Payroll Remediation for Australian Payroll Managers

Something in your payroll hasn’t been sitting right.

Maybe a manager escalated a single pay anomaly. Maybe it’s three or four underpayments over eighteen months that you’ve started to suspect aren’t one-offs. Or maybe there’s nothing you can point to at all, just the knowledge that your award interpretation has never been documented and the people who set it up have moved on.

And there’s a reluctance to pull on that thread, because you don’t know what’s attached to the other end.

If historical pay does turn out to be wrong, the fix is a payroll remediation. That means establishing what every affected employee should have been paid, calculating the shortfall with superannuation and interest, and repaying it with evidence that stands up to a regulator.

It’s a defined process with a known shape. The hardest part is usually deciding to look.

So let us take some weight out of that.

Nobody has this perfect. Not one organisation we have audited has had clean historical pay data, and that includes some of the best-run employers in the country. We have seen every state of record keeping you can imagine and several you’d struggle to. None of it surprises us, and none of it comes with finger-pointing.

The organisations with the strongest reputations run payroll audits and remediations as routine. Compliance moves: awards vary, agreements roll over, systems get replaced, interpretations drift and the people who understood them leave.

Checking is not an admission of guilt. It’s a pursuit of excellence.

This article sets out the payroll remediation process we follow, phase by phase. Joe Woolley and Johanne Matthews walked through the same method in a recent webinar.

Watch the Payroll Remediation Webinar (37:54)

Isn’t the award the hard part?

No, and that surprises people.

SCHADS has earned its reputation, and anyone who has worked through it knows exactly why. But an award is knowable. You can read it, map it and pin down what it requires.

Your data is where this comes apart.

To know whether one person was paid correctly on one shift, you have to solve an equation. Every variable in it has to land:

• Worked time: what happened, not what the roster said
• The award rule: how that time should have been paid
• The rate: base, penalties and loadings
• Classification: the level and the pay point
• Entitlements and allowances: everything sitting on top

Miss one and the error runs through everything you build above it.

Now scale it. The same equation runs for every shift, for every person, across the period in question, which is frequently six years of history.

Where the gaps actually sit

Worked time falls apart first.

Breaks go unrecorded. Shift times change and nobody captures the change. Broken shifts, sleepovers and interrupted shifts all move the pay, and all of them go missing.

But the gaps run deeper than missing punches.

You replace a system and the old data lands in a format nobody can read. The payroll manager who understood how it all fit together has left, and the reasoning left with her.

And underneath everything sit the interpretation errors:

• An agreement that references SCHADS with a gap in a clause
• A classification nobody updated after a variation
• A whole cohort paid on the wrong award for years, quite sincerely, by people doing their best with what they had

None of that is unusual. It is what happens to any large employer over a long enough period, and it is why the work exists.

At scale you’re dealing with millions of data points across thousands of records. The real skill is knowing how to fill a gap. And when you can’t fill it, knowing what to document so the answer holds up in front of Fair Work or a court.

Why we take this personally

We started after the Queensland Health payroll disaster, still one of the worst in the country.

Our CEO, Brett Cowan, was the User Acceptance Test Manager on that program. He flagged more than 2,400 defects before go-live.

The system launched anyway.

In the inquiry that followed, the commission praised Brett as one of the few witnesses who told Queensland Health plainly that it should not proceed.

Other organisations started asking whether we’d look at their payroll too. Hence the name. We now work with around 230 clients across government, healthcare, aged care, retail, education and mining.

How the process works

Six phases, in this order:

1. Establish what correct looks like
2. Work out what data actually exists
3. Build the assumptions register
4. Account for prior corrections
5. Run the calculation engine
6. Verify, and expect a messy first run

Each one has to close before the next can start.

Phase 1: Establish what correct looks like

We map every industrial instrument that applies before we read a single pay record.

We also map how your organisation interprets each one. That gives you one source of truth, and we measure every pay record against it.

This phase catches more than people expect:

• A segment classified incorrectly
• A cohort sitting under the wrong award entirely
• Pay rises written into an agreement and never paid through

Get this wrong and every assumption you build on top of it is wrong too.

Phase 2: Work out what data actually exists

Next we establish what data exists. Not what you think exists. What we can locate, extract and use.

That means four things:

• Worked time: real punch-in, punch-out and breaks
• Pay data: the code, rate and classification on every line
• Employment records: status and classification at each point in time
• Legacy data: anything stranded in a system you replaced inside the audit window

This phase needs candour more than it needs technical skill.

Most organisations assume they’re fine until we look under the hood. That’s not a criticism, it’s just what we find. Tell us early and we can work with it.

It’s also the slowest phase, because we have to gather everything before the audit starts. The worst outcome is being three months in when someone mentions a system nobody told us about, still holding data we’d written off. Then we start again, because the number has to be right.

Missing data does not make remediation impossible. We see three situations, and only one of them is genuinely difficult:

• Complete records: you can retrace every shift, every pay period, for every employee. Painstaking, but straightforward.
• Partial records: some data is there, some isn’t, and the assumptions register does the heavy lifting. This is the common case.
• Limited or no records: shift-level reconciliation gets hard, because that needs time data. But we can still assess rates, classifications and pay point progression.

None of those three is a reason not to start. They change how much work lands in the next phase, not whether you can get to a defensible number.

Phase 3: Build the assumptions register

Every gap goes into an assumptions register when records are incomplete.

For each one we record three things:

• The gap
• The assumption we’ve made
• Why we made it

The split that matters is safe versus unsafe.

• Safe: the evidence points one way. Shift times missing, but the roster intact. You can assume the rostered hours were worked, so document it, apply it and move on.
• Unsafe: the evidence is ambiguous and the call could move the number. We don’t make those for you. They go to your IR team, your legal team or external counsel, and we attach that decision to the audit so the opinion protects you later.

One rule holds all of it together. Decide an assumption once and it applies to the whole population.

Apply one logic to one group and different logic to another and the register falls over the moment anyone examines it.

Higher duties is the classic case. Awards set different trigger periods for the higher rate, and you can often see that higher duties happened without seeing how long it ran. With no evidence either way, the defensible call is to apply the higher rate for the whole shift, as the award provides.

Phase 4: Account for prior corrections

Many organisations have already run some kind of fix before we arrive.

So we need every back payment and correction, by pay item and by period.

It’s easy to underrate this one. A client finds two or three issues, corrects them, then later finds more. By the time we run the full audit, those earlier payments are critical context. We have to know exactly what each one covered, so we neither double-count it nor miss it.

Important: since STP Phase 2, back payments no longer show as distinct line items.

They look like ordinary pay. Which means if you don’t tell us, there’s a good chance we won’t see it.

Phase 5: Run the calculation engine

By now we’ve validated the mapping, documented the assumptions and consolidated the data.

What’s left is the calculation, and this is where the scale of the thing becomes real.

Take the simplest worker you have. Starts, breaks for lunch, returns, finishes. That’s four data points in a day, twenty in a week and around 6,000 across six years for one person.

Now add shift work, part-time minimums, higher duties and leave. Multiply that across your workforce and you’re into millions of lines.

That’s well past what a spreadsheet can carry. We use a remediation engine, a database with tools layered over it. It runs every rule against every data point, for every employee and every shift, at once.

This stops being payroll somewhere around here. It’s closer to data science, and no payroll team should have to do it alongside the day job.

Consistency is the whole point of running it this way. Put four analysts on the same data with the same rules and they should reach the same number. If they don’t, something in the mapping or the assumptions needs another look.

One thing worth being clear about. That output isn’t the raw data. It’s the result, broken into every category, so we can interrogate it.

Does that look right? Is it valid? Should it even be there?

Phase 6: Verify, and expect a messy first run

A second reviewer works through a sample before we accept any result.

The first run is always dirty. We want it that way.

A clean first run would mean we’d assumed everything was correct, which is the one thing we already know isn’t true. We want it to throw up issues, because that’s what tests the gaps and surfaces the assumptions nobody anticipated.

A typical audit runs five to ten times before the result settles. We have never reached a final number on the first pass, on any engagement.

What this looks like when the data is bad

A client came to us with 40 to 50 people they’d been paying on the wrong award, which they’d identified themselves.

No detailed time records. And the correct award turned heavily on shift length and time of day, which was precisely what we didn’t have.

So we went back to the rosters. We used every payslip we could find. And we documented careful assumptions, applied the same way across the whole cohort, to land each person’s number.

Every one of them ended up with a figure we could defend.

Harder than a clean data set. Not impossible.

What you cannot do: offset one period against another

We get this one constantly, usually with a hopeful look. Can you net an overpayment in one period against an underpayment in another?

No.

Unlike items can’t cancel each other either. So you finish a remediation holding both underpayments and overpayments across different pay items, and they don’t net off.

The courts have tightened this further. In the Woolworths proceedings, Justice Perram ruled that set-off clauses only operate within the same pay period. You cannot pool overpayments across the year to cover shortfalls elsewhere.

That single ruling cut through a practice thousands of Australian employers were relying on.

Whether you can recover an overpayment is a legal question rather than a payroll one. Where we think there’s scope, we flag it, you get advice and we apply whatever counsel allows.

Budget on the gross number. Treat anything you recover as a bonus.

Why the legal ground has moved against waiting

Four things have changed:

1. Criminal wage theft: wage theft became a criminal offence in January 2025. The headline is tougher penalties for intentional underpayment. But the part that should hold your attention is the bar it now sets on negligence. The burden sits with you to make sure pay is right.
2. Section 557C: the employee is now the beneficiary of the doubt. Where information is missing, the law presumes it would have favoured them, and we have to audit on that basis. Poor record keeping has stopped being an awkward footnote and started being a number.
3. Real-time reporting: your payroll data already reaches the ATO through STP. With AI running across those data sets, the ATO can spot payment patterns that were invisible a few years ago.
4. Payday Super: from 1 July 2026, you pay super at the same time as wages. It has to reach the fund within seven business days. The old quarterly float is gone.

Now put those together. You do not want to be paying super on overpayments, and you certainly don’t want to be short on underpayments, given the penalties and the tax consequences that follow.

One more point on timing. If you suspect a problem and commission an independent audit, that gives you a measure of safe harbour. If Fair Work later comes asking, you can show you had a concern and acted on it in good faith.

What you walk away with

We run the engine iteratively, refining as we go, until the results settle. Then we present the findings and walk you through the numbers.

You get four things:

1. A complete data set: every line, every employee, every period
2. A methodology and findings report: how we got there, not just where we landed
3. An audit trail ready for regulators: built for scrutiny from the start
4. An interpretation you can rely on going forward: so the same errors don’t start again next pay run

We calculate superannuation and interest on top of the principal.

That package is built to do two jobs at once. It has to satisfy a regulator, and it has to give you something you can take to a board and to your workforce without needing a translator in the room.

And if you need it, we help on the employee side too. These audits often surface former staff who are owed money. We can trace them, confirm current super and bank details, explain the payment and manage the contact.

What needs to be true for this to go well

Remediation has a lot of moving parts. For it to run cleanly, five things need to hold:

1. You give us the full data picture early.
2. Your legal or IR team engages on the unsafe assumptions instead of deferring them.
3. You accept a first run that looks wrong, because that’s the run that finds the gaps.
4. You hold the same assumption across the whole population, even where it costs you more.
5. You move before somebody else decides to move for you.

Why proactive beats reactive

This is the part that decides how the whole thing feels, for you and for your people.

It comes down to two positions:

1. Proactive: you decide who you tell, when and how. You get credit for acting early, safe harbour is on the table and your people see that you moved before anyone forced you to.
2. Reactive, pushed by a complaint or a union or a call from Fair Work: you don’t set the timeline. You don’t set the tone. And the trust you spent years building erodes fast.

Waiting also costs money, because the first number is rarely the final one. Woolworths announced $200 to $300 million in 2019, and by 2023 the figure had passed $750 million. NAB spent $250 million between 2020 and 2022 on errors dating back to 2012, then announced a further $130 million in 2025.

In both cases the exposure grew after the initial finding, because a first look is never the full picture.

We have sat on both sides of this. Telling your workforce “we have a concern, it will take some time, please bear with us while we make sure everyone is paid correctly” is uncomfortable.

It’s nothing like as uncomfortable as the conversation where your people ask how you let it happen.

Where to start

We work with hundreds of organisations. There’s very little we haven’t seen, and none of it comes with finger-pointing. Whatever state your data is in, the job is the same. Get you to compliance with a defensible number and a clear path, and make sure your people are paid correctly.

Something has usually been nagging at someone for a while before they pick up the phone. If that sounds familiar, reach out for a confidential chat.