Why the parallel run is the whole argument
A payroll that produces plausible numbers is not the same as a payroll that produces correct ones. Plausible is easy. Everybody gets roughly what they got last month, and nobody notices the allowance that quietly stopped, the tax slab applied at the wrong threshold, or the contribution calculated on the wrong base until a year end statement arrives and several people ask the same question on the same morning. By then the correction has to run backwards through a closed tax year, which is a different order of work entirely.
The parallel run compares the new system against the current one, employee by employee and component by component, across a full cycle. Every difference gets traced to a cause. Some of them will be the new system being wrong. A meaningful number, in our experience, turn out to be the old system being wrong in a way nobody had noticed, which is uncomfortable and very much better discovered during a parallel run than during an audit.
One cycle is the minimum and two is better, particularly where pay varies month to month. A single quiet month will not exercise overtime, arrears, mid month joiners, leavers with a final settlement, or the loan instalment that reaches its last deduction. We plan the parallel run across months that contain those events rather than across the calmest period in the calendar, because the calm month proves almost nothing. Where the business has a bonus cycle or a seasonal shift pattern, that is the month worth testing against.
- A full cycle compared employee by employee and component by component
- Every variance traced to a cause rather than tolerated as rounding
- Parallel months chosen to contain joiners, leavers, arrears and variable pay
- Sign off by the payroll owner as the condition for switching, ahead of any announced date
- Findings against the existing payroll reported honestly, including where it was the one at fault