Using Edge Computing to Enhance Customer Experience and Operational Efficiency - Part 1
Thu, 10 Aug 2023 12:41:15 -0000
|Read Time: 0 minutes
Edge computing is becoming increasingly valuable for retailers who are trying to stay competitive. Through edge computing, retailers can do real-time analysis of data and management of different machines. Edge computing can also provide solutions to problems previously thought unsolvable, or it could help to improve the efficiency of the business.
The goal of this document is to provide insights regarding current problems in the retail space, along with showcasing possible edge computing solutions to solve those problems. This research document aims to provide a comprehensive analysis of potential edge computing solutions for those current problems in the retail sector, including retail shrink, stocking, and item locations.
I have found three separate problems on my own in the retail sector that may possibly be solved using computer vision on edge computing platforms. For each problem, I will introduce the problem including any relevant background information, a real-world scenario of the problem in action, why it would be beneficial for the company to solve the problem, a description of a possible solution using Dell Technologies’ hardware, and two diagrams explaining the solution in more depth. I will then explain how these solutions can be compounded into a single solution to solve the problems more efficiently, and why edge computing is the right tool for the job.
In addition to the three problems I have found, I interviewed multiple shoppers and asked what kinds of problems they encountered in retail spaces. While the problems are not directly solved by edge computing, the impact we make on store sales and efficiency may also help to lessen the impact of these problems.
Problem 1: Retail shrink
Overview
Retail shrink is a significant problem for many businesses. There are several types of shrinkage commonly found in stores. The three main types of shrinkage in retail are shoplifting, employee theft, and return fraud. Shoplifting is when external individuals steal inventory from a retail store. Employee theft occurs when individuals associated with a business steal from or defraud the company. Return fraud occurs when someone steals a product and then returns it for a refund. Other types of shrinkage include administrative error, which can include mislabeling merchandise for a lower price or refunding merchandise for more than it is worth, and vendor fraud, which occurs when vendors overcharge or add fees to invoices to steal physical inventory from the store when they are onsite.
Shoplifting is a serious issue that causes stores to lose significant amounts of money. According to the Annual Retail Theft Survey, 21 large retail companies with 18,994 stores reported over $136 million in recovered dollars from shoplifters in 2019. This amount does not even include the money lost from shoplifters that they did not apprehend.
Directly stealing is not the only way to shoplift. Some individuals take items where the weight needs to be measured and hold on them gently so that the weight of the item is less than it should be. Others take an item that does not get weighed and weigh it like something it is not, such as taking a valuable notebook and placing a sticker on it saying that it is ten pounds of apples.
Real-world scenario
Suppose I am at a store and come across a fresh bunch of bananas. I proceed to scan the bananas at the self-checkout machine. The machine prompts me to weigh the bananas, and I hold them slightly so that they weigh a quarter of a pound instead of their actual weight of five pounds. As a result, I pay twelve cents instead of 2.5 dollars for the bananas.
Additionally, let us say I purchase a bag of apples from the same store. The store offers two options: organic and regular apples. Although the organic apples are more expensive, I prefer them and decide to purchase them by scanning them as the cheaper regular apples. This action causes the company to lose the extra money they would have made from selling organic apples because of my dishonesty.
Benefits of addressing the problem
Reducing retail shrink can help businesses save money and increase profits. When you reduce retail shrink, you will have more money that you can then reinvest back into the business. Reducing shrinkage can also help businesses improve customer satisfaction.
Preventing shoplifting can benefit stores in many ways. For instance, stores can earn more money since people will not be stealing their items. With fewer goods stolen, stores will have more goods to sell, which means they will have more potential revenue to make. Stores can also save money because instead of having to spend money replacing stock, they can sell the stock they already have. In addition, preventing shoplifting can help stores maintain a positive reputation and customer experience.
Possible solution
This solution I devised requires use of a retail store’s pre-existing or newly installed camera system. You will also need a Dell VxRail cluster for compute power and storage along with computer vision applications for analyzing the incoming video feeds. You will also need to have connectivity with the self-checkout machines.
In this solution, the store will have cameras set up throughout the aisles of the store, and they will have cameras at the self-checkout as well. The cameras will feed live camera data to the VxRail cluster. The cluster will contain pre-trained machine learning models capable of identifying and tracking different products throughout the store. At the self-checkout, the customer will scan their items and place them in the bags. The camera installed at the self-checkout will verify that the barcode the customer scanned is actually for the item (example: making sure the customer did not put a cereal box barcode on a laptop to get it cheaper). If the item is determined to be incorrectly labeled, it will ask the customer to scan it again, and if the item is still incorrect, it can alert a cashier to assess the situation.
Since the models will be able to track the products throughout the store, it can also catch people hiding products in their clothes or leaving the products in the basket at the self-checkout and not scanning them. This way, we can prevent shoplifting from both stolen items and purposefully mislabeled items.
This solution also gives us valuable data that the store can use to better prevent shoplifting and improve the customer experience. The store will know what items are more likely to be shoplifted, and they will be able to take precautions to prevent future shoplifting, such as keeping them in a locked glass door or keeping the items in the back.
Figure 1. Technological diagram
Figure 2. Shelf-checkout sequence diagram
In this diagram, the security cameras send the real-time camera footage over to the cluster. The Cluster then gives the footage to the nodes containing the ML models, which process the footage and give the resulting product info back to cluster. The cluster receives the scanned product from the self-checkout and sends both the scanned product and the product info from the ML model over to the Item Verifier App which verifies that the camera footage matches what the self-checkout scanned. The app sends the validity of the item back to the cluster, which can pass that info back to the self-checkout to continue the customer’s purchase. Th cluster can log any of the attempted shoplifts into a database that the retail employees will be able to view from a dashboard.
Read more
For more information about using edge computing to enhance the customer experience and operational efficiency, see Part 2 and Part 3 of this blog.