Somewhere between the timesheet and the bank transfer, there is a spreadsheet. Someone opens it every month, reads hours off one system, types them into another, and works out what everyone is owed.
Nobody thinks of it as a system. It's just the bit in the middle.
Why it stays invisible
The cost of that spreadsheet doesn't show up anywhere you'd look for it.
It isn't a line item. The person maintaining it has another job title. The errors it produces don't announce themselves — they surface weeks later as a quiet conversation about a short payslip, and get fixed as a one-off adjustment rather than logged as a defect.
So the true cost is spread across three places: the hours spent building it, the corrections it generates, and the trust it slowly erodes with the people being paid from it.
The error rate isn't the problem
Say the person doing it is careful and gets 99.5% of entries right. That sounds excellent.
With 60 staff, that's roughly one wrong figure every four months. Each one is a conversation, a correction, and a small deposit into the belief that payroll is something you have to check.
That belief is the expensive part. Once people think their hours might be wrong, they start keeping their own records — and now you have two sets of numbers that disagree, and every dispute takes an hour instead of five minutes.
Where the errors actually come from
In our experience it's rarely arithmetic. It's:
- A shift that crossed midnight, counted as one day or two depending on who typed it
- An overtime threshold applied per week when the contract says per day
- A leave day that was approved in a WhatsApp thread and never made it to the sheet
- Someone who worked at two sites in one day, entered once
Notice that none of those are mistakes exactly. They're ambiguities — places where the source data didn't say enough, and a human filled the gap with a reasonable assumption that happened to differ from last month's reasonable assumption.
Closing the gap
The fix isn't a better spreadsheet. It's removing the step where a human decides what the data meant.
That means the hours used for pay have to be the same records staff created at the gate — not a summary of them. It means the overtime rule lives in the system rather than in someone's head. And it means approved leave sits in the same place as attendance, so an absence can't look like a no-show.
When those three hold, the month-end job stops being data entry and becomes review. You're checking exceptions rather than transcribing sixty rows — and the exceptions are the only part that ever needed a person anyway.