Every shift a staffing firm delivers gets recorded three times: the shift that was scheduled, the shift that happened, and the shift that was signed. In most operations the three records never fully agree. Staffing margin leakage is what falls into the gaps between them: unbilled hours, payroll corrections, disputed invoices, and compliance exposure.
What are the three stories of a shift?
It is the simplest honest model of the staffing middle office we know.
Scheduled is what was planned: who, where, when, at what rate, under which credentials.
Happened is what took place: the late start, the swapped clinician, the extra two hours, the missed break, the last-minute coverage.
Signed is what was approved and documented, the record that becomes the invoice and the paycheck.
The glossary entry keeps the short definitions in one place. The rest of this piece is about what happens between them.
Why do the three records disagree?
Not because anyone is careless. They disagree because they were created by different people, in different systems, at different moments, for different purposes. A coordinator builds the schedule in one tool. The shift happens in the real world, and the evidence of it lives in a text thread, a time clock, or a sheet of paper. A supervisor signs something days later, and a processor keys that into a payroll system.
Each of those handoffs is reasonable. Together they guarantee drift.
Where do scheduled and happened split?
A late start. A swapped clinician. Two hours of overtime nobody logged against the right code. Coverage filled by text message at 6am.
Every one of these is somebody's afternoon spent reconciling, and every one of them is invisible until payroll runs. This is the gap your coordinators live in. It is why the shadow spreadsheet is often the real schedule and the group text is the actual coverage system.
Closing it means the schedule has to be the place where change happens, not a plan that reality wanders away from. When a swap, a callout, or an ad-hoc shift is handled inside Scheduling & Shift Management, the scheduled record updates as the happened record forms, and the swap carries its own approval.
Where do happened and signed split?
Work performed but never approved. Approved at the wrong rate. Signed on paper that arrives three days late, or twice, or not at all.
This is the gap that costs money directly. It becomes a payroll correction, a client dispute, or an invoice you never send. It is also where rate cards matter: if the on-call or callback rule is applied by a person reading a spreadsheet, the signed record can be internally consistent and still wrong.
Closing it means hours are captured close to the work, whether by GPS-validated clock-in or a photographed paper timesheet, priced by one rate engine, and approved in one queue before anything reaches the client. That is the job of Time, Rates & Billing.
What does the gap cost in a real week?
One national, multi-brand healthcare staffing organization ran its paper timesheet process for one brand the way it always had. The week's timesheet processing started on Sunday and finished late Tuesday or Wednesday. Every sheet was hand-keyed, and then a supervisor cross-referenced every keyed timesheet back against the paper. That step existed because everyone knew the numbers were sometimes wrong. Nobody called that a problem either.
After the hand-keying step was replaced by software that reads the paper, with people still reviewing every sheet, the same work took one day. Two days of the week came back. The workflow went from 10 steps to 5, a 50% reduction in steps per timesheet, and from 6 systems touched to 2. In the same week, five weeks of backlogged expenses finally got cleared, work nobody had expected to reach, because the timesheets were already done.
The clearest measure in this layer is not cost per transaction. It is how long the operation spends not knowing what it is owed.
A payroll processor said it in week three: "I never thought I could have every punch ready to import by Monday afternoon."
Why can't AI just fix this?
Integration in the shallow sense moves data between systems. The real work is getting three accounts of the same shift to mean the same thing.
AI is genuinely good at parts of this. It reads the mess: handwriting, inconsistent formats, dozens of versions of the same document. It handles volume without fatigue, so the two-hundredth document gets the same attention as the first. It catches duplicates, like the same timesheet emailed four times because nobody replied.
But it cannot decide which of three conflicting versions is true. When the schedule, the timesheet, and the signed record disagree, AI processes the disagreement faster. It does not resolve it. Judgment and exceptions, such as on-call rules and multi-sheet edge cases, still belong to people. And if nobody owns a workflow, pointing AI at it produces a faster broken workflow.
Automating on top of disagreement just produces faster disagreement. Operational truth comes first.
How do you get from three stories to one?
Start with one workflow, not the whole layer.
- Talk to the people holding it together. The workarounds your team built are the diagnosis. Every improvised spreadsheet and every "only Deb knows how that works" is a flag planted exactly where your systems failed.
- Map one placement-to-payroll flow. At each handoff ask whether scheduled, happened, and signed agree. Someone is absorbing that difference by hand, and margin is leaking with it.
- Follow the relief. Start where the suffering is deepest, not where the ROI slide is cleanest. The numbers follow the pain, because both are made of the same hours.
- Pick one workflow and go. Expect iteration, and choose a partner who expects it too.
Expect the fix to keep moving. In one deployment, a client raised 7 issues and optimizations on a Thursday at 2pm, and all 7 were deployed to production before 7am on Friday, under eighteen hours later. In the same deployment, 45 minutes spent watching one reviewer do her actual job surfaced 8 frictions nobody had complained about. All 8 shipped within two days. Nobody had asked for them. They had adapted.
You will never finish the middle office. You are building the capability to keep finding the next crack and closing it. When the three stories of a shift agree, Analytics & Revenue Intelligence can finally tell you what you made, what you did not, and why.
The full paper, including the pressure-test checklist, is summarized in the Field Guide summary. For the wider definition, read what the middle office in staffing is.