How Long

Working time records attract a retention period under the regime, usually counted in years from the period they relate to. Separately, payroll and employment records have their own periods, usually longer.

The workflow in “How Long” becomes more reliable when scheduled hours, actual time and later corrections can be distinguished. For teams exploring 7 minute rule payroll, this practical implementation page can add operational time and project context, provided data collection is proportionate, permissions are limited and every important exception receives human review.

The system that holds the detailed data has its own setting, chosen during the implementation by somebody weighing storage, and it is routinely shorter than both.

For an independent reference relevant to “How Long”, consult the GitLab remote-work handbook. Use it to test working-time definitions, recordkeeping, access, retention and exception handling against the organisation’s real process rather than treating one software report as conclusive.

The three sources and why they conflict

The regime specifies a minimum, frequently two or three years for working time records, with sector variations that can be longer.

The organisation's retention policy covers employment records generally, usually six years or similar, and may not mention working time at all.

The system's configuration defaults are typically shorter — ninety days of detail is common, with summarised data kept longer. That default silently governs whether anybody can answer a question about last year.

The specific thing that gets lost

Shift-level detail. Many systems keep weekly or monthly totals indefinitely and purge the individual punches after a few months.

That purge destroys the ability to compute rest, which depends entirely on start and end times. An organisation can therefore be fully compliant with its retention policy and unable to demonstrate daily rest compliance for any period more than a quarter old.

Checking what yours does

Ask for a shift-level export for a week eighteen months ago. Not a report — the underlying records.

If it comes back, the retention is adequate. If it comes back summarised, the detail has gone. If it comes back empty, the retention was shorter than anybody believed and the gap extends back to whenever the system was installed.

The averages are a record too

The rolling average for a person at a point in time is derived, which makes it tempting not to store it. It is also the figure that demonstrates compliance with an averaged limit, and recomputing it years later requires the underlying weekly data to still exist.

Storing the computed average alongside the weekly figures costs almost nothing and removes the dependency. It is also the form in which the question will be asked.

Opt-outs and the date problem

Agreements should be retained for as long as they are relied on plus the retention period, and the date signed is the field that makes them useful. An agreement with no date cannot be shown to have predated the hours it covers.

Where an organisation has old agreements with no dates, that is worth knowing now rather than at the point somebody asks, because the remedy — re-papering going forward — only works prospectively.

Backups, which outlive the policy

Data purged from the live system persists in backups until those age out. That is normally an acceptable position provided somebody can state the backup cycle.

It also means a deletion request or a retention claim needs the honest answer: removed from live systems on this date, present in backups until that date. Organisations that claim immediate and complete deletion are usually wrong about their own infrastructure.

Migration, where records actually disappear

The most common way working time history is lost is a system change. The new system imports current balances and summary data; the shift-level history stays in the old system, which is decommissioned eighteen months later.

Put historical working time data on the migration checklist explicitly, with a decision about whether it is migrated, archived in a readable form, or deliberately allowed to lapse. All three are defensible; losing it by accident is not.

The page to write

For each category of working time record: what the retention is, what the source for that period is, which system holds it, and what format it can be produced in.

Four columns, six or seven rows. On most sites at least one row has no source behind its retention figure, and it is usually the shift-level detail — which is also the row that matters most.

Deleting on request

A person asking for their data to be deleted presents a conflict: some of it must be retained for working time, tax and employment purposes, and some of it need not be.

Working out which is which in advance, in one paragraph, makes the response straightforward. Working it out during the statutory response window, while also establishing what the system can actually delete, is considerably less comfortable and tends to produce an answer that is defensive rather than accurate.