Detecting timesheet variance without measuring people

This note tests a local question: where do recorded hours differ from the approved schedule, the leave record, or the payroll line, and what explains the difference? The unit is one worker-day. The method describes variance; it does not judge productivity.

Join the schedule, the timesheet, the leave record, and the payroll register by worker and date. Record the time zone and the cut-off. Mark missing joins rather than dropping them.

Use a small taxonomy: approved overtime, unapproved overtime, leave mismatch, holiday treatment, schedule change, entry error, and unknown. Unknown is a real category and should be reported.

Take a stratified sample by week and by team. For each item, open the approval and the payroll line. Where the difference is a process issue, note the step that failed.

Report the variance count, the value, and the class mix. Pair it with the number of items reconciled and the number still open. Do not convert the result into a performance score.

The method depends on consistent records and may miss informal schedule swaps. It measures recorded agreement, not whether the work was useful.