The extent to which retail investigations have been transformed by analytics is a good reflection of just how quickly things move in a technology age. We’re not so many years removed from when we were introduced to the idea, and now it’s hard to think of managing loss prevention without it.
“I really can’t imagine a world of AP without data analytics,” said Nathan Bandaries, a twenty-six-year LP veteran who manages the fight against organized retail crime for Safeway in the Denver area. Abraham Gonzalez, CFI, LPC, fraud mitigation manager at Bloomingdale’s, agreed: “Analytics are starting to consume how you conduct every investigation.”
Data analytics act as a force multiplier, providing visibility where actual eyeballs have previously been needed. For example, Gonzalez noted that pass-off cases, in which an associate slides merchandise to a coconspirator, would typically be difficult without physically witnessing the hand-off, and many other schemes have been just as reliant on someone being at a station to view it. “Now, transaction times, when they begin and end, how many items are cancelled or line voided during the transaction, and noting when it’s outside the normal range—now you have data analytics that let you see it.”
LP teams are increasingly getting used to this new way of “seeing” theft and conducting investigations. In interviews with LP leaders, the consensus view was that, as an industry, we are getting better and more comfortable with investigative analytics; data analytics are generating a rising share of cases taken up by AP investigators; education is improving, leading to more savvy professionals; and we’re capturing more data points, which is letting us be more predictive. Analytics have also been key to burnishing LP’s image, transcending the security guard stereotype and providing a platform for LP to better partner with other functions.
The relationship with technology vendors is also in a positive cycle, experts believe. For example, video can now yield actionable data with analytics while returning fewer false positives, so retailers are spending less energy and gaining greater trust in the data. And as retailers place greater confidence in the technology, technology companies are willing to invest more in building better, easier-to-use LP tools, which spurs even more trust in, and greater reliance on, analytics.
Home Depot is among leading retailers upping their investment in analytics to better leverage the power to identify problems, combat theft and organized retail crime (ORC), and identify opportunities to improve the business through more predictive modeling. “On the reactive side, we spend time reverse engineering how one person caused a problem and recreating it to see who else may be engaged in the same activity. Because if one person is using an opening to commit a bad act, then chances are that others are too,” explained Nate Cunningham, director of AP forensics at the nation’s largest home improvement retailer. “And we run proactive patterns on data, looking for ways people may be taking advantage. And sometimes you miss, and sometimes you hit, which then allows you to create an ongoing alert for that activity.”
A critical part of driving effective investigations through data analytics is, well, data. The more, the better, according to Scott Pethuyne, senior manager for asset protection at Designer Brands. “We’re either doing everything exactly right, or we’re collecting data,” Pethuyne explained during a presentation at the 2019 NRF Protect conference in Anaheim.
As Home Depot has undergone some data infrastructure transition in recent months, this has made more data available for data analysts on its AP forensics team—but it’s not a process that retailers “complete,” suggested Cunningham. “We’re always adding new data,” he said. “It’s a never-ending process, especially as new information is collected and connected and as we come up with new ideas.”
Lisa Brock, national director of investigations at Best Buy, made a similar point in her NRF Protect presentation, noting that it’s important to open and tap different data streams to combat theft trends as they emerge. Each new retail sales idea—such as buy online, pickup in store—brings risk with it, and while it’s often “hard to get analytics around it,” data is critical for devising countermeasures, said Brock.
For example, she noted that Best Buy investigators are in a better position to combat porch pirates and other losses in ship-from-store orders because they track data on which stage of the order a package was lost, such as after customer delivery or after packaging but before shipment pick-up. Similarly, she said it’s beneficial to work with human resources to get employee home address data to check against ship-from-store orders for an alert if employees ship orders to themselves.
Analytics have enabled Safeway’s AP team to build higher-quality cases, but it’s not the only benefit. “We’re having another positive effect, in that we’re also able to react to cases far faster than previously,” said Bandaries. “Before analytics, you might have had someone getting away with their fraud for a year, but now you’re recognizing it after a couple of times, way faster than in the past.” Data is the difference, highlighting discrepancies as they emerge instead of waiting for word of mouth, a tip, or stumbling across an inconsistency by accident.
But even if data is laying the groundwork for investigations these days, it is still people driving the show.
The People Piece
While essentially a numbers game, the leveraging of investigative analytics remains a very human endeavor. It requires the marriage between number crunching and the human intelligence that investigators bring to the equation. “When I think of analytics, I think of the partnership between the data analytics staff and the investigative team,” explained Bandaries. “Without that partnership, we wouldn’t be able to do what we do.”
Gonzalez similarly stressed the value in strong partnerships, noting that many LP departments hand off information between analysts and investigators, “but a lot can be lost in translation,” he warned. But when there is partnership, synergy, openness, and alignment—with both teams learning from each other’s skill set and becoming well versed in the other’s function—better investigations are inevitable.
“It’s the concept behind quality over quantity. You want that quality in your investigations and not just have data analysts hand you a sheet,” said Gonzalez, who has fifteen years of experience in e-commence and ORC investigations, software implementation, and data analytics. “Here at Bloomingdale’s, it’s okay to have an opinion, to challenge one another, for investigators to challenge data analysts and vice versa.”
It can help to have central investigative teams and data analysts work in close proximity, suggested Home Depot’s Cunningham—and, if not physically, then at least through process alignment. “It’s not workable to have groups working in different silos; you need to be in lock step,” he said. “You need a good communication path, to regularly share ideas and concepts, with both sides learning from the other.” It’s also important to enhance the connection between data analysts and business partners to understand loss and shrink trends and changes in risk. “The organization needs to be aware of the importance of keeping AP in the loop of issues that could create exposure,” Cunningham added.
Onboarding the right talent to the analytical team that supports AP is also playing a critical role in helping Home Depot get investigative value from analytics. The perfect candidate—a brilliant data scientist with extensive LP experience and expertise—is a “unicorn,” Cunningham said, so they look for skill and adaptability and toss in continuous learning. “We look for that strong statistical background, but we also find people who show they can be a quick study and then pair them with AP investigators. And that concoction has proved to be a good recipe.” Because of the value data analysts bring, keeping and retaining talent is also a growing challenge facing LP departments.
To provide even more data support to investigators, Home Depot has recently added new staff positions in its three AP investigative divisions. “It allows us to take advantage of analytics capabilities that field investigators—running and gunning out in the field—don’t have time to dig into,” said Rory Stallard, senior manager of AP investigations in its western division. The new hires reflect the company’s continued investment in analytics and its drive to be more efficient. “We should be able to work smarter, get results with less effort, connect dots that we may have missed, and accelerate our closure rate,” said Stallard.
Data analytics are ingrained into the AP group at Safeway, a banner within the Albertsons Companies. Most divisions employ a full-time data analyst in the AP department, working forty hours a week, as well as analysts on the corporate team—and they play an outsized role in generating cases. “It produces a large yield for us,” said Bandaries. “We rely on data analysts and analytical tools to conduct investigations; the role of analytics in our AP world is huge.”
He described a typical case flow in today’s retail environment. Data analysts are constantly monitoring common presets for excessive voids, for example, as well as creating their own data sets to monitor for normal and abnormal activity. When analysis identifies out-of-range activity, the analytics team digs in, teasing the story out of the data, the plot from the numbers.
“That’s when the true analysis and research happens,” said Home Depot’s Cunningham, noting that data analysts play a critical role in identifying whether discrepancies are malicious in nature, present a training opportunity, or if “looking below the surface may unwind a different huge opportunity.”
The initiation of an ORC case is typically different than in years’ past, noted LP executives. Some cases are still born from incidents in stores, but it has become more common for investigations to be pushed forward by trends observed in data. Analytics similarly indicate which cases are valuable enough to make it worth investigators’ time to go after. But that is not to say the value of in-store intelligence has been lost, noted Cunningham. “We haven’t turned our back on what associates see and tell us; in fact, we’re doing a better job than ever crowdsourcing and getting tips, and then using that to inform our data analysis,” he said.
Often, data analysts will access store video or other sources of data to help prove up the case, or perhaps to identify if the anomaly is instead the result of a training issue or a systems snafu, such as a glitch in the point of sale. Whatever the cause, analytics provides the AP team the ability to rapidly identify and remediate causes of loss and inefficiency.
“Identifying training issues can be a huge area for us, such as a process that has not been communicated correctly to associates, which they can see through the data,” said Safeway’s Nathan Bandaries. In those cases, the ability to act quickly and correct the training gap prevents loss that typically exceeds even cases of widespread fraud.
Although computation extracts the narrative, there is still strong reliance on people looking in the right places. The ability of 1s and 0s to do their job hinges on experience, intuition, creativity, coordination, and relationships, suggested LP leaders.
For example, to improve identification of fraud schemes, inefficiencies, or problems that need solving, there must be continuous refinement of data presets and monitors, which is typically reliant on human skill and coordination. “It’s based primarily off the knowledge of the analyst and their partnerships with their fellow analysts, as well as the understanding of corporate teams to understand trends in other divisions that could carry over into those divisions that are not currently experiencing the issue,” Bandaries said. “Also, analysts will often come up with their own analytics and their own ideas based on their experience, trends, and understanding of history.”
Analytic tools are improving, which can remove some of the technical work from investigative equations, but savvy and experience are still required to extract maximum value from data input. Gonzalez, who serves in both the role of data analyst and investigator, described some of the skills and strategies that serve teams well. “You need people who are intrinsically skilled to see outliers, who are disciplined, understand statistical theory, and can put it in context alongside all the other data,” he said. “And you need to be savvy enough to understand when you have a false lead, and this is where a good feel for the industry and understanding what’s going on in the building helps.”
Looking at data from all sides is also important, as Gonzalez highlighted in a recent LP success story. The data seemed to be all positive, with a sales associate ringing up impressive sales numbers on mobile transaction devices—until it was discovered that a significant number of them were fraudulent and made using stolen credit cards. “It’s not something we would have noticed except that we examined nontraditional points of data and were willing to consider what is driving the data even as it’s being celebrated,” said Gonzalez. “You have to look for fraud on both ends of the spectrum and look for outliers on both the negative and positive side. When you have too much of anything you need to question it.”
The interplay between investigators and data analysts often continues all the way to the interview room door, said Bandaries. “There is a bunch of back and forth throughout the process. I know of times when investigators would step out of room as they are ready to go into an investigation and will make one last call to the analysts, to make sure they are on point with their investigation and that they have a thorough understanding of the loss avenue—and analysts support them in that.” Data analysts also fill the role of educator for new investigators, helping them to understand how actions and illicit behavior are observable in the data.
As it suggests, the partnership between AP investigative teams and data analysts is more important than ever to breaking up theft rings and reducing shrink. “The partnership with the analysts is huge. It’s the most important part of our having been able to conduct large successful internal cases, breaking up a prolific external refund fraud operation, and even in smaller cases,” said Bandaries. “Often, analysts put together everything for us so that literally all we need to do is conduct the interview.”
Data analysts typically generate cases—identifying data to monitor, focusing on unusual findings, developing information, adding corroborating evidence such as photos or video if available in a case file, and then putting that in an email to AP investigators. “Very rarely do we do some digging and find out that there is not something there or that there is not a case. Most of the time it is dead on,” said Bandaries, while also suggesting that a successful long-term partnership can’t be a one-way street. “One thing we’re always sure to do is to include data analysts in the conclusion of the case, and to say, ‘this was the yield of that case’ and ‘here is what that person admitted to.’ Completing the circle helps give them some ownership. Letting them know how cases turned out is extremely important to us moving forward with future cases.”
Designer Brands shared their success story at NRF Protect. To combat a trend in rising shrink at stores, its AP team began to focus on improved data analysis to make effective decisions, according to Adam Gilvin, the footwear retailer’s director of asset protection. “Seven or eight years ago you started to hear about big data. Well, we have a lot of data—we’re satisfied with our data—but what do you do with it? Where do you push it?” For answers, AP is centralizing its analytics function and hiring more analysts to build out an elite team that is highly skilled in “data-driven decision making,” said Gilvin. “We were too often making decisions on anecdotal evidence or based solely on experience.”
Pethuyne is one of those recent experts hired to kickstart a more analytical AP approach. Already he is finding success in identifying fraud patterns in online sales, such as whether customer service representatives are sending packages to friends and family. “That has been successful for us, a good case generator,” he said. “We also did the same thing with store associates that might be doing it, and we were getting hits right away. It was a good case generator from minimal work.”
Pethuyne noted the importance of working off a robust data platform (in their case, Profitect) that permits easily transitioning between data in different parts of the business. “You want to be able to easily jump from the sales module to data on the back of the house,” he said. “That ability to pivot easily [between data] provides additional value.”
A data-driven approach is cutting costs as well as reducing fraud at Designer Brands. For example, data noted a significant overspend by stores on expensive EAS tags, and by using analysis to rank shoe brands by theft risk and targeting price points they wanted to protect, they redesigned their EAS strategy, reduced shrink, and secured significant savings for the company by reducing its spending on tags. “That is where having data really shines,” said Pethuyne.
Still, the industry could improve its use of analytics, said industry leaders. For example, several leading LP executives suggested that the industry can do a better job conducting post-case analysis to show the return on investment of ORC investigations. By putting analytics to work on sales data and inventory turns, the financials of ORC rings, shipping data from fence operations, and other sources, LP teams can put a reliable price tag on investigative cases and cite specific harm in victim impact statements. Taking the time to apply analytics in this way can dispel the commonly held idea that there is no good way to measure the value of an ORC case.
Finally, it’s not only human skill that shapes the value of investigative analytics, it’s also human limitations. A burgeoning area of analytics is built around trying to put information extracted from data analysis into formats that the human mind can more easily make sense of. Home Depot’s Nate Cunningham, for example, said his AP data team has focused on tools to provide better data visualization for its field partners, to find creative ways to help them make use of data without overwhelming them.
Visual simplification is also a driving force behind the growing use of maps in connection with analytics. “If anyone is not using GIS (geographic information systems) with their LP analytics and business analytics, they’re missing out,” said Scott Peacock, director of analytics and insight for Walmart Global Investigations in a 2018 interview with LP Magazine. “Adding a geospatial component opens up lots of different avenues that otherwise might have been missed or misunderstood.”
LP collects myriad data describing the time of events and locations. Geospatial analysis uses this data to build maps and other visuals to depict where and how changes are taking place, to anticipate and prepare for future changes, and to make complex relationships understandable. Such geographic profiling can spatially focus an investigation and yield insights that are impossible to see in the data alone. “There absolutely is unique value in seeing things expressed visually. There are often patterns of things that cannot be understood unless you view them spatially,” explained Lorie Velarde, GIS analyst with the Irvine Police Department in California.
Using transactional databases to search inventory and product sales, geospatial analytics allows analysts to visually identify emerging patterns and gain advanced security intelligence. Maps show where ORC activity is occurring and where it’s likely to head next. “We’re able to leverage predictive analytics to identify where and when bad actors are going to hit, and we can then stage targeted surveillance, do some target hardening, or prepare for an apprehension,” explained Peacock.
A Smarter Future
Analytics already are at the center of generating cases for investigation, but there is still “a lot of growth potential,” said Bandaries. Safeway, for example, has plans to integrate more video analytics into its investigative processes.
Bloomingdale’s is also taking advantage of the “monumental changes” in video analytics that Gonzalez sees. “Who is lingering in what area, putting two things in the bag and ringing up just one—the technology of the cameras and the ability to translate that video information into numbers for analysis is allowing for a much better understanding of what’s happening in our environment,” he explained.
Retailers are also pulling in data from facial recognition or other identification solutions. Using such tools can permit ORC investigators to piece together crime events, each one of which could be its own police report, in a more comprehensive way—to fundamentally disrupt an ORC operation rather than make a single arrest. Managing the privacy component of this and other identification technologies will be a test for retailers and could be a near-term drag on adoption, said some LP leaders.
The quality of analytic dashboards grows more important, as even more data heads LP’s way. Continuous analytics of all sales reducing activities (SRAs) and related data is inevitable in retail, as it provides a reliable, economical way to fix systemic problems, reduce shrink, and ultimately improve profits. Eventually, all data correlates with future sales figures.
Machine learning and artificial intelligence (AI) is also transforming the analytics landscape. Leading global retailers are already leveraging it for a variety of purposes. For example, a mix of accelerated analytics and deep learning is helping retailers with pricing strategies, and retailers are running daily profit-optimization calculations to know how best to distribute which products to which stores. Online, AI works on customer data to help retailers provide more personalized shopping experiences and to help e-commerce platforms adapt to the needs and interest of online shoppers. AI can also fuel predictive price and forecast simulations to boost revenue by fractions of a percent, which for giant retailers can add up to millions.
For loss prevention, these advances lay the groundwork for increasingly smart pattern detection in retail transactions. Models become more capable over time at identifying previously undetected fraud patterns and better at distinguishing between problematic and legitimate transactions. Previously hard-to-identify collusion cases—those spread across large numbers of customers, for example—are drawn into sharp relief. Investigators become more efficient. Predictive analytics are more easily added to those focused reactively on shrink. Data analysis transforms into a 24/7 function.
In this emerging world, a large retailer can compile a list of all its top risks—regulatory compliance, brand protection, and so on—and then turn machines loose to identify when trends start to emerge or behavior changes. A slight anomaly in the cashing of IRS refund checks, for example, could mean a retailer is processing more counterfeits.
How will LP fit in this new world order? How is the role of a LP professional likely to change? Retailers at the forefront of this trend shared some of the changes that they expect to result:
- Greater centralization of analytics
- A shift toward internal process improvements and internal theft and away from external theft
- A lowering of criticality, such that increasingly smaller loss events can become targets of prevention
- More responsive LP countermeasures as retailers identify instances of fraud in real time
- A reduction in overall LP headcount
- Greater value assigned to LP practitioners with specialized skills who can wear multiple hats
- Greater focus on using data to target issues that can be corrected or problems that can be solved with existing personnel resources.
One misconception about the future is just how far off it is. Some leading industry experts said that the open source nature of AI problem-solving makes it less labor intensive, less expensive, and easier for retailers to pull off than they probably think. Given the results it can yield, cost for a major retailer isn’t even an issue, some say.
And though perhaps fewer in number, people—and the relationships between them—will still play a critical role in an AI future, say industry leaders. Data scientists and criminal behavioral analysts will work side-by-side, with analysts identifying tools they need and software teams taking company data to start building those tools within hours.
Loss prevention professionals will have a role in a machine-learning world, said Gonzalez. “People, more than machines, can look at items for the X factor that might escape data’s attention, to see more than what is in front of them, to make inferences from their understanding of the battle,” he said. “You can get very good info from machine learning, truly actionable data, but you need a person to interpret, and to contextualize, and to see a data point that a machine might have missed.”
Far from subtracting from the value of LP professionals, increased reliance on analytics and more powerful analytic tools may ultimately enhance it, suggested LP leaders. Data offers LP a clear path to becoming the full business partners we strive to be, they said.