Hannaford Bros. and Pathmark are using intelligent systems that interpret video images to identify employee theft and shoplifting
At any major casino, trained security professionals are stationed in a video surveillance room closely observing the activity at the high-stakes gaming tables for hints of fraudulent behavior that would put the casino at risk.
While the individual stakes are decidedly lower at supermarkets, the cumulative impact of fraudulent activity — by shoppers and especially by store employees — weighs heavily on the industry. The Food Marketing Institute's 2007 Supermarket Security & Loss Prevention report put the median shrink figure for food retailers at 1.52% of sales (down from 1.69% the previous year); about 39% of shrink is attributed to employee theft, while 32% is tied to shoplifting.
Like casinos, supermarkets use video technology to deter fraudulent activity. But since food retailers can hardly afford a full-time video surveillance team, they are increasingly turning to video systems that have built-in intelligence — or are supplemented by intelligent software — that makes it possible without continuous human oversight to detect illegal behavior on the part of shoppers or employees, especially at the checkout lanes.
“We can't pay someone to watch 12 registers at 100 stores,” said Tom Perkins, director of loss prevention at Hannaford Bros., Scarborough, Maine. “We want technology to tell us [when something happens].”
Of course, closed-circuit television (CCTV) cameras have long been a staple of the supermarket environment. In recent years, retailers have upgraded from analog to digital video cameras and recorders, making it easier to locate video images and tie them to POS exception reporting data, which identifies potential employee theft issues.
But advances in video technology are now allowing retailers to bring more powerful tools to the task of tracking and preventing shopper and employee theft, addressing problems as they occur or shortly thereafter, and using far less human oversight.
Hannaford Bros., for example, is testing software from StopLift Vision Systems, Bedford, Mass., that analyzes video feeds from overhead front-end cameras to determine whether cashiers are engaging in “sweethearting” — deliberately moving an item through the front-end without scanning it. Big Y and Safeway are also evaluating the system.
StopLift's software “watches how cashiers interact with merchandise, tracking fine motor skills, and distinguishes between what they should and shouldn't be doing,” said Malay Kundu, chief executive officer and founder of StopLift, who began developing the video recognition system while a student at Harvard Business School.
The system, for example, can detect when a cashier passes an item around or over the scanner or faces the bar code away from the scanner. It rapidly dispatches reports, identifying the cashier and the date and time of the incident. Other kinds of theft, such as void and coupon fraud, can also be tracked.
Sweethearting has proved to be a particularly vexing problem for retailers because it doesn't show up consistently on exception reports, which can only track scanned items. StopLift's system “is the first product I'm aware of that really gets at sweethearting,” said Hannaford's Perkins. “This is one of the most significant leaps in loss prevention technology I've seen in more than a decade.” He put the cost of sweethearting to the industry at $13 billion per year.
Perkins pointed out that the StopLift system flags each instance of sweethearting whereas exception reporting only spotlights fraudulent activity that has built up over time. As a result, StopLift is able to target individuals involved with a minimal amount of theft — say $5 per day — “who operate under the radar,” he said.
Hannaford has been testing the StopLift system for about a year, Perkins said, declining to say how many stores are in the test. “It's worked very well. It's identified sweethearting that we normally would never have known of.” Hannaford does not have the staff to react to StopLift's reports immediately, but incidents are addressed in short order, and no more than a week after they take place, he said.
Hannaford plans to install the system in an unspecified number of additional stores, Perkins said. He declined to comment on the cost of the system.
Perkins said that Hannaford initially used the system to identify instances of sweethearting without acting on the information. “If we had let the cat out of the bag, we would have had no more data to collect,” he explained. Hannaford is still deciding how to discipline employees found to be sweethearting.
After employees at the Hannaford test stores were made aware of the system, very few instances of sweethearting turned up. “Now it's a deterrent, which is what we want,” said Perkins.
Hannaford is also using the system as a training tool to help cashiers who are not deliberately sweethearting but are still failing to correctly scan items.
In all of the stores testing StopLift (including Hannaford's), an average of 30% of employees were determined to have engaged in sweethearting at least once, according to Kundu. “It was lower at Hannaford, but I'm not surprised the average was 30%,” said Perkins.
Another new video technology that is attracting interest in the supermarket industry is the LaneHawk Visual Scanner system from Evolution Robotics Retail, Pasadena, Calif. This system employs a video camera placed near the floor of the checkout lane to monitor whether products are being carried at the bottom of the shopping cart. Using what's called visual pattern recognition (ViPR) technology, the system is able to not only tell that a product is under the cart, but identify the product and bring it to the attention of the cashier, who asks whether the shopper wants to purchase it or not.
ViPR technology, adapted from robotics and military applications, uses a process that is akin to fingerprinting, explained Alec Hudnut, CEO of Evolution Robotics Retail. The packaging for each product has a unique set of visual characteristics that the system records and is able to recognize when it sees the item in the lane.
The system's database stores the characteristics of more than 600 common bottom-of-the-basket (BOB) items such as large bags of dog food and 12-packs of Diet Coke. Sometimes an item may be identified as belonging to a family of items, or not identified at all; in these cases the cashier asks the shopper to fill in the blanks.
Evolution Robotics Retail has so far publicly disclosed the names of two retailers using LaneHawk — Pathmark Stores, Carteret, N.J., a division of A&P, and Shoppers Food & Pharmacy, a division of Supervalu serving Washington, Baltimore and northern Virginia. But Hudnut said that 1,000 or more stores will have the system deployed by the end of 2008. It costs under $2,000 per lane, with a return on investment of less than a year, he said.
Pathmark has implemented LaneHawk in 10 of its stores, starting installations last September following an extensive pilot period. Bob Oberosler, senior vice president of Asset Protection for Pathmark, said one of the strengths of the system is that it requires the cashier to respond to the presence of a BOB item in order to complete the transaction. “We tried other BOB technology like cameras and bells but by the second week the cashiers would just ignore it,” said Oberosler, who is leaving Pathmark at the end of February as a result of its merger with A&P.
Like StopLift, LaneHawk has a deterrent effect on theft after being in place for a while. “Once the customer is asked about BOB items a few times, they start putting the items on the counter,” said Oberosler.
Oberosler observed that LaneHawk has had a substantial “anecdotal” impact on BOB items during major promotion periods. The system has prevented enough BOB losses in those cases that the chain has been able to reduce the “on-hand” adjustments needed to compensate for lost product by 90%.
Pathmark has also observed overall “inventory shrink improvement” at the one or two stores that have had the system the longest, said Oberosler, who declined to be specific. In a statement released by Evolution Robotics Retail last September, Pedro Ramos, in Pathmark's loss prevention department, said that since Pathmark started using LaneHawk in its initial stores, it has “been able to reduce BOB shrink by 90%, resulting in a 50 basis point improvement in our grocery shrink.”
This summer, Evolution Robotics Retail plans to introduce another loss prevention product, VideoHawk, which will track POS activity through cameras aimed at cashiers and shoppers. The system will use ViPR technology to identify items that are not being scanned because of sweethearting or because they have been left in the shopping cart; it will also identify items that have an incorrect bar code.
Hudnut pointed out that the ViPR technology, in both LaneHawk and VideoHawk, works to address potential fraud in real time, alerting the cashier immediately of items that have not been scanned.
Pathmark uses LaneHawk in concert with another loss prevention technology, StoreVision, from Agilence, Camden, N.J. StoreVision captures scan data on every item at the POS and ties it to a 10K video image of each scan, as well as a video of an entire transaction.
Pathmark, for example, was able to ask StoreVision to call up examples and images of large orders with BOB clears, which “helped us identify sweethearting cases,” said Oberosler. StoreVision has also helped the chain identify cash loses as well as credit card, check and coupon fraud, and new theft trends that emerge.