Payroll system migration: how to stop historical errors moving into your new platform

A new payroll platform does not fix your payroll system.

The platform is one part of it. The rest is your data, your award and agreement interpretation, the processes around the software and the people who run them.

Most non-compliance starts in that second group rather than in the technology. It moves across with your data, and it keeps accruing. Claims under the Fair Work Act reach back six years.

Here is what to check, and when.

Why organisations migrate payroll systems, and what each reason carries

Most migrations start for one of four reasons.

1. Software at end of life, where the vendor has stopped investing in the interface, the data storage and the award interpretation
2. Automation, to remove the manual steps holding the process together
3. Integration, where payroll, time and attendance and HR no longer talk to each other properly
4. A remediation program, where an underpayment has been found and the system is part of the cause

Three of those four put the attention on the technology. The data, the interpretation and the processes then get less scrutiny during the project, even though they are often what is driving the change.

What crosses over in a payroll data migration

Historical payroll errors survive the migration

Configuration defects run for years without being noticed.

One of Australia’s largest resources companies disclosed in 2023 that leave had been incorrectly deducted on public holidays since 2010. The correction came to about $430 million across 28,500 current and former employees.

A Group of Eight university disclosed a design flaw in its timesheet system in 2024. Timesheets stayed visible for only two weeks, so unprocessed ones never surfaced. It ran for 11 years and covered 2,290 current and former staff.

Fair Work Ombudsman enforceable undertakings name the same two causes over and over: governance and the payroll system itself. A tier-two public university paid $21.4 million across 10,443 employees. A dual-sector regional university paid over $4 million, attributed to a decentralised payroll system and improper configuration.

Payroll metadata is lost, the seven year record obligation is not

Your pay codes migrate. The record of how they were read does not.

What gets lost:

• How each pay code was interpreted, and when that interpretation last changed
• Who owned which field
• Why a classification was set the way it was

The people who held that knowledge have usually left by the time anyone needs it. Section 535 of the Fair Work Act still requires you to keep employee records for seven years, and section 544 lets a claim reach back six.

Legacy calculation rules follow the manual processes across

Termination payments, long service leave and back pay often keep their legacy rules.

The same manual steps then run on new data, and they rarely get tested.

Four payroll migration red flags before the project starts

• Only legal questions coming your way: the questions are about privacy and record retention, never about whether the calculations are right.
• No budget to check the historical data: the business case covers the new system, not the entitlements going into it.
• Stakeholders who have never met: payroll, workplace legal and workforce planning report to different executives, and all three change how people get paid.
• A compressed testing window: vendors are measured on time and budget, so user acceptance testing is the first thing to shorten.

One state health department shows what the last flag produces. Its project board revised the defect severity definitions in July 2009, then changed the exit criteria for the final round of testing. The commission of inquiry that followed called the result a catastrophic failure.

Six payroll data checks to complete before go-live

1. Cleanse payroll data before the project starts

Not midway through, when the load templates are due and the timeline is already fixed.

2. Establish what correct pay looks like under every industrial instrument

Map every award, agreement and contract that applies, and how your organisation interprets each one. Every pay record then gets measured against a single standard.

This is where cohorts sitting under the wrong award show up, along with pay rises written into an agreement and never paid through.

3. Document the payroll configuration knowledge your team holds

Capture the reasoning behind your configuration while the people who hold it are still in the business.

4. Validate the old payroll system against the new one, and keep the output

Check and record four things:

• Pay code mapping
• Employee ID mapping
• Organisational structure
• Span of control

Save the results somewhere you can find them. Anyone reconstructing the past later starts from that record instead of from nothing.

5. Run a full parallel pay run, not a sample

That same health department tested a parallel pay run across ten per cent of employees, eight months before go-live. Its software vendor had recommended a full parallel run and the project board declined it, because of the size of the task.

A sample confirms the common cases. Errors concentrate in the uncommon ones: broken shifts, sleepovers, overtime, allowances and terminations.

6. Run the legal workstream alongside the build

Settle four questions before testing finishes.

• Whether you notify an underpayment when you find one
• Whether you pay interest
• Whether you need superannuation and tax advice
• Who briefs the executive

Since 27 February 2024, section 557A means a serious contravention of the Fair Work Act needs only recklessness. It no longer has to be knowing and systematic.

What to do when your payroll records have gaps

You will find entitlements you cannot test, because the data was never captured.

There are three defensible ways to handle it.

• Pay it: assume everyone was entitled and pay it. Lowest risk, highest cost.
• Ask for it: run an opt-in claim supported by a statutory declaration. Lower cost, and it puts the effort on employees.
• Assume it, and document why: build an assumption from the data you hold and apply it consistently. Lowest cost, and it needs the clearest paper trail.

Section 557C puts the burden on you to disprove a claim where the records were not kept. The Fair Work Ombudsman expects employees to get the benefit of the doubt where records are deficient. Its remediation guidance treats six years as a minimum review period rather than a limit.

Three payroll compliance checks after go-live

Test the payroll calculations that sit outside the build

Termination payments, long service leave and back pay run on new data through old rules. Check them against real cases inside the first three months.

Set a review point for every award and agreement change

Each annual wage review, each agreement expiry and each acquisition changes what correct looks like. Put a date against each one.

Name an owner for each payroll data domain

Put data quality and change control in a position description, with a name against it.

Errors that reach production stop being a project issue. They become a liability that accrues every pay run until somebody finds them.

Payroll Experts runs payroll audits, remediation, system implementations and processing support for about 230 organisations across healthcare, aged care, government, education, retail and mining.

Sources: Fair Work Act 2009, sections 535, 544, 557A and 557C. Fair Work Ombudsman, Payroll Remediation Program Guide (April 2025) and published enforceable undertakings. Company and university disclosures. State audit office and commission of inquiry reports into a public health payroll replacement. Organisations are not named.