Ghost workers: your agency is paying people who do not exist

A worker leaves your agency in March. Nobody closes the assignment. Hours keep arriving against their name, the wage keeps landing in the same account, and because your roster turns over every week, nothing about it looks strange. That is a ghost worker: a salary going out every cycle to somebody who is not working for you. The median fraud runs about twelve months before anyone notices.¹

Most operators hear "payroll fraud" and picture a finance problem, something for the auditors. It is not. It starts in the roster, it is committed by somebody who can edit a schedule, and it is paid for out of the same margin you are trying to protect on every filled shift.

The first two numbers come from the same report, covering 2,402 fraud cases worldwide. The gap between them is not a gap in sophistication. It is a gap in time. A ghost worker on a modest hourly rate costs you almost nothing in month one and a great deal in year four, and nothing about the scheme has to change in between. The third number is the same mechanism running unchecked in a public payroll, at a scale that made it a court case.¹

Why a staffing roster hides a ghost better than a payroll does

The leaver who never left the schedule is the first of two shapes. In the second, somebody with access to the roster submits hours for a shift that was never worked, either against a real worker's record or against a name they invented. Neither is clever, and neither needs an accomplice in payroll.

Both rely on the same thing: that the list of people you rostered, the list of people who actually turned up, and the list of people you paid are never held next to each other in the same week.

In a fixed-headcount business, an unfamiliar name on the payroll is an anomaly. In a staffing agency, it is Tuesday.

That is the real exposure. A manufacturer with 300 permanent staff notices a 301st. An agency running 400 shifts a week across nine sites, with 40 people joining and leaving in the same period, has no baseline to notice against. The churn that makes the business work is also free camouflage.

Then there is who sits where. In most agencies the same coordinator assigns the worker, receives the timesheet and approves it for payment. One person, three controls. No finance function would allow that arrangement anywhere else in the business, and it survives in staffing because it is fast and because everyone trusts the coordinator, usually correctly.

The most common way this gets found is that somebody tells you

Across those 2,402 cases, the single largest share of frauds were first uncovered by a tip, and more than half of the tips came from employees. Internal audit and management review, the two controls an operator would actually point to on a slide, come second and third by some distance.²

Read that as an operator rather than an auditor. If the most likely way you find a ghost worker is that a coordinator mentions something in passing, you do not have a detection control. You have a reporting channel and some luck. The reason schemes run for years is not that they are hard to see. It is that nothing is scheduled to look.

The reconciliation playbook

None of what follows needs new software before it starts working. Steps one and two can be run this month against exports you already have.

1. Put three lists side by side

   Assignments, attendance, payroll. Who you rostered, who actually turned up with evidence attached, who you paid. Any line on the payroll file with no matching attendance record is the whole investigation. Most agencies hold all three and have never once compared them in the same sitting.

2. Automate the exceptions worth a human look

   Two workers sharing a bank account. Somebody who has never missed a shift and never taken a day off in fourteen months. Hours identical to the minute, week after week. Check-ins with no location attached. A name on your payroll that never appears in the client's own gate log.

3. Verify attendance where the work happens

   Geofenced check-in from the worker's phone or a site device, with liveness verification where identity substitution is a genuine risk. Evidence captured once, at the moment of the shift, rather than reconstructed from a paper sheet on Friday afternoon.

4. Put the approval gate before the money, not after the audit

   Anything that changes what a person is paid carries human approval and a log with the value before and after the change. Closing this loop on a 300,000-worker deployment cut payroll leakage by roughly 90%, which is our own operating figure rather than an industry one.⁵

5. Re-run it on a schedule nobody controls

   A monthly job that runs whether or not anyone remembers, with the output going to someone outside the branch being reconciled. A control that the reconciled party can quietly skip is not a control.

What three years cost Nigeria

The IPPIS case is worth following because every stage of it is dated, and the dates are the point.

1. 2023
   A systems study flags irregularities

   The ICPC reviews the federal payroll platform and finds signs of personnel who cannot be matched to real people.³

2. Apr 2024
   587 suspected ghost workers identified

   A joint investigation with the Office of the Accountant-General traces fictitious payroll identities across ministries, the police and several universities.³

3. 2025
   120 of them turn out to be real people

   A verification exercise clears 120 civil servants and reinstates them on the platform.³

4. Jul 2026
   ₦941,994,079.86 permanently forfeited

   The Federal High Court in Abuja orders final forfeiture of the traced funds to the federal government.³

5. 19 Aug 2026
   A forensic audit of the whole payroll is ordered

   The Federal Executive Council commissions an audit of Nigeria's ₦9.5 trillion payroll, three years after the first flag.⁴

Two details in that sequence matter more than the headline figure. The first is that 467 bank accounts in the case are still linked to people nobody has been able to identify, three years in. The second is that roughly one in five of the workers the process accused had done nothing wrong, and spent time off the payroll while it was sorted out.³

That second number is the argument against running this as a sweep. A blunt detection rule produces victims who are your own staff, and in staffing the people it hits hardest are the ones least able to absorb a missed pay cycle. Exception rules need a second signal before they touch anyone's money, which is the same lesson the geofence taught us, arrived at from the opposite direction.

Nigeria's payroll is unusual in scale and in publicity. It is not unusual in structure. A list of names, a list of hours, a list of payments, and nobody whose job it is to check that the three agree.

Nobody discovers a ghost worker. They discover how long one has been there.The rule, in one line