Correcting a Week That Has Been Approved
Approved does not mean right. What a correction has to preserve, who has to agree to it, and why overwriting is the one thing never to do.
Correction, period 7
Actually reviewed
Submitted
Original 37.5h, project A
Approved
Corrected 22h A, 15.5h B
Time taken
11 days later
Requested by employee, approved by both leads · Both versions retained.
The week as submitted, and as it happened
M
T
W
T
F
S
S
As submittedAs worked
Errors surface after approval constantly: a code that was wrong, hours booked to a closed project, a day attributed to the wrong week, an absence that should have been recorded. The approval does not make them right, and the process for fixing them is usually undocumented.
The approval issue in “Correcting a Week That Has Been Approved” becomes easier to diagnose when the record shows both the submitted hours and the operational context around them. A team evaluating Monitask for interview reimbursement policy should define what an approver must actually check, how a disputed entry is returned and which activity signals are context rather than proof that the work occurred.
What matters in a correction is not the new figure. It is that the old figure, who asserted it, and why it changed all survive, because the question asked later is almost never "what are the hours" and almost always "why does this document say something different from that one".
For a separate benchmark relevant to “Correcting a Week That Has Been Approved”, consult the Xero accounting resources. Use it to test record quality, approvals, retention, employee rights and exception handling against the real workflow rather than treating a software report as self-explanatory evidence.
What a correction has to preserve
The original entry as submitted, with its timestamps and its approver. The corrected entry. Who requested the change, who authorised it, when, and a reason.
A system that overwrites gives you none of this. It gives you a record that now says the right thing and has no memory of having said anything else, which means any report, invoice or claim produced from the earlier version can no longer be explained.
This is the most common serious defect in timesheet administration. It is invisible until somebody compares a historical report against the current data and finds they disagree, at which point nobody can account for the difference.
Who has to agree
The employee, where the correction changes their hours or their pay. A correction made to somebody's record without their knowledge is a problem in itself, regardless of whether the new figure is more accurate.
The original approver, or their successor. If a different person authorises the change, the record should say so rather than appearing to be a second act by the original approver.
Any downstream owner whose output has already gone out. If the hours have been invoiced, claimed or paid, the correction has consequences in those systems and the correction process has to trigger them rather than assuming somebody will notice.
Correction versus adjustment
Two different things that get conflated. A correction says the original record was wrong: the hours were not as stated. An adjustment says the original record was right and something else changed: a rate was wrong, a code was retired and work moved, a cost was reallocated.
Keep them apart. A correction affects what the record asserts about the work and needs the employee's agreement. An adjustment affects how the organisation accounts for it and generally does not. Systems that offer one mechanism for both produce records where it is impossible to tell whether the hours changed.
The window
Most organisations need a point after which the period is closed and corrections follow a heavier route. Without one, records change indefinitely and no report is ever stable.
A workable structure: free correction until the approval deadline, correction with approval until the period closes in finance, and after that a formal adjustment with a documented reason and a named authoriser. Three tiers, each with a clear owner, and the boundaries aligned to the finance calendar rather than invented separately.
Volume as a signal
Count corrections per period, by type and by team. It is one of the few leading indicators of record quality that is cheap to compute and genuinely informative.
A rising correction rate in one team usually means a code problem or a new project whose structure nobody understands. A high rate concentrated on one code means that code is ambiguous. A high rate immediately after approval means the approval is happening before the person has finished, which is a deadline problem wearing a different costume.
The thing to insist on in any system
Version history on entries, retained for as long as the records themselves. It is the single technical requirement that makes everything in this section possible, most products have it, and a significant number of implementations have it switched off or purged on a short retention because it makes the database larger.
Turning it on costs storage. Not having it costs the ability to explain your own records, and that cost arrives at the worst possible moment.
Correcting the record of somebody who has left
The hardest case administratively and the one most likely to be handled badly. The person has no account, often no access to any evidence, and frequently no interest in cooperating.
The requirements do not change: the original stands, the correction is recorded with its reason and its authoriser, and the person is told if it affects what they were paid. Where it reduces a payment already made, the recovery rules apply with more force rather than less, and an approach to a former employee about an overpayment should be in writing and prompt. The cases that become disputes are the ones raised late with no explanation of how the error arose.