Notice to Quit: Auditing a Service Before You Switch It Off
Photo: Loominade, CC0, via Wikimedia Commons
Closing something down attracts a fraction of the planning that launching it received. A switch-off date is agreed, a short note goes to users, and the work is treated as an administrative tidy-up rather than the reverse of a change programme. Then the plug is pulled and the dependencies appear: a team in another region, a statutory return, a supplier still invoicing against a contract nobody remembered, a spreadsheet that only one person could interpret.
Retirement is a change like any other, with the same requirements for ownership, communication, capability and evidence. The difference is that its failure modes are quieter. A badly launched service annoys people loudly. A badly retired one simply leaves gaps that are filled by workarounds and are not seen again until an audit or a complaint surfaces them.
The date is not the plan
The switch-off date is one line in a document. Around it sit obligations with their own clocks: contractual notice periods, statutory retention rules, the financial year end, licence renewal dates, and the point at which the last data extract must be taken. A retirement plan that begins with the date and works backwards only as far as the communications note is not a plan. It is an intention.
The first practical step is therefore a calendar of every date that constrains the exit, assembled before anything is announced. Where notice must be served on a supplier, and where that notice has a minimum period, the sequence is fixed by the contract rather than by internal preference.
Who is still using it
The assumption that a service is unused is almost always slightly wrong. Residual users fall into recognisable categories: small teams whose needs were never large enough to attract attention, external partners or public bodies receiving outputs, the person who built a local workaround around the system, and the successor process that quietly depends on an output nobody has traced.
The test is not to ask whether people use the service. It is to ask three narrower questions: what task does it perform for you, what will you do instead, and has that alternative been tried? Answers to the second question are frequently optimistic and answers to the third frequently absent. Where an alternative has not been tested, the retirement is carrying an unquantified risk.
Ownership, before and after
Every retirement needs a named owner who is accountable for the exit, not merely an IT contact who will action the switch-off. Separately, someone must own the residual risk after the date: queries from former users, archived data requests, contractual tail obligations and correspondence that arrives eighteen months later.
The most common failure is that these two roles are assumed to sit with the same person and, in practice, sit with nobody. Naming both, in writing, with a review point at ninety days after switch-off, costs very little and prevents most of the post-exit confusion.
Data: three decisions, not one
Data is usually treated as a single question, when it is three. What is transferred into a successor system or archive, what is retained under a statutory or contractual obligation, and what is deleted. Retention periods differ by record type: financial records, HR files, case notes, regulatory returns. The obligations may be sector-specific and are not always understood by the team doing the retiring.
There is a further practical test that is often skipped. Before the system is switched off, retrieve a sample of archived records through the route that will exist afterwards, and confirm that the record is readable, complete and intelligible. An archive that nobody can search is a storage cost, not a record.
Contracts and notice
Supplier agreements contain obligations that outlive the service: notice periods, exit assistance, data extraction in a specified format, licence tails, support during wind-down and sometimes penalties for late notice. These are frequently held by a commercial team that has not been invited to the retirement conversation. A short review of each relevant agreement, specifically for exit clauses, is a necessary step and not a legal nicety.
Communication that reaches people
Three audiences need to hear about a retirement: the people who use the service, the people who receive its outputs, and the people who will inherit the work. An intranet notice reaches almost none of them reliably.
A simple check is worth the effort: after the second communication, ask five randomly chosen users what is changing and when. If fewer than four can answer accurately, the communication has not landed and the switch-off date should be reconsidered.
What the next design should inherit
Every retirement should produce a short service biography: what the service was for, who depended on it, why it closed, what was discovered late and what would be designed differently. This is the cheapest form of organisational learning available and it is almost never captured, because by the time the work is done the team has moved on.
Better still, exit criteria can be written into the design of new services. A named owner for retirement, a defined data exit route, a scheduled relevance review and an indicative decommissioning cost, agreed at launch. Services designed with an exit are easier to run and considerably easier to end.
A pre-retirement audit in six questions
Who owns the exit, and who owns the residual risk afterwards? Which users remain, and have their alternatives been tested? What data is transferred, retained and deleted, and can each be retrieved? Which contracts require notice, and by when? Have the three audiences confirmed they understand the change? What will be recorded for the next design?
Six questions, answered in writing, before the date is fixed. The organisations that do this find that retirement is not the end of a service but the last stage of its management, and that the discipline applied to closing something well is the same discipline that makes the next thing easier to open.