A crew comes through a convenience store on a Tuesday afternoon. Three people, about four minutes, two shelves of razor cartridges and pain relievers into a bag, out the door before anyone behind the counter finishes deciding what to do. The camera over the aisle records all of it, in good light, from a useful angle. The clip gets pulled. A report gets filed. Nothing else happens, because one clip of three strangers is not a case.
Six weeks later the same crew works a store eleven miles away. That store files its own report, with its own clip, from its own camera. Each store now has an incident. Nobody has the pattern.
That gap is where organized retail crime actually lives. It is worth saying plainly that it is not a gap in camera coverage.
The national numbers got smaller, and then quieter
Before anyone reaches for a statistic here, it is worth being honest about what the industry knows. In December 2023 the National Retail Federation revised its own organized retail crime report to remove the claim that nearly half of the previous year's shrink was attributable to organized retail crime. The figure had traced back to a 2016 survey of inventory losses from every cause, including plenty of causes that had nothing to do with theft. NRF withdrew the assertion, and with it any dollar estimate of what organized retail crime costs the industry.
Then the broader measure went away too. In October 2024 NRF confirmed it would stop publishing its annual shrink report, ending a series that had run for more than three decades. Its final edition put shrink at 1.6 percent of sales, or $112.1 billion, for fiscal year 2022, a total it reached by applying a self reported rate to all of US retail. An NRF spokesperson explained that "a broad study about retail shrink is no longer sufficient for capturing the key challenges and needs of the industry." That 1.6 percent is still the most recent national shrink rate anyone can point to. It describes a fiscal year that ended four years ago.
The research that replaced it tells a third story. NRF published The Impact of Retail Theft & Violence 2026 on July 30 of this year. Surveyed retailers reported, on average, a 12.4 percent decrease in shoplifting incidents and an 8.1 percent decrease in merchandise theft incidents between 2024 and 2025. The report carries no shrink rate at all. Those averages come from 66 retail companies representing 143 brands, nearly all of them operating at a scale where a dedicated loss prevention team is normal.
Put the three together and the conclusion is not that theft stopped. It is that a national average was never going to tell one operator anything useful about one store. The number that got quoted for years turned out to be wrong, the series that produced the honest number got retired, and the survey that replaced it describes companies that look nothing like a single store on a corner. If you run three stores, none of this is your data.
The law is moving toward aggregation
While the measurement got quieter, the legal machinery got more specific. It is now asking for something most stores cannot produce.
The Combating Organized Retail Crime Act of 2025 passed the House on May 12, 2026 and went to the Senate the following day, where it was read twice and referred to the Judiciary Committee. It has not passed the Senate, and it is not law. What matters for an operator is the mechanism it proposes. Federal theft statutes today turn on a $5,000 threshold met in a single transaction. The bill would add an alternative, inserting goods "of an aggregate value of $5,000 or more during any 12-month period" alongside the existing test.
That is aggregation, and the states got there first. The bill's own findings note that since 2022, "more than 30 State laws have been enacted to address organized theft, allow for aggregation of thefts, and adjust penalties and enhancements."
Aggregation quietly changes what evidence has to be. One four minute incident under the threshold is a police report that goes in a drawer. The same crew, twelve incidents, five stores, eleven months, is a case. Getting from the first thing to the second is not a surveillance problem. It is a records problem: can you show the same people, using the same method, across dates and locations, with each incident tied to what the register did at that moment?
What a camera cannot do
A camera is an excellent witness to one event and a poor witness to a pattern. The reasons are structural, and no amount of resolution fixes them.
- Footage sits on a recorder until somebody has a reason to pull it, and the reason usually arrives after the loss.
- A clip has no transaction attached. The video cannot tell you whether the merchandise that left was paid for, returned, voided, or simply carried out.
- Each store's recorder is its own island. Nothing compares a Tuesday here against a Thursday eleven miles away, because nothing is looking at both.
- Retention runs out. By the time a series of small incidents looks like a series, the earliest clips have often rolled over.
- Somebody has to watch. A manager running a store does not have spare hours to scrub video for a pattern they do not yet know exists.
None of that is an argument against cameras. It is an argument that the recording was never the missing piece.
The join is what makes a pattern visible
What turns a pile of clips into something an operator or a prosecutor can use is the join: the incident matched to the transaction record, the transaction matched to the shift, and all of it comparable across every location on the same timeline. 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.
In practice it means an event is not filed as a clip. It is filed as a record with a time, a lane, an operator, a basket, and a location, which makes the ordinary question answerable: has this happened before, here or anywhere else we run? A single store cannot answer that. A group of stores sharing one data layer can, and the answer takes seconds instead of an afternoon.
We would rather be careful about what this is. It does not identify anyone, and it does not decide that a person committed a crime. It surfaces that the same method has now appeared eleven times across four stores, with the transaction records that go with it, so a human can look at something real. Often enough the pattern turns out to be a broken process rather than a crew, which is its own kind of useful, and a good deal cheaper to fix.
If you run stores and you suspect the same thing is happening in more than one of them but cannot prove 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.