Several years ago, I waited 45 minutes in the queue just to buy a simple beverage and I thought whether we really need cashiers in retailers, or not. I knew the answer was no. In my opinion, it was possible to walk in, grab whatever you like and leave with a self-checkout system. However, because of the lack of technical knowledge, I could not come up with a brilliant idea. The only idea came to my mind was RFID (Radio-frequency identification) tags. The idea was simple; tag every product with RFID tags (like we see on books or DVDs) that will enable customers to scan each product with their smartphone –Near Field Communication (NFC) technology is standard for most of the smartphones today- and add the products to their cart and pay online.
Although the idea seems good, it was not functional. Changing the barcode system is not that easy. Additionally, even if it reduces the labor cost, tagging each product with RFIDs decreases the gross profit margin. As you can imagine, I failed. Still, the pitch of the idea provided me an A from Entrepreneurship course.
The only person around the world thinking about autonomous checkout systems could not be only me. I discovered that all the big players like Amazon and Walmart were working on autonomous checkout systems. So, I have done some research.
Different approaches to self-checkout
There are different traditional checkouts. While some of them are commonly, some are rare and they have pros and cons.
Some convenience stores have already been using self-checkout machines in their stores for a few years. They have some advantages:
- Shorter lines
- Shorter waiting times
- Faster checkout
- Greater accuracy
- Reduced labor cost
However, they are not the ultimate solution for fully automated checkout (I was in the queue for one of these machines). Moreover, implementation costs are very high. A typical four-lane setup costs $125,000 to install – they certainly are not cheap! For more information, you can read The Conversation’s article about self-service checkouts.
Jed Griffin and Scott Deuty stated that shopping-cart level checkout is possible with RFID technology in 2014 (Yes, it appeared that I was not alone). They said in their blog post that both the technology and methodology now exist for making cost-effective tags that have enough range to be read as the cart passes through a reader bay.
A company called Automated Stores produces customized vending machines to provide automated checkout experience. Even it automates the checkout experience and it is more secure than other approaches, vending machines put a glass between products and customers and eliminate the touch.
Walmart Scan & Go App
Walmart’s solution for autonomous checkout system is really simple (It resembles my idea actually; I thought RFID tags are better in terms of security). Walmart introduced the Scan & Go program in select stores between 2012 and 2014. The app basically allows customers to scan the products with their smartphones and pay by scanning a barcode for their total purchase and pay for their goods.
The first attempt at the program was a failure. Customers were complaining about scanning problems and there was no troubleshooting. The company ended the program in 2014 due to overwhelmingly negative customer experience.
The company has the program still running in a dozen stores in Texas, Florida, South Dakota, Arkansas, Georgia, and Kentucky. For the customers who do not have a smartphone, Walmart provides hand-held scanners and for the customers who are not willing to link their credit cards or payment method, there are options.
However, the concept is not a fully automated checkout system since you do have to have a Walmart employee verify the receipt on your smartphone and clear you before you can leave.
AI and image recognition will boost self-checkout
In general, the main reason for long waiting times in long checkout lines is Point of Sale (POS) systems and POS devices. Traditional payment process is what causes increase in waiting times whether you are in a convenience store or supermarket or H&M. There are software solutions for payments. One of the key features of automated checkout systems is online payment with pre-registered credit card. At the end of the day, you wait to pay what you get from shelves.
Than, another problem shows up. How to determine
If you recently searched for autonomous checkout systems, you probably encountered with the video of 2 people randomly picking items from shelves to demonstrate Standart Cognition‘s technology.
The company uses computer visions to track people and the products in real-time in the store. Deep learning and image recognition technics enable Standard Cognition to recognize items. Brandon Ogle, co-founder and an engineer said that the system is currently correct 98 percent of the time. You can find out more about the technology in the article written by Rachel Metz on MIT Technology review.
Online retailer Amazon opened its beta version of the cashierless store on 5th of December in 2016 in Seattle. The 1800-square-foot retail store was only for employees.
Amazon also uses computer vision and deep learning algorithms like Standard Cognition. Amazon called the technology just walk out technology similar to technology used in autonomous driving vehicles. After shopping, customers can simply leave the store. Amazon automatically charges the customers’ Amazon account and receipt are sent to the app.
The store has yet to be opened in new locations as Amazon is still improving the system. Since last December, Amazon employees have been trying to fool the system with though challenges like wearing Pikachu costumes while shopping. Even the system passed the costume test, it has hard times to detect customers who shop in groups, such as families. Still, Amazon is close to expanding this beyond a pilot. According to Bloomberg, Amazon is hiring construction managers who would build the first stores.
Funnily, the Go stores which are meant to reduce queues are causing queues as people line up to see how they work:
Long before Amazon and Standard Cognition, Imagr published a video showing its SmartCart powered by image recognition technology. You link your phone to the cart with an app and that’s it. Scanning is not necessary, artificial intelligence recognizes the products in the cart with image recognition technology. Then you just pay online via the app and leave the store.
Wheelys is a company like Uber for retails, not transportation. Wheelys is a company provides different mobile sales vehicles.
Recently they developed a store called Moby where customers can enter through sliding glass door with an app including their pre-registered credit card information. In the store, there is no employee, no cashier, no queue, no waiting. Customers scan and bag the products (including fruits, potato chips, coffee, magazines, and even sneakers) with the app and then leave. The app automatically charges the customers.
The Swedish company is testing the 24-hour mobile grocery store in Shangai. The store is also a self-driving vehicle. However, the self-driving technology that depends on AI has not been implemented yet.
They opened a website where they answered questions and claimed that Amazon Go has a copy of their idea.
Cons of self-checkout systems
Even it sounds like amazing and provides peerless customer experience while reducing labor cost, it has some disadvantages.
- All purchases and transactions are made online with pre-registered credit card information. This may be tempting for hackers.
- Inventory Loss
- No matter how advanced the technology is, there will be people trying to steal because employees are unable to monitor customer transactions as closely on self-checkouts.
- Lack of Personal Interaction
- Although the key benefit is eliminating the customer engagement to decrease waiting times, many customers prefer to have a one-on-one interaction with cashiers.
- Adaptation difficulties
- Sometimes bar codes and coupons don’t scan properly, products require age verification, or customers need assistance
Interested in trying this out at your business and looking for an AI vendor? Let us know