Why should video be high on loss prevention’s list of priorities? Because a fundamental change has created a wave of innovation and will fuel advances for years to come.
Once, back when Canon made a splash with its “Image is Everything” advertising campaign, it really was. CCTV offered images of sales floors and stockrooms, with the visual information acting as a surrogate for human observation. The pictures provided the value. Cameras offered extra sets of eyes.
It’s a world of difference today. Visual information is no longer the only—and perhaps not even the primary—component of a video surveillance system. Images are still valuable, but it is the data inherent within them that provides for new applications and value. A camera is now a computer with a lens. Video once provided LP with extra eyes—now it also offers brains.
Harnessing this new intelligence has become a true test of loss prevention leadership. And it’s not a stretch to say that a retail organization’s success is increasingly linked to its ability to take advantage of emerging video surveillance tools to reduce loss, enhance situational awareness, improve store operations, aid compliance, and boost sales.
By some accounts, this is an area where loss prevention teams and retailers are falling short. Experts we interviewed on all sides—manufacturers, integrators, and end-users—said it’s not uncommon for many deployed video system features to go underutilized and, worse, completely unused. Some cited specific gaps—between the use of video surveillance tools and what they’re capable of—in the areas of integration, remote monitoring, and analytics. One was harsh, saying it’s as much the rule as the exception for retail security cameras to be positioned incorrectly, poorly maintained, and insufficiently managed.
In some cases, personnel may not be up to the task. As one loss prevention director shared, the video loss prevention technology now available to stores is, in car terms, like a Ferrari, yet we are asking it to be driven by novice drivers. It is not uncommon for retail security guards and store associates to use but a small fraction of the potential capability of the video systems installed. And in the end, a system is reliant on those who have the time, skills, or inclination to use it.
A lack of effective strategic security management can also be to blame. Developments in video technology can outpace loss prevention leaders’ understanding of how to exploit it, and while snapshots like this article may help, truly addressing a potential knowledge gap demands effective and consistent communication between those who follow security technology and those in charge of developing strategies to use it. That said, we think it’s worth noting the issues and applications that rose to the top when we examined new research and solicited expert opinions from end users and video technology providers on where we are, where we’re going, and what pitfalls we may find along the way.
Coverage and Integration
Panoramic HD cameras, remote monitoring, and integration are among the video uses providing LP with significant value. The consensus view is that video surveillance succeeds best in applications where it offers a clear benefit to users, is intuitive and easy to use, and offers robust performance. So how are retailers benefiting from today’s video solutions? Some of the strategies frequently mentioned include:
- Getting More from Less. “High-resolution panoramic cameras, which can cover more of a store more cost effectively, is a video technology that has been providing a very positive practical contribution to users,” said Dan Reese, director of vertical market applications for Bosch Security Systems.
- Going Beyond You. The integration of cameras into store infrastructures—such as doors, store alarm systems, EAS gates, and lighting—creates new opportunities for value creation, and many experts think linking security video with other systems and databases is where you’ll find its greatest value. “Integration is a huge area of opportunity,” said Hedgie Bartol of Axis Communications, a network camera manufacturer. For example, the integration of high-definition video with point-of-sale data enables investigators to more quickly and successfully review video surveillance footage associated with a specific transaction. A retailer can expect, at a minimum, 10 to 20 percent of timesaving compared to traditional investigations, according to Scott Emery of NAVCO, an electronic security system integrator.
- Being Generous. Some sharing of security video is common, such as with risk management to manage slip-and-fall risk, but widespread video sharing across a retail enterprise isn’t always common. One loss prevention executive said it’s important to think of store video as a tool for providing business benefits—not as a departmental asset. Cross-functional corporate executives can use video to obtain feedback on customer traffic, customer service execution, merchandising presentation, procedural compliance, employee safety, and general store operations, among other issues. And there is a bonus benefit, say experts: it’s easier to justify investments in video—and to demonstrate return on investment—when there are more internal users.
Led by analytics, video is more than ever central to strategies to improve sales and cut loss. Video analytics—computer-assisted video monitoring, analysis, and indexing—changes what a surveillance camera network “sees.” Software using signal processing and pattern recognition techniques automatically generate meaningful or semantic data from video images. It is technology that is rapidly improving according to LP and security executives we consulted. They noted specific and significant progress in analytic functions such as motion analysis, object detection and tracking, identification, and activity recognition.
What’s the hottest trend? “Video being used and providing value for more than just loss prevention and expanding to in-store operations,” said Reese. “To make sure things are running smoothly, to monitor queue length, examine dwell times, do calculations on conversion rates, and collect other business data to assist with marketing and merchandising—this is what is up and coming.”
Interest in business intelligence and retail analytics stands to be a boon for loss prevention. The same cameras that can run analytics to measure customer impressions of promotional displays or to generate heat maps of customer traffic patterns are the ones used for loss prevention. “Now you can get dual use out of your cameras,” said Reese. “So now the LP pro has more people potentially funding the cameras.”
Whether or not this promising future is fully realized significantly depends on whether retailers come to find that video surveillance data does, in fact, enable improved retail performance. Can video information, as is currently being promised, enhance customer experience, optimize store performance, reduce operational costs, and ultimately translate into higher profitability? On this critical question, there is good reason to be hopeful.
Researchers at York University in Toronto examined retail sales and video monitoring data collected by video analytics provider i3 International from six US sporting goods stores over five months. Video tracked the times that customers entered and exited stores, their direction, and whether or not a customer entered alone, among other things. Three of the six stores used the data to adjust promotion and sales techniques, while the other three stores did not use the data. The study, “Cameras Tracking Shoppers: The Economics of Retail Video Surveillance,” published in the December 2015 edition of Eurasian Business Review, was unambiguous in its conclusion: “Under our most conservative estimates, the surveillance technology increased total sales per hour by 16.93 percent, increased the number of transactions per hour by 9.69 percent, and increased the average size of each transaction by 9.59 percent. There is less than a 1 percent chance that these findings are due to random chance.”
Improvements in analytics are also poised to aid loss prevention. The hope has always been that video surveillance saves money by identifying thieves and discouraging crime. Proof that it does has been surprisingly hard to come by, however, given the reliance of retailers on CCTV to address store theft. A review in May 2016 of forty CCTV studies by University of Leicester Professor Adrian Beck—Amplifying Risk in Retail Stores: The Evidence to Date—concluded that results paint “a very mixed picture.” Overall, research suggests that visible video deters opportunistic offenders but that “professional thieves are likely to look for ways to defeat it.”
But experts generally agree that advances—specifically, better analytics, the ability of machines to “learn” from their environment what is a noteworthy event, and the ability for cameras to run analytics and detect anomalies at the edge—are improving video’s theft prevention prospects. Video analytics already do a good job outside by identifying blocked fire lanes, or inside to alert staff to blocked fire exits, and they “continue to get better and better,” said Bosch’s Dan Reese. Advances open new possibilities, such as using video analytics to alert when an employee is processing a return when there is no customer in front of them.
Improvements in analytics now allow cameras to track individuals throughout an entire store, using their gait, height, and other personal features to keep them in view. And it’s no longer a fantasy to envision a video surveillance system that can—given enough data sets to go on—identify behaviors that suggest an individual is about to commit a theft and to alert store personnel. New algorithms are being developed that can learn the difference between normal and anomalous behavior. These “deep learning” algorithms could potentially be applied to automatically detect sweethearting and other behaviors related to internal theft. “This is emerging to the point where it’s just becoming practical. It’s something that’s not far off,” according to Reese. Recent scientific studies have shown that novel algorithm approaches significantly reduce false alarms and enhance precision in detecting nonscans versus scans by cashiers at checkout, for example. In short, network cameras plus high resolution plus rapidly advancing analytical capabilities equal a powerful platform to drive value for retailers by both boosting business and reducing loss.
Technological improvements and an expansion in use cases are making facial recognition technology more viable in retail. Using facial recognition, equipped security cameras in a retail location are capable of comparing images of individuals who walk into a store against a database of images. If a match is found, personnel can be alerted and provided whatever information is known about the individual. For safety and security, that can alert an asset protection associate to the presence of known shoplifters, members of organized retail crime syndicates, or other persons of interest, such as a disgruntled ex-employee or a violent ex-spouse. Expanding a system’s database by incorporating images provided by local law enforcement or other retailers can make a system more valuable.
In a study of potential privacy issues, researchers from the Government Accountability Office identified retail as a potential primary user of facial recognition technology but admitted that the extent to which it’s already in use is unclear. A representative of the National Retail Federation (NRF) said many retailers are looking into facial recognition systems for security purposes, but that it didn’t have data on the level of use. The general sense is that retailers in the US are not currently using the technology broadly as concerns over customer reaction has forced a go-slow approach. But as more retailers investigate positive uses for facial recognition, something that retailers in Europe and Asia have been doing, retailers may increasingly test the American public’s willingness to accept the technology.
Facial recognition may have started as a security technology, but consumer commercial applications are growing fast. For example, cameras embedded into digital signs can determine the demographic characteristics of a face, such as approximate age and gender, and show targeted advertisements in real time. In a 2015 survey by Computer Sciences Corporation of 150 UK retailers, including more than half of fashion retailers, 27 percent said they use facial recognition technology (FRT).
Saks has been one of the technology’s early adopters as the company described in a presentation at the 2014 NRF Loss Prevention Conference. Sharing Saks’ experience, Patrick McEvoy, director for AP systems and technology for Hudson’s Bay Company, Saks’ parent, suggested the technology had its limitations but was capable of delivering desired results. The presentation suggested that some of the historical causes for a lack of adoption—faulty technology, too expensive, lack of benefits—were being overcome. In addition to identifying bad actors, stores are finding other ways to use the technology including identifying VIP customers, conducting targeted marketing, collecting data on customer demographics, and for general physical security applications such as access control and capacity management. Again, video surveillance cameras can do double duty—serving security needs and enhancing marketing and customer service.
In discovering a wider range of uses for facial recognition, retailers are following the lead of casinos. Jessie Beaudoin, former executive director of surveillance for Hard Rock Hotel & Casino Las Vegas, explained that the casino has successfully used face recognition for years to spot known criminals, prostitutes, previously ejected individuals, and card counters from its photo database of known “cheats” and problem individuals. Police mug shots of local criminals and discs of facial images sent to them by other properties enhances Hard Rock’s database of people it wants to know are on the property. But the casino also leverages FRT for value beyond security—to immediately identify when high rollers or celebrities come on its premises. This allows staff to greet and tend to them—perhaps point them to high-stakes games in progress—a service that encourages top customers to keep coming back.
It’s that type of customer service application that is expected to drive adoption of FRT in retail. In a December NRF poll, 82.5 percent of retailers said it will be “very important” to personalize the shopping experience in 2017, and facial recognition is perhaps the ultimate personalizing technology. “A valuable customer who has opted into a program can get a personal welcome when he comes into the store and personalized coupons sent to his phone,” explained Reese.
T-Mobile has FRT in more than 1,000 stores and captures tens of thousands of faces per day, the wireless carrier explained in an ISC West conference presentation. But it regards the technology differently than most—not to provide “recognition” but to provide “face capture events” with the goal of associating every transaction with a face. Although it’s not the stereotypical “face in the crowd” application, it provides evidentiary value. In an identity theft event, for example, investigators can search and find the face that goes with the transaction, and then pull up all other occasions when the suspect individual has been in any T-Mobile’s store. That alone often reveals their true information as a database search of a fraudster’s face often returns legitimate transactions when they used their real identity. Also because an account number has a face associated with it, when a different face makes a change to the account, security can be alerted to the anomaly. The technology did not have an immediate impact on T-Mobile’s security headcount, but it did leverage the value of the people it has, significantly improving investigative closure rates, investigation time-to-closure, and financial return per investigator.
Emerging Recognition Technology
In 2017, the most talked-about camera and recognition technology may have nothing to do with faces. Four major retailers, including a convenience store and a big box, are currently in varying stages of testing cameras embedded with an algorithm that pick-ups signals from shoppers’ WiFi- and Bluetooth-emitting devices, like smartphones, Fitbits, and Apple watches. The system assigns each individual a unique but anonymous identifier, which facilitates data collection such as where that individual goes in the store, how long they shop, and—so long as they’re in possession of the same device—whenever the individual returns or visits another of the retailer’s locations. The technology gives retailers that are interested in tracking shoppers—but scared away by the privacy concerns associated with personal identifiers—a way forward. With it, retailers can learn about a shopper’s personal habits and whether they’re someone store agents should be suspicious of, but that knowledge is never linked in a database with unique identifying information, like a Mac address, a name, or a face. The technology’s not new, and its inclusion in surveillance cameras allows retailers to take advantage of it without needing to alter the IT infrastructure.
The technology has obvious marketing potential, and LP executives are finding useful applications as well, according to Jim Paul of ClickIt, which makes the cameras currently undergoing testing. For example, a criminal who covers his face during a robbery may nonetheless be easily caught by searching for other times his device was in a retailer’s location and reviewing corresponding video to see his unobscured face. Or if a known problem individual enters any of a retailer’s locations, store personnel can receive an alert. On the “crawl, walk, run” continuum, Paul says the technology is currently at the “walk” stage—but that it would soon be a game-changing application for major retailers. Technology that can provide unique security and business intelligence but sidestep privacy concerns seems likely to find its way into retailers’ technology arsenal.
Effectiveness Does Not Require Perfection
A willingness to accept less-than-perfect technology is a promising development that may help to drive adoption of new loss prevention tools like facial recognition. “Human recognition systems are inherently probabilistic, and hence inherently fallible,” concluded researchers in Biometric Recognition: Challenges and Opportunities, a report by the National Research Council. It’s not the kind of characterization that is likely to inspire confidence in a retailer flirting with adoption of facial recognition technology. Indeed, according to existing retail users, there are substantial technical requirements and a long list of issues that can complicate implementation and degrade performance:
- Multiple cameras are often needed to effectively capture faces at a double-door entrance,
- Multiple servers are often needed to run face capture and database matching,
- Database photos must be of a certain resolution for optimal use in matching live video, and
- Accurate system testing is really only possible in the live, operational environment of real-world users.
Sometimes, warn researchers, improving the performance of one component of a facial recognition system degrades performance of the larger system. For example, upgrading a face image detector can allow a system to find more faces in more poses, but then the number of off-angle and partial faces sent into the comparison process increases. Additionally, there is no way to definitively determine the impact of component changes on system-level performance until the components have been inserted and the system is tested as a whole. Finally, with scant data on operational evaluations of FRT, retailers are often dependent on claims made by the FRT vendors themselves.
For a retailer conditioned by Hollywood interpretations of technology, the many caveats may suggest that technology like FRT is simply “not ready.” However, perhaps the best way to approach new video applications are not with a mindset of pass-fail based on their original promise, but instead to recognize what these tools actually can do—and to decide whether those capabilities can return value to your security and retail operation.
That approach may be growing increasingly widespread. Several Super Bowls ago, facial recognition was hyped for its use to spot terrorists, promotion that didn’t do the industry any favors, according Stephen Russell, CEO and founder of Prism Skylabs and former CEO of 3VR Security. Hype obscured the fact that even if 88 percent accuracy is as good as we can ever get in such a “face in the crowd” scenario, FRT can still pass a cost-benefit test. Ask yourself whether at that percent of recognition you would find bad guys you would not find otherwise. Then ask if the volume of investigating “possibles” would be too great for your staff to handle. If “yes” and “no,” respectively, are your answers, then the technology is useful despite a lack of absolute precision.
While it may sound like an industry trying to lower the bar, there are countless case studies of companies extracting real value from less-than-perfect video technology applications. For example, people counting functions may not be 100 percent accurate—people huddled at exits will cause a miss here and there—but the data may still provide for more informed business and security decisions. As advanced video surveillance tools have been increasingly used for business, there has been more variation in the way end users characterize “effective,” said one industry expert.
Following are some additional insights from new research and perspectives that appear to be widely shared on the state and future of video:
- Measuring customer response to products and understanding their level of engagement is necessary for the future of brick-and-mortar retail.
- On-board processing power of video surveillance cameras will continue to increase, with more applications being performed at the edge.
- Collecting data on shopping experiences can be done by cameras and image processing algorithms, “but the appearance of new kind of sensors like iBeacons or radio frequency identification [on shopping carts, for example] is leading to new hybrid systems with great performance,” according to “Improving Retail Efficiency Through Sensing Technologies,” a study published in the October 2016 issue of Pattern Recognition Letters.
- 4K video surveillance is accompanied by extensive marketing hype, according to IHS, a market analysis firm in a research report, Top Video Surveillance Trends for 2016. “Yet make no mistake, the video surveillance market is going to 4K cameras; it’s only a matter of when rather than if.”
- Mundane factors like tight budgets are restricting adoption of advanced video technologies, but there is also a level of fear that, at times, their promise is oversold and that systems try to do too much, which harms performance and reliability.
- Analytics and more efficient compression technologies can effectively reduce video storage requirements, but alone can’t offset the burgeoning demand, according to IHS. An approach incorporating multiple storage types will often be the most cost-efficient solution, the firm concluded. “In its most basic form, this means a combination of cloud and local storage with one unified platform.”
- The trend of linking of public and private video surveillance networks will accelerate.
Will You Benefit?
The big change noted at the outset—that video cameras have moved from single-function pieces of store equipment to a platform technology with limitless business potential—requires effective management to leverage value. Enhanced video functions are less meaningful if they outstrip the ability of loss prevention departments to make effective use of them.
As camera technology has improved and options have expanded, it has become more important, when purchasing video solutions, to have clearly articulated goals and strategies. It’s more important now than back when a camera was just a camera and loss prevention was the sole stakeholder. As such, experts frequently identify collaboration as a linchpin to success so that all potential stakeholders can provide input on how they might be able to leverage a retailer’s investment in advanced video technology. LP departments should also be wary of letting capabilities “drive the debate,” warned one security executive. The focus should be on how those features can be effectively used to benefit your LP operation.
Experts offered additional advice for squeezing maximum value from future video investments:
- A retailer’s video technology buying process becomes more important alongside the growth in the number of functions it performs. That makes the tests retailers perform of new video technology increasingly important. Said one executive we spoke with, “Staged product demonstrations always go perfectly. They are often done in a fixed environment and are well rehearsed, so I take them with a grain of salt. I like to see products proven under my own set of conditions.” Unless you thoroughly vet systems under the most extreme conditions in which they are going to operate, you won’t get the balance right—between false alarms and detection, for example—and may even select the wrong system.
- A staunch dedication to open architecture and a willingness to investigate manufacturer claims of it is important for capitalizing on advances in video technology no matter who develops it. “You don’t buy the camera for the video feed solely. You want all the other nice features,” said Mike Dunn, a vice president at Best Security Industries (BSI). “But you may not know how your retail organization will want to make use of them for months to come,” he warned. Open systems can accommodate those future demands.
- Utilizing expert counsel from integrators is valuable even as systems become easier to use. It may be easier now to tackle sophisticated video projects without help from integrators, but they can still serve a valuable role in a plug-and-play world. For example, one security pro told us that because end users frequently focus on “capabilities over practicalities,” integrators often return value to a project.
- It is incumbent upon loss prevention executives to ensure that connected video devices don’t become a threat vector. Experts we consulted said it’s not sufficient to hope that integrators become more diligent or that vendors do more robust testing. According to research presented in October at the 6th International Workshop on Trustworthy Embedded Devices, studies on surveillance systems have revealed vulnerabilities that have had a “large-scale impact in real life,” including theft from retail organizations. The study, Security of CCTV and Video Surveillance Systems: Threats, Vulnerabilities, Attacks, and Mitigations, offers a clear warning to LP executives: “The variety of vendors and vulnerabilities disclosed in those studies and security advisories clearly indicates the unhealthy state of cyber security of video surveillance systems.”
- Retailers should have a strong baseline of privacy for use and retention of video as concerns will undoubtedly grow alongside more “futuristic” applications. It’s easy to imagine how a smart video camera system—capable of identifying shopper demographics and tracking individuals around a store—might spark lawsuits and raise concerns for the protection of images and data it records. Federal law does not expressly address the use of facial recognition technology by the private sector to identify or track individuals, but that only reflects the slowness of government to adapt laws to reflect new technologies, not necessarily a lack of interest in doing so. With the potential of increased liability and regulations on the horizon, LP executives might want to track industry self-regulation efforts and implement, adhere, and adjust privacy policies accordingly. For example, the Point of Purchase Advertising International’s Digital Signage Group has developed a code of conduct for data collection, the International Biometrics & Identification Association has issued IBIA Privacy Best Practice Recommendations for Commercial Biometric Use, and the Federal Trade Commission has issued a best practices document entitled Facing Facts: Best Practices for Common Uses of Facial Recognition Technologies.
- Tracking general advances in video-related applications and technology helps prepare loss prevention departments to utilize them. Video technology is evolving rapidly, and several experts we interviewed recommended frequent brainstorming as a way to help LP departments capitalize as technologies mature and applications commercialize—and to help avoid potential problems and obstacles.