Improving Loss Prevention by Applying Swarm Intelligence
In retail, theft and intentional damage to merchandise is a common problem. Retailers have several loss-prevention solutions in place, but theft still occurs as shoplifters find new ways to steal a product or even part of a product. Monitoring systems at retail stores use traditional video surveillance where the person behind the camera determines unusual behavior that needs investigation. But nowadays, retailers need to step up loss-prevention techniques in big ways, as merchandise loss can happen not only at the store, but also in the warehouse, distribution center, depots, while in transit, etc.
Particularly within stores and common shopping areas, shoplifters today are using newer techniques to be as invisible as possible: while in the act, they create distractions by working as a flash mob. Retailers are challenged to monitor through surveillance every suspicious activity within a large crowd, especially during the high season when there is a significant uptick in store traffic. When shoplifters work in flash mobs, the loss due to theft is greater.
Using Science and Technology to Identify Shoplifters
Retailers and wholesalers, who have much to lose, employ additional help during high season to get more boots on the ground. But there’s a fine line that retailers need to walk: you cannot make prospective customers uncomfortable by scrutinizing them, as they might not return. Using today’s technology, the art of identifying shoplifters can be converted into a science.
Analyzing the behavior of shoplifting mobs offers insight into the merchandise or the department/class of merchandise that is most at risk. Each store places its merchandise based on a planogram, which in turn, is based on the analysis of flow patterns of prospective customers. Security and monitoring systems are placed where high value merchandise is displayed and shelved. Applying software simulations to shoplifting behavior can help retailers identify and prepare for a number of shoplifting scenarios.
Applying Swarming Intelligence
By using Swarm Intelligence (SI) algorithms such as Particle Swarm Optimization (PSO), the Bees Algorithm, the Ant Colony Optimization, the Bat Algorithm, etc., retailers can create simulations to understand potential threats based on how merchandise is placed at the store. PSO is a good algorithm to apply to the flash mob scenario, as it helps retailers understand the behavior of each individual (particle)—and ideally, of the leader coordinating the flash mob—who is browsing at a store. The store format may have multiple levels (n-dimensions) with various departments for different categories and merchandise, and the flash mobs organize themselves based on the information they gather moving through the store.
For example, the information swarm members might communicate includes: “there is too much security around here", "lots of people in this area", "no security here", or "many store associates are on a break." This information is enough for the entire group to swarm in if they see criminal opportunity or swarm out if the risk of being caught seems likely. Simulations of these models can help security and store associates determine the risk of theft by department and type of merchandise, so they can prepare accordingly.
Knowing When a Swarm Event Is Happening?
Simulations aside, we can identify an actual swarm event and known flash mob individuals via their smart devices. Smart devices offer multiple ways to zero in on individuals within a specific area. A number of TV shows, such as CSI, NCIS, ID TV, have introduced us to the idea of using radio signals bouncing and triangulating off cell towers to hone in on a person’s location.
Bluetooth and Wi-Fi beacons within smart devices can be used to conduct similar triangulation methods in stores. In retail locations, the quality of a General Packet Radio Services (GPRS) signal is spotty; moreover, only service operators and law enforcement have access to it. However, Bluetooth and Wi-Fi beacons can be intercepted by any device listening to them. These beacons have a unique ID that identifies the device, the device class, the device manufacturer and the signal strength. By capturing these beacons, similar to a heartbeat, and comparing the suspect device Machine (MAC) IDs in a retailer’s database against its store setup plan, retailers can determine potential alignment of actual movements to simulated swarm behavior. Every time the suspect ID enters the store, a quick correlation with other suspect IDs can be used to predict an impending event. The store security and associates can then be alerted to take appropriate action.
Understanding Shoplifting Patterns
Analyzing beacon data over time is crucial for building a trending data-set with the correlations needed to perform swarm analysis. Using position analysis and the path traversed by the suspect (MAC) IDs, the amount of time a suspect individual remains in a certain spot, and the person’s proximity to specific merchandise, retailers can establish a visual map projected against a store's planogram to reveal shoplifting patterns based on the merchandise setup in the store aisles. They can generate heat maps that visually inform store associates and security if multiple individuals are acting suspiciously, learning the patterns of their behavior related to scouting expeditions or an actual shoplifting event. Associating these projections with store operation data can provide further insights on the time the incidents occurred, the number of associates on the floor, and the peak flow pattern of customer Point of Sale (POS) transaction periods, etc.
Of course, privacy issues concerning collection of this data, especially the capture of information about the device and the traversal data of individuals, should be considered. Similar to procedures around surveillance videos, retailers will need to put the right security measures in place to ensure data is consumed and used for security purposes only.
Other Use Cases for Swarm Intelligence
PSO can be applied to other areas, such as in stadiums during a sporting event or at a concert, in a theater for a play or a movie, or in a theme park. Most consumers/patrons have smart devices that can be used to detect presence and provide location-specific data. There are a couple of hardware products on the market that capture the phone’s digital fingerprint (MAC ID) data from smart devices. They have a good range and can be tuned to capture data at rates that fit a variety of use cases; the higher the fidelity of the data, the greater the opportunity to serve consumers and patrons better while keeping track of unsavory individuals.
To put these programs in action, a knowledgeable partner can help you build the algorithms for your specific use cases and the insights you want to derive from this data. Do not fall into the trap of "once and done," as you will have to fine tune these algorithms over a period of time. Remember flash mobs change their tactics, and so should you. Running simulations and applying them to various store formats can help you understand the gaps in store security and your vulnerabilities.
-Kuruvilla Mathew (Mat), Chief Innovation Officer at Ness
Particularly within stores and common shopping areas, shoplifters today are using newer techniques to be as invisible as possible: while in the act, they create distractions by working as a flash mob. Retailers are challenged to monitor through surveillance every suspicious activity within a large crowd, especially during the high season when there is a significant uptick in store traffic. When shoplifters work in flash mobs, the loss due to theft is greater.
Using Science and Technology to Identify Shoplifters
Retailers and wholesalers, who have much to lose, employ additional help during high season to get more boots on the ground. But there’s a fine line that retailers need to walk: you cannot make prospective customers uncomfortable by scrutinizing them, as they might not return. Using today’s technology, the art of identifying shoplifters can be converted into a science.
Analyzing the behavior of shoplifting mobs offers insight into the merchandise or the department/class of merchandise that is most at risk. Each store places its merchandise based on a planogram, which in turn, is based on the analysis of flow patterns of prospective customers. Security and monitoring systems are placed where high value merchandise is displayed and shelved. Applying software simulations to shoplifting behavior can help retailers identify and prepare for a number of shoplifting scenarios.
Applying Swarming Intelligence
By using Swarm Intelligence (SI) algorithms such as Particle Swarm Optimization (PSO), the Bees Algorithm, the Ant Colony Optimization, the Bat Algorithm, etc., retailers can create simulations to understand potential threats based on how merchandise is placed at the store. PSO is a good algorithm to apply to the flash mob scenario, as it helps retailers understand the behavior of each individual (particle)—and ideally, of the leader coordinating the flash mob—who is browsing at a store. The store format may have multiple levels (n-dimensions) with various departments for different categories and merchandise, and the flash mobs organize themselves based on the information they gather moving through the store.
For example, the information swarm members might communicate includes: “there is too much security around here", "lots of people in this area", "no security here", or "many store associates are on a break." This information is enough for the entire group to swarm in if they see criminal opportunity or swarm out if the risk of being caught seems likely. Simulations of these models can help security and store associates determine the risk of theft by department and type of merchandise, so they can prepare accordingly.
Knowing When a Swarm Event Is Happening?
Simulations aside, we can identify an actual swarm event and known flash mob individuals via their smart devices. Smart devices offer multiple ways to zero in on individuals within a specific area. A number of TV shows, such as CSI, NCIS, ID TV, have introduced us to the idea of using radio signals bouncing and triangulating off cell towers to hone in on a person’s location.
Bluetooth and Wi-Fi beacons within smart devices can be used to conduct similar triangulation methods in stores. In retail locations, the quality of a General Packet Radio Services (GPRS) signal is spotty; moreover, only service operators and law enforcement have access to it. However, Bluetooth and Wi-Fi beacons can be intercepted by any device listening to them. These beacons have a unique ID that identifies the device, the device class, the device manufacturer and the signal strength. By capturing these beacons, similar to a heartbeat, and comparing the suspect device Machine (MAC) IDs in a retailer’s database against its store setup plan, retailers can determine potential alignment of actual movements to simulated swarm behavior. Every time the suspect ID enters the store, a quick correlation with other suspect IDs can be used to predict an impending event. The store security and associates can then be alerted to take appropriate action.
Understanding Shoplifting Patterns
Analyzing beacon data over time is crucial for building a trending data-set with the correlations needed to perform swarm analysis. Using position analysis and the path traversed by the suspect (MAC) IDs, the amount of time a suspect individual remains in a certain spot, and the person’s proximity to specific merchandise, retailers can establish a visual map projected against a store's planogram to reveal shoplifting patterns based on the merchandise setup in the store aisles. They can generate heat maps that visually inform store associates and security if multiple individuals are acting suspiciously, learning the patterns of their behavior related to scouting expeditions or an actual shoplifting event. Associating these projections with store operation data can provide further insights on the time the incidents occurred, the number of associates on the floor, and the peak flow pattern of customer Point of Sale (POS) transaction periods, etc.
Of course, privacy issues concerning collection of this data, especially the capture of information about the device and the traversal data of individuals, should be considered. Similar to procedures around surveillance videos, retailers will need to put the right security measures in place to ensure data is consumed and used for security purposes only.
Other Use Cases for Swarm Intelligence
PSO can be applied to other areas, such as in stadiums during a sporting event or at a concert, in a theater for a play or a movie, or in a theme park. Most consumers/patrons have smart devices that can be used to detect presence and provide location-specific data. There are a couple of hardware products on the market that capture the phone’s digital fingerprint (MAC ID) data from smart devices. They have a good range and can be tuned to capture data at rates that fit a variety of use cases; the higher the fidelity of the data, the greater the opportunity to serve consumers and patrons better while keeping track of unsavory individuals.
To put these programs in action, a knowledgeable partner can help you build the algorithms for your specific use cases and the insights you want to derive from this data. Do not fall into the trap of "once and done," as you will have to fine tune these algorithms over a period of time. Remember flash mobs change their tactics, and so should you. Running simulations and applying them to various store formats can help you understand the gaps in store security and your vulnerabilities.
-Kuruvilla Mathew (Mat), Chief Innovation Officer at Ness