.
| Assertion | Demonstration | Evidence to retain |
|---|---|---|
| We can release | A receiving engineer releases the agreed change through the normal path | Commit, artefact identity, configuration, deployment result and observer |
| We can diagnose | The team investigates an agreed fault using its own access | Alert, timeline, useful logs and escalation route |
| We can recover | A scoped exercise reaches the agreed user task | Recovery point, timings, data checks, result and limits |
| We own the service | Access and responsibilities work without the outgoing team | Identity checks, named owners and remaining dependencies |
The table is an illustrative method, not a mandatory checklist for every service. A short internal tool and a nationally used transactional service need proportionate evidence. Both need clarity about what has actually been demonstrated.
Keep the evidence attached to the decision
Link to the result, not merely to a directory containing thousands of build files. Record which test ran, against which revision, in which environment, with what outcome. A successful pipeline from last month does not establish the behaviour of today's release.
Distinguish evidence produced by the supplier from a receiving team exercising the delivered procedure. Both have value. A supplier demonstration shows that someone knows how to operate the system. A receiving-team rehearsal tests whether that ability has transferred.
AWS's operational-readiness guidance describes a continuing review process informed by requirements and past incidents. The acceptance record is one decision within that lifecycle, rather than a permanent certificate that nothing else needs checking.
Treat remaining work as a decision
An open action does not disappear because it has a ticket. Give it an owner, a deadline, a consequence and an authorised decision about whether operation can begin before it closes.
Suppose a supplier-operated reporting feed has not been rehearsed. The service owner might accept limited operation while a manual route is available and a named person monitors the feed. Or the dependency may be essential, making the rehearsal a launch condition. Writing 'accepted with actions' without recording which of those decisions was taken leaves the next incident team to reconstruct it.
Use explicit outcomes such as accepted within a stated scope, accepted subject to recorded conditions, or not yet accepted. The organisation chooses the terminology. The record should explain its effect.
Rehearse without hidden assistance
Ask the receiving team to perform the work using the access and instructions it will retain. Keep the outgoing team available as observers. When it has to intervene, capture what was missing: permission, knowledge, an undocumented dependency or an incorrect instruction.
This is not a test of individual competence. It is a test of whether the operating arrangement survives the departure of the people who built it. Fix the gap, update the procedure and repeat the affected step.
Keep acceptance useful after the meeting
Record review triggers: a significant architecture change, a new critical dependency, a changed support arrangement or a recovery exercise that exposes a failed assumption. The next team needs to know when the previous decision should be reconsidered.
A practical record contains the scope, revision, evidence, decision maker, outcome, conditions and next review. Our downloadable illustrative acceptance record can be adapted to the service. It is a blank working aid; completing it does not independently validate the contents.
Operational acceptance is useful when the receiving organisation can explain what it now operates, what it can do unaided and which limitations it has chosen to carry. That is a stronger finish than a complete folder whose instructions nobody has followed.
See our delivery approach, use the recovery exercise record or the supplier transition checklist, or discuss a project.
