A truck backs into the lot on a Thursday morning. Twelve cases of sports drink come off it, the driver hands over an invoice for twelve, and somebody signs. Eleven cases go to the back room. That night the invoice gets keyed for twelve, because twelve is what the paper says.
Nothing in that sequence looks wrong inside any one system. The register sold what it sold. The back office received what the invoice claimed. The camera recorded a delivery that looked like every other delivery. Three months later the count comes back short, and by then four hundred more deliveries sit between the store and the answer.
Most of what a store loses does not happen inside a system. It happens in the space between two of them, and that space does not belong to anybody.
A third of shrink is nobody stealing anything
Appriss Retail published its 2026 Total Retail Loss Benchmark Report from a survey run in December of 2025. It puts United States shrink at $90 billion and splits it by cause. Employee theft is 29 percent, about $26 billion. Inventory errors are 21 percent, about $19 billion. Operational inefficiencies are 13 percent, about $12 billion. Organized retail crime is 10 percent, about $9 billion. Returns account for another 20 percent, and 7 percent has no explanation at all.
Read the middle two again. Inventory errors and operational inefficiencies come to 34 percent of shrink, roughly $31 billion, and neither one is theft. Nobody took anything. A quantity was keyed wrong. A vendor credit was never claimed. A price changed at the register and not in the back office, or in the back office and not at the register. A case was received on paper that never came off the dock.
What makes those two categories interesting is what does not exist around them. There is an industry built to sell you something against the 29 percent and the 10 percent, and you already know what it looks like: cameras, mirrors, locked cases, a guard on Friday night. There is no product category called the invoice was keyed wrong. No alarm sounds. No one walks out of the building.
One honest caveat. A benchmark built from surveyed retailers is not a measurement of your store, and the split between causes depends on what each retailer could attribute in the first place. That is closer to the point than an argument against it: the categories near the bottom of that list are defined by what nobody could explain.
A wrong number costs more than the missing item
The obvious cost of a bad inventory record is the item you no longer have. The larger cost is the one nobody sends you an invoice for.
Four researchers published a study in 2025 on inventory record inaccuracy in grocery retailing, covering roughly 24,000 items across 11 stores. They ran a field experiment to see what changes after an audit corrects the records. Counting the store produced an 11 percent store wide sales lift, and nearly all of that lift landed on one specific kind of item: the ones where the system believed stock was there and the shelf was empty.
That is the second loss, and it is quieter than the first. The system thinks it has eight, so it does not reorder. The shelf is empty, so nobody buys. The store is short the item and then short every sale of that item until the next count. The authors argue that a stock count is better understood as a way to increase sales than as a cost you absorb once a year. They also found inaccuracy runs higher on perishable goods, which in a convenience store means the cooler, the deli case, and anything with a date printed on it.
The gap is widest where there is nobody to close it
This is not a problem small operators invented. PMC and Retail Economics surveyed more than 100 senior retail and brand leaders for a study called Race To Unified Commerce, published in July of 2026. Seven in ten direct to consumer brands said their core systems are still not fully integrated. On the effects, 56 percent said fragmented systems were hurting profitability and customer experience, and 54 percent said operational effectiveness had suffered.
Those are companies with technology teams and integration budgets. Now look at who runs American convenience retail. NACS counts 151,975 convenience stores in the United States, and 95,672 of them, 63 percent of the total, belong to a company that operates ten stores or fewer. More than 122,000 of them sell fuel.
A store with fuel has more systems than a store without one, not fewer. A dispenser controller, a register, a back office package, a fuel supplier feed, an ATM, a lottery terminal, a food service system, a scheduling tool, a camera recorder. Each was bought in a different year, from a different vendor, to solve a different problem, and each one is competent inside its own boundary. Nobody chose a fragmented store. It accumulated one purchase at a time, and in a company with four locations there is no person on the payroll whose job is the space between two vendors.
What joining them actually looks like
The useful version of this is not another dashboard. It is one shared timeline that several systems write to, so a moment in the store carries more than one record of itself. That is what we build, and it runs on the cameras and registers a store already owns: CASH crossed with POS crossed with CAM crossed with LABOR, continuously.
In practice it means being able to ask a question that crosses a boundary, which is what no single system can answer today:
- Did the quantity we received match what actually came off the truck, at the door, on the video, at that hour?
- Is this shelf empty during the same days the system says we are holding eight of them?
- When the price at the register disagrees with the price in the book, which shift was on, and how many times has that happened this month?
Two things this does not do, and we would rather say them plainly. It does not fix a receiving process. It makes an existing one visible, and the repair is still a conversation with a driver and a change in how somebody signs a slip. And it does not decide that a person did something wrong. It reports that a discrepancy has now appeared nine times, that seven of them fall in the same delivery window, and it sets the records beside each other so a human can look. Most of the time the answer is not theft. It is a step somebody skips because skipping it has never yet cost them anything they could see.
That is the whole argument for joining the systems. Not that it catches more thieves, but that it gives the third of your shrink that nobody stole somewhere to show up.
If your counts keep coming back short and no report you own can tell you where it went, 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.