The retail landscape is changing like never before. Brick and mortar retailers face stiff, and what may seem like unfair, competition from online retail options. New-age shoppers are tech-savvy and store-smart, demanding more from each shopping spree. The prevalence of technology continues to grow and influence more aspects of shopping, offering personalized experiences to win over the loyalty of intelligent customers. Retail tech not only helps understand the customer better but also decipher the store’s shortcomings. It proposes solutions, reduces shrinking, and ensures greater ROI. This analysis is made possible with real-time intelligent video analytics (IVA) and Edge computing.
A closer look at the different ways in which retailers use IVA-powered AI gives insight into the functioning of intelligent retailing.
Fraud and Theft Detection
Fraud and theft are two challenges of retail business that can be curtailed only through constant vigil and continual law enforcement. According to the 29th annual National Retail Security Survey released by the National Retail Federation, theft and fraud losses in the United States totaled USD 61.7b in 2019, much higher than the preceding years. The corresponding global numbers stand at USD 100b. It is estimated that fraud and theft account for almost two-thirds of retail shrink every year. With such startling figures, retailers are always on guard, using a mix of conventional methods and advanced technology.
The cameras connected to the IVA systems at the store entry and exit points help capture people’s count and use biometrics and face recognition software to identify customers visiting the store. When a customer is involved in fraud or theft, the facial data is evidenced for future use. Consider Japanese start-up Vaak’s AI theft-detection system that is said to be well over 80% accurate in identifying potential and actual fraud and theft by processing massive amounts of data analyzed through deep learning algorithms run on the IVA. The system can immediately identify and sound an alarm when a blocked customer tries to enter the store to take corrective actions.
However, fraud and theft detection is not just limited to the point of entry. There are possibilities of crime throughout the store, including at the time of payment and checkout. For instance, Malong Tech’s RetailAI Protect consists of an overhead fixed-dome camera that captures the footage of unscanned and suspect-barcode items and sends the information to the backend AI model to decode. If the system detects foul play in terms of either mis-scans or barcode ticket switching, an alert is raised at once.
IVA makes for easy review of footage from multiple cameras in a matter of minutes, rather than hours or days of manual review, to investigate losses and crimes in stores. Operators can use filters to search for people or objects that match a specified description, extracting crucial details, gather evidence, and accelerate investigations.
A lot can happen along the aisles of a retail store. From finding the right item that a customer has been frantically looking for to complete frustration at a poor shopping experience from stock-outs, effective aisle management can make or break sales. Apparel giant H&M uses AI to keep popular items stocked up by analyzing purchases and store receipts. With insights into the aisles that are popular, the customer dwell times in each of them, and demographics of the customers captured through IVA, retailers can increase on-shelf stocks of fast-moving products, improve merchandising, and offer instantaneous promotions to increase revenue and provide an enticing shopping experience.
Analytics technology used on in-store video at the aisles also helps store staff record and understand what kind of purchases the customers make and their average spend and read emotions and identify customer dissatisfaction. All these inputs help staff attend to those shoppers who might need more attention. For example, select outlets of clothing store Uniqlo have AI-powered UMood kiosks that determine customers’ moods by studying their reactions to different colors and styles, enabling better aisle management for Uniqlo.
In a customer-centric business world, retailers go all out to engage customers. Customer Intelligence (CI) gives retailers a competitive edge by consolidating and analyzing all available customer data to improve communication, study and influence buying behavior, and drive better sales through predictive recommendations. In a 2018 Harvard Business Review Analytic Services study in partnership with several big IT players, 83% of respondents said that the ability to translate data into actionable insights at the right time is essential to customer experience. Still, only 22% were successful at it.
Observing how many customers come in and when is a major analytic challenge (?). When analytics are changed to insights, they can improve the relevancy of the customer experience, ensuring that the retailer resonates with the customer. These analytics provide operational and branding insights, along with several other aspects of CI. CI provides the context within which purchase decisions are made and can be used to increase ROI.
Several retailers in the past decade have also been experimenting with the augmented reality concept of Magic Mirrors. These cameras in these mirrors feature a time-delay display that allows customers to turn around and see a 360-degree view of themselves while trying on clothing. The digital wall at Rebecca Minkoff’s flagship store in Manhattan, New York, cannot be missed. Not only does the interactive mirror help try out clothes virtually by using IVA, but it also offers customers suggestions on accessories to go with the total look. They can also order a drink and request staff assistance when needed. The connected mirrors in the fitting rooms also allow you to browse through the available collection and order the right size.
In-Store Staff to Customer Ratio
The ability to accurately count and analyze customer traffic allows retailers to ensure efficient operations. IVA from footage captured at store entry along with other sensors that capture analytics provide real-time traffic data. IVA helps differentiate the staff from the customers based on dress code detection and staff facial and biometrics recognition. By regularly analyzing this data, retailers can ascertain traffic patterns that improve staffing levels based on customer demand. Deducing the staff to customer ratio helps retailers decide to reallocate staff within different sections of the store or bring in additional staff during days and times of high footfall and cut down overstaffing costs during slow hours.
Busy checkouts and long queues can be avoided using IVA. AI can automatically analyze the networked video footage and raise an alert when high traffic is detected at checkout counters. Based on instant analysis from connected cameras and sensors at aisles and shelves, AI can help predict the possibility of long queues, over-occupancy, and customer surge at checkouts. This can help retailers prepare ahead to open more billing counters. IVA can also auto-trigger alerts to staff to further accelerate the checkout process. Specific queue management algorithms can just as well calculate the specific wait time for each customer in the queue as it can delay servicing a customer. Such insight helps identify issues, optimize the number of queues, and reallocate customers to different queues for faster billing.
Stores like Futuremart in China and Amazon Go in the US go a step further with IVA tech with completely cashier-free and cash-free autonomous stores. Facial recognition for entry, QR codes for purchases and payment apps for billing, along with sensors and cameras all over the stores, make for a queue-less experience. Futuremart even features a Happy Go meter that offers customers bigger discounts based on their smiles.
Counterfeit products are so good these days; it is almost impossible to tell them apart from the real thing. Forgers have become adept at using AI themselves, making it easier to engineer their products to pass off as genuine. The World Customs Organization (WCO) estimates that 7-9 percent of global trading relates to fake products, making it a profitable business by itself. But selling counterfeit products can lead to loss of revenue, reputation, future sale. This makes it critical that retailers focus on implementing AI to detect fakes.
Image Recognition and Object Detection techniques can help retailers regulate store checks and get consistent results. Using deep learning neural networks, it is possible to compare products on shelves and establish authenticity and genuineness. The neural network can be trained using images to recognize those products that differ from the original in any single or inconspicuous manner. A miss during painstaking manual checks can be instantly caught through IVA running on these deep learning algorithms.
Companies like Entrupy and Authentic Vision have been working towards leveraging IVA, advanced data science, and optical machine learning to identify counterfeit goods in real-time and deliver a high-quality user experience. However, a known limitation is that there are no fool-proof counterfeit detectors. Even the most accurate ones may not pass all the tests.
What IVA Signals for Retail
IVA aggregates video data over time to provides businesses with the intelligence to understand trends, make smart decisions, and develop powerful strategies. It goes a long way in providing valuable inputs towards a delightful customer experience and minimizing shrinkage. It helps bridge the gap between online and physical presence. With more businesses taking advantage of AI-led IVA, the retail terrain is changing. IVA combines visual acuity with analytical strength, cataloging information to provide retailers with coils of rich, instantaneous insight, closing the chasm of human error.
This article was written by the following authors:
Vinod Bijlani is an AI & IoT Expert with over 2 decades of experience.
Utpal Mangla is a Global Leader in the Telco, Media, Entertainment industry.
Mathews Thomas is Distinguished Engineer and has over 25 years working with many of the major Telecom and Media companies.