How Long to Keep It, and on What Basis
Building a retention schedule for leaver records: why one period for everything is wrong, and how to record the basis, not just the number.
Set different periods for different categories, and record why each one is what it is. A single retention period applied to the whole leaver file is the commonest approach and it is wrong in both directions at once — it keeps some things far longer than any purpose requires, and destroys others before the organisation stops needing them.
The recordkeeping discipline in “How Long to Keep It, and on What Basis” should also apply to workforce technology. When a team assesses this workforce tool in relation to gdpr employee monitoring, it should document purpose, access, retention and deletion, then preserve only the evidence needed for the handover, payment or review decision.
The reason to get it right is not abstract. A file destroyed too early cannot answer a question that arrives, and a file kept with no basis is a holding of personal data nobody can justify.
For an independent reference relevant to “How Long to Keep It, and on What Basis”, consult the ICO employment-practices guidance. Use it to test record quality, access, retention, fair process and exception handling against the organisation’s real departure workflow.
What drives each period
Four things, usually: a legal requirement to keep something for a stated time, a limitation period within which a claim could still be brought, an operational need such as pension or reference enquiries, and anything specific to the sector.
Those pull in different directions, which is exactly why one number cannot serve. Payroll records often have a statutory minimum. Process documents are driven by limitation periods. Pension-related material may need to outlive both.
| Category | What usually drives the period |
|---|---|
| Payroll and tax records | A statutory minimum, often several years |
| Contract and variations | Limitation periods for contractual claims |
| Disciplinary and process documents | Limitation periods for employment claims |
| Expired warnings | Usually a short internal period, stated in the policy |
| References given | Operational need, plus the person's rights |
| Pension and benefit records | Potentially decades |
| Health and safety or exposure records | Sometimes very long, by regulation |
What each period actually is differs by jurisdiction and sector and changes. That is a question for somebody qualified in the place concerned, and it is worth asking once, writing down, and reviewing on a schedule.
Recording the basis, not just the date
A retention schedule that lists categories and numbers is half a schedule. The other half is the basis — the requirement, the limitation period or the purpose that produced the number.
Without it, nobody can tell whether the period is still right when the law changes, and nobody can defend it when asked. With it, a review is an hour's work rather than a project.
Expired warnings
Warnings that have lapsed under the organisation's own policy are a special case and a common failure. A policy saying a written warning remains live for twelve months implies something about what happens afterwards, and in most organisations nothing does.
Decide what lapsed means: removed from the file, retained but marked expired, or retained for a stated further period for a stated reason. Relying on an expired warning in a subsequent process, because it was still sitting in the folder, is a recognisable error.
Keeping the whole file as one
It is tempting to treat the leaver file as a single object with a single date. It is simpler and it guarantees the wrong answer for most of the contents.
The workable compromise is a small number of groups rather than eleven individual dates: core employment record, process documents, payroll, and anything with a long statutory tail. Four dates is manageable and gets the answer approximately right, which is far better than one date that is definitively wrong.
Setting the date on the last day
The retention date should be written on the file at closure, not derived later by whoever is doing a clear-out. Derived dates are calculated from whatever date is easiest to find, which is usually the wrong one.
It should run from the termination date, which is why that date has to be recorded separately and unambiguously. A retention clock started from the last working day will be short by the length of any garden leave.
Retaining less in the first place
The cheapest retention problem is the record that was never created. Much of what accumulates in an employment file is duplicated, superseded or was never needed: draft letters, forwarded chains, three versions of the same scanned document.
Deciding what goes into the file is therefore part of deciding how long to keep it. A file assembled deliberately is smaller, more defensible and faster to search when somebody asks, which is three benefits from one decision taken on the last day.
Reviewing the schedule
Retention periods change, and a schedule written in 2019 is not self-maintaining. Put a review in the calendar annually, confirm the periods, and date the confirmation.
That dated confirmation is worth as much as the schedule itself. An organisation that can show it reviewed its periods last March is in a different position from one whose schedule has no date on it at all, and the difference costs one line a year.