The Exit File: How UK Organisations Can Prove a Cloud Service Has Been Retired
Photo by Photo by Andrew Neel on Unsplash on Unsplash
Most UK cloud estates contain at least one service that no one actively uses but everyone continues to pay for. It may be a collaboration tool from a previous acquisition, a reporting platform whose sponsor left two years ago, or a storage tier attached to a project that has long since closed. The subscription line looks small enough to ignore, yet the service remains in scope for security reviews, backup routines, audit queries and licence renewals.
Retiring such a service is not simply a matter of sending a cancellation email. Deletion is a project with evidence. The aim is to reach a defensible position where the organisation can show that data was handled correctly, access was removed, dependencies were checked and the commercial tail was closed. In 2026, with UK regulators and customers asking sharper questions about data retention and third-party risk, that evidence matters as much as the saving.
Cancellation Is Not Decommissioning
A cancellation changes a billing status. Decommissioning changes the estate. Between those two points sit identity records, DNS entries, email routing, integrations, API keys, support contracts and the possibility that a live process still depends on the service in a way no one has documented.
The practical distinction is simple: cancellation is owned by finance or procurement; decommissioning is owned by a named individual in IT or operations, supported by the business owner. Without that ownership, the service tends to linger in a half-retired state, still discoverable by attackers and still visible to auditors.
Start With a Dependency Map, Not a Deletion Request
The first artefacts in the exit file should be a service owner, a business justification for retirement and a dependency map. The map does not need to be elaborate. It should record which users, groups, applications, data flows and business processes touch the service.
Useful sources include single sign-on logs, finance records, helpdesk tickets, network traffic, email gateway logs and the corporate directory. Where the service exposes an API, review the application registrations and service accounts associated with it. Where it holds data, identify any downstream reports, exports or integrations that consume that data.
This stage often reveals the awkward dependency: a scheduled report that feeds a regulatory return, a contractor who still authenticates through the service, or a webhook that quietly posts updates into a team channel. Finding these before cancellation is far cheaper than discovering them afterwards.
Data Export and Retention Proof
The data question has two parts. First, what must be preserved? Second, what must be deleted, and how will the organisation prove it?
For UK organisations, the retention position should be set against the UK GDPR, any sector-specific rules and the organisation's own retention schedule. The exit file should record the data categories held, the export format, the destination system, the date of export and the person who verified completeness. A simple checksum, record count or reconciliation report can turn an assertion into evidence.
Deletion proof deserves equal attention. Many providers offer a deletion request process but not a deletion certificate. Where that is the case, retain the request, the provider response, the date and any final confirmation. If backups or archives will persist beyond the service closure, document the retention period and the controls around restoration. A future subject access request or audit may require the organisation to explain exactly where residual data sits.
Identity and Access Teardown
Identity is where retired services most often leave a residue. The visible user accounts may be disabled, but application registrations, enterprise applications, OAuth grants, service principals, API keys and shared credentials can remain active.
The decommissioning checklist should include removal or disabling of the service from the identity provider, revocation of tokens and grants, deletion of service accounts, rotation or destruction of API keys, and confirmation that no break-glass or emergency access path remains. Where the service was federated, check whether the trust relationship has been removed at both ends.
Access removal should be evidenced with screenshots, export logs or configuration records. The point is not bureaucracy for its own sake; it is to show that the organisation did not leave a dormant door unlocked after the tenant was supposedly closed.
DNS, Email and Domain Records
DNS and email records are easily forgotten because they sit outside the service console. A retired service may still have a CNAME, MX record, SPF inclusion, DKIM selector or domain verification TXT record pointing to it. Those records can create security exposure, misrouted mail or confusing ownership long after the contract ends.
The exit file should include a DNS review before and after retirement. Check for service-specific subdomains, verification records, mail routing rules, auto-forwarding and any domain registration or renewal linked to the service. If the service used a dedicated domain, decide whether to retain, redirect or retire it, and record that decision. For email, confirm that no mailbox, alias, distribution list or journaling rule still depends on the service.
Commercial Tail and Licence Traps
Cloud contracts rarely end at the cancellation date. There may be notice periods, minimum terms, auto-renewal clauses, early termination charges, committed spend, support credits or data egress fees. Procurement should confirm the final billing date and obtain written acknowledgement that the contract is closed.
Licences need the same treatment. A retired service may have been bundled with a wider agreement, and removing it could affect pricing or entitlement elsewhere. Check whether any licence keys, support entitlements or partner agreements remain associated with the service. Record the closure in the software asset register so that future renewals do not resurrect the line item.
The Sign-Off That Closes the File
The final step is a short, signed record. It should state that the service has been retired, list the evidence held, confirm that no live dependency remains, and name the business owner and IT owner who accept the closure. Any residual risk, such as archived backups held by the provider for a fixed period, should be noted explicitly.
This sign-off is not a guarantee that nothing will go wrong. It is a reasonable, documented position that the organisation can defend. If a dependency is discovered later, the exit file shows what was checked, by whom and when, which makes remediation faster and accountability clearer.
A Practical Sequence
A workable order is: appoint an owner, map dependencies, export and reconcile data, obtain retention and deletion evidence, remove identity and API access, clean DNS and email records, close commercial tails, then sign off. Each step should produce a record. The exit file can be modest — a folder with a checklist, configuration exports, correspondence and sign-off — but it must be complete enough for someone outside the project to follow.
Retiring a cloud service is a small governance exercise with a long memory. The organisation that treats it as a cancellation email will eventually pay for the loose ends. The organisation that treats it as an evidence project closes the door, keeps the keys and can prove it did so.