Notes from a recent webinar on how top retailers are winning on shrink, and what the data actually showed about repeat offenders, investigations, and the case for facial recognition in loss prevention.
By Craig Greenberg, Chief Growth Officer, Gatekeeper Systems and FaceFirst®
I spent an hour recently on a webinar with Greg Buzek of IHL Group and Dr. Cory Lowe of the Loss Prevention Research Council (LPRC). The topic was how top retailers are winning on shrink. The data Cory walked us through was the kind that sticks with you, and I want to share what I took away from it.
Before I get into the numbers, a quick frame. I have been working in retail loss prevention for over 30 years, and Gatekeeper has been focused on two things that drive the highest per-event losses retailers see: pushout theft and repeat offenders. About 15 months ago, we added FaceFirst to that picture. What we have learned since then is that the visibility problem in loss prevention is bigger than most retailers realize, and the cost of not seeing the full picture compounds quickly.
The journey of a single offender
For years, Gatekeeper has stopped pushout attempts at the door. A cart locks, the would-be thief walks away empty-handed, and a store associate logs the incident. That is a good outcome on its own. What we did not see was where that same person went next. Did they hit another store the same afternoon? Did they come back the next week? Did they then visit stores two states away?
FaceFirst answered that question for us. LPRC’s analysis of a FaceFirst client’s data confirmed what we anecdotally knew – repeat offenders drive a disproportionate amount of the loss dollar volume. LPRC helped us quantify just how meaningful the impact of a few repeat offenders can be. These are not opportunistic thieves grabbing a cart on a bad day. They are professionals running a route, and they are organized.
The LPRC research provided hard numbers on what we instinctively understood. In one 60-day study, the top repeat offender had 374 FaceFirst probable match events across 66 different stores spanning five states. An average of more than six per day. The path of travel was clear and shocking, once you looked at it. On one day, a repeat offender hit 12 stores, traveling 241 miles over a nearly five-and-a-half-hour stretch. That single offender’s potential losses extrapolated out to numbers a retailer cannot ignore.
Building cases prosecutors will actually take
Visibility is the first benefit. The second one matters just as much, and it gets less attention. The data builds a case.
Cory walked through a study of two equally qualified investigators assigned to the same retail case. One was unassisted by facial recognition technology. The other had FaceFirst. The unassisted investigator spent 904 minutes on the case, reviewing video from store after store, plotting paths manually, and trying to connect incidents. The investigator using FaceFirst spent just 117 minutes, but his case value came in at 4.3 times higher, and he identified the subjects in twice as many store locations.
That gap is not a minor efficiency gain. It is the difference between a case that ends up on a prosecutor’s desk and one that does not. Law enforcement and prosecutors have to triage. They prioritize the cases that are serious and well-documented, because their time is finite. If we hand them a thin file with one incident, they cannot do much with it. If we hand them aggregated evidence showing the same offender hit 12 stores over six weeks, with timestamps, vehicles, and loss values attached, the calculus changes.
This is where Purchek® and FaceFirst together do something neither tool does alone. Purchek stops the cart. FaceFirst can help connect that cart-stop event in one store to that same individual’s other offenses across the entire enterprise. The retailer gets a turnkey case. The prosecutor gets something worth pursuing. The community gets a repeat offender off the street for a while.
The conversation we still need to have
I will be honest about one thing. The biggest barrier to facial recognition in retail right now is not the technology. It is the public’s understanding of what the technology actually does.
My father-in-law is in his 90s. When I told him Gatekeeper had acquired FaceFirst, his reaction was, “I don’t know if I would like that. They will know my name is Jerry and where I live.” That was a moment for me. If a thoughtful retiree thinks the grocery store is going to start greeting him by name from a facial scan, we have not done a good enough job explaining how this works.
FaceFirst does not index photos of every customer who walks in. It does not know names. It works from a biometric template against a private database of known offenders that the retailer maintains for safety and security purposes. It alerts a trained and designated associate when someone with a documented history of violence, threats, or theft walks in. The data for everyone else who passes through is deleted automatically. The data retention limits, the human oversight requirements, and the authorized use audit trails are all built into FaceFirst.
The LPRC research Cory presented about bias and accuracy reinforced something I had been thinking about already. All models have bias. All people have bias, too. The honest question is not whether AI is perfect. It is whether humans assisted by AI are more accurate and less biased than humans on their own. In the LPRC study, loss prevention practitioners who did not have facial recognition assistance identified the right offender only 23 percent of the time. With FaceFirst’s assistance, they were correct 62.9 percent of the time — nearly three times more accurate, across every demographic group. Researchers also noted that with more training on the technology, those results would improve further. That is the conversation worth having.
Where this leaves us
Greg made a point during the webinar that I keep thinking about. The sales winners in retail are 33 times more likely to be using computer vision than the average retailer. That is not a marginal gap. It tells you something about where the industry is heading and who is going to be left out of the next chapter.
For loss prevention specifically, the case is no longer just about stopping the theft at the door. It is about understanding the network, building cases that hold up in court, keeping associates and customers safe, and giving prosecutors something they can run with. The tools to do all of that are here now. The harder work is helping the rest of the world understand what they actually are, and what they are not.
I am grateful to Greg and Cory for the conversation, and to the LPRC for the rigor they bring to this work. If you missed the live session, the slides and research report are worth your time – you can find them here.
About the author
Craig Greenberg is Chief Growth Officer at Gatekeeper Systems and FaceFirst, a retail loss prevention technology company specializing in pushout theft prevention, repeat offender identification, and facial matching for retail safety. Gatekeeper Systems has supported retailers in cart-based loss prevention for more than 30 years.
