Parallel pay runs: proving the new system to the cent

A parallel pay run proves the new payroll system by processing the same workers, hours and rules in both systems, then explaining every difference before anyone is paid.

What a parallel pay run actually proves

Think of the old system as the control and the new system as the test. Both receive the same employee details, timesheets, allowances, leave, deductions and pay dates.

You then compare the results employee by employee and line by line. Checking only the total net pay is not enough. Two errors can cancel each other out and leave the total looking right.

For a construction payroll, the comparison should cover:

• ordinary hours and base rates
• overtime and penalty rates
• travel, fares, meals, tools and site allowances
• rostered days off and accrued time
• annual leave, personal leave and leave loading
• taxable and non-taxable items
• PAYG withholding
• super guarantee and other super contributions
• deductions and child support amounts
• net pay
• job, project and cost-code allocation
• Single Touch Payroll categories

The aim is a zero unexplained difference. A difference caused by an approved rounding method may be valid. A mystery cent is not.

You can find the broader controls around pay processing in our payroll operations hub.

Lock down the test before comparing anything

Start by freezing the source data. Use the same employee master file, timesheet cut-off and approved pay adjustments in both systems.

Do not let one team repair data in the old system while another team changes the new one. You will end up comparing different inputs and blaming the calculation engine.

Record at least these details for every parallel cycle:

1. Pay period start and end dates.
2. Payment date.
3. Employee population included.
4. Timesheet and leave cut-off times.
5. Award, enterprise agreement or contract version.
6. Tax and super settings.
7. Software configuration version.
8. Manual changes made after the first calculation.

Use a closed pay period for the first test. Once the basic calculations match, run current data with real late timesheets, corrections and terminations. That is where awkward problems usually show up.

Check the dated control figures

A parallel run should use the rules applying on the actual payment date. Do not copy settings from an earlier implementation workbook and assume they remain current.

Useful control figures include:

• The super guarantee rate is 12% as at October 2026.
• The Australian resident tax-free threshold is $18,200 as at October 2026.
• The resident marginal tax rate from $18,201 to $45,000 is 15% as at October 2026.
• The standard Medicare levy is 2% as at October 2026, although exemptions and reductions can apply.
• Payroll records generally need to be kept for 7 years as at October 2026.

These figures are checks, not a home-made PAYG formula. PAYG withholding should follow the ATO schedule and employee declarations applying to the payment date. Super also needs an ordinary time earnings assessment. Not every payment attracts super merely because it is taxable.

Award and enterprise agreement rates need separate version control. Record the clause, classification and effective date used. Construction allowances can change because of work location, duties, tools, travel or project conditions.

Compare lines, not just totals

Export both systems into the same layout. Give each employee and pay item a stable identifier. Descriptions alone are risky because one system might say “Site allowance” while the other says “Construction allowance”.

A useful comparison sheet looks like this:

Area	Expected tolerance	What to compare	Evidence to keep
Gross earnings	$0.00	Hours, rate, multiplier and amount by pay item	Timesheet, rate table and calculation report
PAYG withholding	$0.00	Tax scale, declarations, study loan settings and payment date	Employee declaration and ATO-based output
Super	$0.00	Ordinary time earnings base, rate, fund and exclusions	Super report and pay-item mapping
Leave	$0.00	Opening balance, accrual, leave taken and closing balance	Leave ledger and employee history
Net pay	$0.00	Gross less tax, deductions and adjustments	Payroll register and bank file control total
Cost allocation	$0.00	Project, job, phase and cost code	Costing export and general ledger mapping
STP reporting	Exact category match	Income type, allowances, overtime, leave and termination fields	STP validation report

If the old system rounds each day while the new system rounds at the end of the week, you may get a cent difference. Do not simply post an adjustment. Document the calculation order, decide which method complies with the governing instrument, and configure the correct method.

Our construction payroll compliance audit can help organise the clauses, pay items and evidence behind those checks.

Use workers who will expose weak spots

A test made up only of salaried office staff proves very little for a builder or subcontractor.

Include workers with different patterns, such as:

• a full-time tradesperson working ordinary hours
• a worker with weekday overtime and weekend work
• an apprentice who changes year or stage
• a casual with overtime or allowances
• a worker taking annual leave with leave loading
• an employee using an RDO
• a worker with salary sacrifice or extra super
• someone with child support or another deduction
• a new starter without a full pay period
• a terminated employee receiving unused leave
• a worker moving between projects or states
• an employee receiving back pay

Run at least two ordinary cycles and one cycle containing irregular events. This is a practical testing approach, not a statutory minimum. If your roster repeats over four weeks, test the whole pattern.

Trace each variance to its cause

Give every variance an owner and status. Useful cause categories include source data, configuration, interpretation, rounding, interface and manual entry.

For example, a super difference might come from an allowance being mapped as ordinary time earnings in one system but excluded in the other. An overtime difference might come from the new system applying a multiplier to the loaded rate instead of the base rate. A leave difference could start with an incorrect opening balance rather than the current pay calculation.

Reprocess after each correction. Keep the original result, the reason, the approved change and the new result. That history matters if a manager, auditor or employee later asks why the systems differed.

Never make a balancing entry just to force the totals to agree. It hides the cause and can carry the error into future pays.

Sign off before money or reports leave the business

The test environment should not release bank payments, super contributions or duplicate STP reports. Confirm how files are marked and who can transmit them.

Before cutover, payroll and finance should sign off employee totals, bank totals, PAYG, super, leave, costing and general ledger entries. The person responsible for awards or enterprise agreements should approve the rule mapping as well.

Keep a fallback plan for the first live run. That means retaining access to the old records, preserving approved inputs and knowing the latest point at which the team can stop without creating duplicate payments or reports.

“To the cent” does not mean ignoring why the systems calculate differently. It means every cent has either matched or been traced to a documented, compliant reason.