Three Systems That Disagree

Most operations hold the same day in three places. The rota says who was supposed to be there. The attendance or absence record says who was. Payroll says who was paid for what.

The record in “Three Systems That Disagree” becomes more useful when the original time entry, later correction and reviewer decision remain distinguishable. If Monitask resources for download time tracking software supports download time tracking software, administrators should test exports, amendments, permissions and retention before launch so a dashboard does not replace the underlying evidence.

They disagree, routinely and in both directions, and because no one reconciles them the disagreements persist as quiet errors: people not paid for cover they worked, absences never recorded, agency shifts invoiced against a slot somebody internal also filled.

For an independent benchmark relevant to “Three Systems That Disagree”, consult the UKRI award-management guidance. Use it to test scheduling, working-time limits, attendance records, employee rights and exception handling against the real operation rather than treating a software report as self-explanatory evidence.

Where the disagreements come from

Cover arranged informally. Somebody swaps, stays late or comes in, and the rota is never amended. Payroll pays the original pattern and the person chases it, or does not.

Absence recorded in one place only. The shift record shows a gap; the HR system shows nothing, because the person who filled it in was covering a shift.

Rota changes after publication that reach scheduling and not payroll.

Agency bookings recorded by the agency and not by the operation, so the invoice is the only record that the shift happened at all.

The reconciliation

Monthly, by person, for one site: rostered hours against recorded hours against paid hours.

Where all three agree, nothing to do. Where they do not, the difference has a cause and the cause is usually one of the four above. Half a day for a site of a hundred, less once the common cases are known.

Most operations have never run it, and the first run is always uncomfortable: the discrepancies are larger and more numerous than anybody expects, and some of them are people who have been underpaid for months.

What to do with what it finds

Underpayments: correct promptly, in full, and tell the person what happened. This is both the right thing and, in most jurisdictions, an obligation that does not depend on them having noticed.

Overpayments: raise them promptly, because recovery is constrained by time and amount in most places and because the conversation is far easier at one month than at nine.

Unrecorded absence: record it, with the operational facts if they are recoverable.

Unrecorded cover: this is the interesting one, because it is where the invisible absorption described elsewhere shows up as a number.

The pattern rather than the instances

Individual corrections are the small part. The valuable output is which category dominates.

Mostly unrecorded swaps means the swap mechanism is not working. Mostly rota changes not reaching payroll means an interface problem. Mostly absences recorded in one system means the recording route is too hard at the moment it is needed.

Each is a specific fix and each removes a recurring class of error rather than one instance of it.

Why it is nobody's job

Payroll reconciles its own figures and does not see the rota. Scheduling maintains the rota and does not see what was paid. HR holds the absence record and sees neither.

Three partial views, each internally consistent, and the discrepancy only exists in the space between them. Giving one person the monthly job of looking across all three is the whole intervention, and it takes less time than the chasing it replaces.

The side benefit

A reconciliation that runs monthly also produces, as a by-product, the cover-by-route data this collection keeps asking for: how many shifts were covered internally, how many by bank, how many by agency, and what each cost.

That is the same join described for cost attribution, arrived at from the other end, and an operation doing the reconciliation is three-quarters of the way to the quarterly figure without any additional work.

Running it the first time

The first reconciliation is the hardest and produces the most findings. Do it for one site and one month rather than for everything, because the volume from a full run is unmanageable and the exercise gets abandoned.

Fix what that month shows, change whatever produced the largest category, then run it again the following month. By the third run the discrepancy count has usually fallen by most of its starting value and the job takes an hour, which is the level at which it becomes a standing monthly task rather than a project.