Work TimeTimekeeping · Monitoring · Evidence

Timekeeping & the Law

Correcting Time Records: The Process That Keeps You Out of Trouble

Section
Timekeeping & the Law
Written
2026-08-06
Last checked
2026-08-06
Law and programme rules in this area change. This article states the position at the time of writing and is revised when it moves. It is general information, not legal advice.

Time records need correcting. People forget to clock in, systems fail, a break gets interrupted, someone works from home and records it three days later. None of that is a problem.

For further product context on online timesheet corrections, consult here.

The governing federal recordkeeping text is available in the federal recordkeeping regulations in 29 CFR Part 516.

What creates a problem is a correction process that is hard to use, invisible in the record, or that people have learned not to use.

The principle

Corrections should make records more accurate, and the fact that a correction happened should be visible.

A record that has been silently altered is worth less than one that shows an original entry, a correction, an author, a timestamp and a reason. In a dispute, a clean audit trail is evidence of a functioning process. A record that could have been changed by anyone at any time without a trace is evidence of nothing.

What a defensible correction contains

  • The original value
  • The corrected value
  • Who requested the change
  • Who approved it
  • When it was made
  • Why

The last one is the one systems most often leave optional and employers most often skip. "Corrected" is not a reason. "Forgot to clock out; left at 5:30 per building access log" is.

Never overwrite without a trail. If your system permits silent edits, that is a configuration to change and, if it cannot be changed, a reason to change systems. See choosing a timekeeping system.

Who may correct what

The employee should be able to request a correction to their own record. Making them ask a manager for every forgotten punch adds friction, and friction is what produces incomplete records.

A manager approves. That approval is an attestation, and managers should know what they are attesting to. See timesheet approval.

A manager should not be able to reduce recorded hours unilaterally, without the employee's knowledge, and certainly not without a trail. This is the highest-risk permission in any timekeeping system, and it is frequently granted by default.

Payroll or HR handles retrospective corrections beyond the current period, with documentation.

Timing

Same period is routine. A correction inside the current pay period, before approval, is ordinary business.

Prior period requires more. Documentation of the reason, and a payroll adjustment. Do it — an employer who declines to correct a prior period because it is inconvenient is choosing to leave an underpayment in place.

Long after the fact still requires payment. If unpaid hours surface a year later, they are owed. The correction is the cheap part; the exposure is the pattern that produced it.

The signal to watch

A near-zero correction rate is a warning, not an achievement.

In any real workforce, punches get missed, breaks get interrupted, and remote work happens outside recorded hours. A team of thirty people over a month with no corrections at all is not a team with perfect records. It is a team whose correction process nobody uses.

Reasons people stop using it:

It is cumbersome. Three screens and a manager conversation for four minutes of work means the four minutes go unrecorded.

It draws attention. If corrections are treated as errors and errors are treated as a performance issue, people will avoid them.

A manager discouraged it. The most damaging cause. "Don't worry about it this time" produces unpaid work and teaches the employee not to raise it again. See off-the-clock work.

They were told corrections cost the team. Where a manager is measured on overtime spend and nothing else, this is a rational thing for them to say and a structural failure on the employer's part.

Track the correction rate by team. A team well below the organisational average is worth a conversation with the manager, not with the employees.

Corrections that are not corrections

Two patterns that look like corrections and are something else:

Systematic downward adjustment. Corrections that overwhelmingly reduce hours, in one team or under one manager, are a pattern rather than a series of errors. This is what an investigator looks for, and it is easy to find.

Bulk edits. Multiple records changed in one action, with one reason, is not a correction process. It is a payroll adjustment and it should be documented as one.

Making it work

Make requesting a correction take under a minute, from a phone.

Allow employees to see their own records and flag issues, without asking anyone.

Set a routine. Employees review their own time weekly; managers approve on a fixed day. Corrections found in a routine review are cheap.

Say plainly that corrections are expected. Not tolerated — expected. People need to hear that recording an extra twenty minutes is the correct behaviour, not an admission.

Audit the trail quarterly. Sample corrections and check that reasons are meaningful, that the direction is not systematically one way, and that no team is at zero.

The check worth running

Pull every correction from the last quarter and sort by direction and by manager.

If corrections cluster under one manager, or run predominantly in one direction, you have found something. It is very unlikely to be a coincidence, and finding it internally is considerably better than the alternative.