January is when compliance teams make plans. Audit the WFM configuration. Review the payroll mappings. Retrain the managers on award changes. Build a reconciliation process. Maybe even get budget for it this year.
By March, they're fixing payroll errors.
I've watched this cycle play out at a dozen mid-market employers and the pattern is identical every time. The plan is good. The intent is real. But the first batch of pay run exceptions lands in late February, and the compliance initiative quietly dies on a whiteboard while the team spends three weeks reconciling underpayments from configurations that drifted over Christmas, and by the time they surface for air it's April and the next quarter's problems are already compounding.
The reason this keeps happening is architectural. Not cultural. The systems being governed, Deputy, Humanforce, KeyPay, Xero, are automated. They process thousands of calculations per pay run without hesitation. A mid-market employer with 800 staff on fortnightly pay generates somewhere around 3,000 to 5,000 shift-level calculations every cycle, each involving a base rate, a loading, a penalty, maybe an allowance. The system does this in seconds.
The governance layer checking whether those calculations are correct?
That's a person with a spreadsheet. Maybe they spot-check 5% of transactions. Maybe they pull a sample of Sunday shifts and compare rates manually. Maybe they run a quarterly reconciliation that takes two weeks and is already outdated by the time it's finished. However they do it, they're catching errors at human speed while the systems generate them at machine speed.
The gap only widens.
The spreadsheet ceiling
There's a threshold, and I'd put it at roughly 200 employees across multiple awards, where manual cross-system reconciliation becomes physically impossible at a frequency that matters.
Below 200, a competent payroll manager can hold the whole picture in their head. They know which casuals are on junior rates, which sites have different allowance triggers, which awards apply to which teams. They catch drift because they remember what the numbers should look like.
Above 200, it's too many variables. The reconciliation shifts from "I know this is wrong" to "I need to build a process to check whether this is wrong," and that process is always manual, always periodic, and always slower than the systems it's trying to validate, which means the errors accumulate between checks and compound silently until someone stumbles across them.
A casual loading misconfigured in week one runs for six fortnights before the quarterly review catches it. By then it's a remediation. Not a correction. Multiply that by every derived value in every award across every employment type, and you understand why the compliance plan from January never survives contact with the March pay run.
The problem isn't discipline. The team cares. The team is skilled. The problem is that you cannot manually govern automated systems. The governance layer has to run at the same speed and frequency as the systems it validates. Otherwise you're not preventing errors. You're discovering them later.
Every compliance plan I've seen assumes the team will have time to be proactive. None of them account for the fact that the systems will keep generating work faster than the team can check it.