The timesheet for last Tuesday looks perfect. Two names, both clocked in at 5:58 a.m., both clocked out at 2:04 p.m. Eight hours each, no exceptions, nothing for a manager to approve. The camera over the register tells a different story. From six in the morning until a little after nine, there is one person behind the counter, not two.
Nothing in payroll will ever flag that Tuesday. The punches are clean, the arithmetic is right, and the checks go out on Friday. That is a ghost shift: hours that exist in the payroll system and nowhere else in the store.
Labor is the biggest number in the store and the least verified
Labor is now the largest controllable cost inside a convenience store. At the 2026 NACS State of the Industry Summit, NACS Research reported that labor is responsible for $1.60 per transaction rung inside the store, the chief component of direct store operating expenses, and that those expenses have grown faster than inside gross profit for the second straight year. The same research put the average basket at $7.69 and calculated that, after every expense is subtracted, the typical transaction inside the store lost seven cents in 2025.
Put those numbers next to each other and the exposure is plain. There is very little room inside the store to absorb an hour that was paid but not worked. The hours are not cheap either: average hourly wages in the industry just passed $15, across roughly twenty employees per store. One unworked hour a day at that rate is about $5,500 a year in a single store, before payroll taxes. Across twenty stores it is more than $100,000. It never shows up on any report as a loss. It shows up as normal payroll.
Now compare how the two big controllable costs actually get checked. Merchandise gets counted. Deliveries get signed for and matched against invoices. Shrink gets its own meeting. Labor, the larger number, gets verified by asking the time clock whether the time clock is right.
What a timesheet actually records
Federal law is specific about what an employer has to keep. Under the Fair Labor Standards Act record keeping rules at 29 CFR 516.2, an employer must record the hours worked each workday and the total hours worked each workweek for every nonexempt employee. The same regulation offers a shortcut for employees on a fixed schedule: the employer may note that the scheduled hours were worked and write down only the weeks that differ.
That shortcut is entirely legal, and it was written in an era of paper records. It also describes the blind spot exactly. The record says the schedule happened. It does not say the work did. Everything downstream inherits that assumption: the payroll run, the labor percentage, the schedule built for next week.
The patterns that live in the gap are rarely dramatic.
- One employee punches in a coworker who is still ten minutes away.
- Someone clocks in from a phone in the parking lot, then takes a while to reach the counter.
- A fifteen minute break runs to forty minutes and nobody punches out.
- The closing punch reads 11:00 p.m., but the lights went off and the door was locked at 10:30.
- A manager edits punches after the fact, and nobody reviews the edits, because nobody has a reason to.
Each one is small. Small and daily is how a labor line bleeds without ever looking wrong.
No single system can see it
Ghost shifts survive for a structural reason. The time clock knows the punch, not the person who made it. The camera knows who was on the floor, but not who was being paid. The register knows which operator ID rang which sale, but not who was scheduled to be there. Each system holds a third of the answer, and none of them talk.
Joined, the same three streams read very differently. A cashier is on the clock from three in the afternoon until eleven at night. The register records no transaction under their operator ID after ten past eight. The camera shows one person behind the counter, not two, from eight until close. Any one of those observations is noise. Together they are a specific question about a specific shift, and a manager can ask it the next morning, while people still remember the night.
That join is what we build. CASH crossed with POS crossed with CAM crossed with LABOR, continuously, on the cameras and registers a store already owns. The labor side of it surfaces as the Workforce Honesty Score, which is not a verdict about any person. It is a measure of how much of the paid schedule the rest of the store can corroborate.
It has to work in both directions
A tool that only ever accuses people is not worth installing, and it will not survive contact with a real store. In practice the joined record clears people at least as often as it raises a question. An employee who says they came in early to meet a delivery now has a camera that agrees with them. A cashier blamed for a drawer variance can be ruled out, because the record shows they were stocking cooler doors while the drawer was open.
It also catches the failure that runs the other way, which is the more expensive one. Hours worked and not paid is not a shrink problem. It is a wage claim, and it usually begins as an honest scheduling mess rather than anything deliberate. A store that can show what happened on the floor is in a far better position than one holding a timesheet and a memory.
None of this requires new hardware. When NACS Research walked operators through cutting store expenses, the advice on shrink was practical and plain: write it down, watch for patterns, and "look at your cameras if you have to". That is good advice for hours as much as for merchandise. It is also the part that quietly never happens, because a manager running a store does not have spare hours to scrub footage. Doing that review automatically, every shift, is the only version of the advice that survives a busy week.
If your labor line keeps climbing and you cannot say which hours are actually behind it, we would be glad to compare notes. Argus is in private beta with convenience, gas station, and grocery operators. You can talk to us, or write to support@useargus.co.