Amazon Go: The Future of Grocery Stores

Analyzing the operational benefits of Amazon Go grocery stores.

By now, most of you have probably heard about Amazon Go – Amazon’s sans-checkout grocery stores. Through a combination of computer vision, sensor fusion, and machine learning1, Amazon allows customers to shop without checking out. There are plenty of articles explaining how the technology works,2 so this blog post will focus on the operational process improvements to Amazon and its customers.

The Shopping Process

Traditionally, customers enter a grocery store, put items into a cart as they shop, and visit a cashier before leaving. Since shopping patterns are variable, there is often a line of customers waiting for a cashier or a plethora of cashiers waiting for customers. This leads to unhappy customers and lower labor utilization. This problem isn’t new: over the years, companies have introduced “10 items or less” lines and self-checkout stations to try and alleviate it.

At Amazon Go, a customer must create an Amazon account and download their app before visiting the store. To enter, the customer scans her smartphone into a subway-like turnstile and begins shopping. Instead of putting items into a cart, the customer puts items directly into a bag. Once the customer is finished, she simply scans her phone at a turnstile on the way out. Within a few minutes, she will receive an email with a detailed receipt of the items she purchased.

Customer Enthusiasm

In a survey of 1,000 US consumers, 75% said they would be “extremely likely” or “very likely” to shop at an Amazon Go if a store opened nearby. Additionally, 25% of the respondents said they would pay higher prices to avoid waiting in a checkout line.3 While I believe these results are overall positive, I note that only 30% baby boomers were open to shopping at an Amazon Go.

Cost Savings

Increased customer demand allows grocery stores to spread out labor costs and fixed costs over a larger denominator, holding labor constant. However, Amazon Go will be able to employ fewer people, as they will no longer need to employ cashiers and baggers in their stores. A survey of six managers of major Midwest retail and grocery stores, five believed that these entry level jobs would be eliminated. Furthermore, all six thought the technology would make inventory management more efficient and reduce price/coupon errors.4 Overall, it appears there will be significant cost savings for grocery stores.

The Future of Amazon Go

Amazon has been slowly expanding the number of their stores. As of November 2018, there were three stores in Seattle, two stores in Chicago, and one store in San Francisco, with three more in the works.5 I expect Amazon to continue to slowly roll out stores in the next one to two years as they perfect their technology and wait for sensors and computing power to get cheaper.

In the long-term, it is hard to ignore their acquisition of Whole Foods and wonder whether Amazon plans to use their technology in all 400+ Whole Foods locations. It will not be a trivial task to scale the technology to the size of a Whole Foods store. Current Amazon Go stores have a relatively small footprint with a limited product offering. As you increase the footprint of a store, you will need to add cameras and sensor-enabled shelves throughout the store. Additionally, you will need to add granularity to the technology to be able to track more customers throughout a larger space and to be able to differentiate between a larger number of products. The data Amazon is collecting in their six existing locations will be critical to this effort.

New Bottlenecks

In the current format, customers must scan their smartphones to enter and exit stores. As more people begin to shop in these locations, I could imagine a queue of people waiting to enter and exit these turnstiles. Instead of saving customers time, Amazon may have just shifted the bottleneck from the checkout counters to the entry/exit points. Amazon would still have the cost savings associated with decreased labor costs but they may not have as many customers itching to shop with them. Going forward, I think Amazon could improve this technology to make entering and exiting the grocery stores more fluid. They could, for example, install sensors throughout the entrance of the store trying to catch a customer’s unique QR code as they enter a store. They could have one employee by the entrance who could greet customers and manually scan in any customer who the sensors didn’t catch.


The technology Amazon is developing is going to have major effects on grocery stores going forward. But how generalizable will this technology be? Will it change how we dine out? How we check into our flights? How we buy clothes? (784 words)


1, accessed November 3, 2018.

2, accessed November 3, 2018.

3 Redman, R. (2018). Amazon Go seen as welcome grocery option. Nation’s Restaurant News, July.

4 Polacco, A. & Backes, K. (2018). The Amazon Go concept: Implications, applications, and sustainability. Journal of Business and Management, 24 (1), March, 79-92.

5, accessed November 3, 2018.


Get Em While They’re Hot: Dynamic Ticket Pricing in Major League Baseball


Machine Learning at Emirates Airlines

Student comments on Amazon Go: The Future of Grocery Stores

  1. Yes, I think is this technology was available at scale it would transform almost every buying experience of physical items. Much the way amazon disrupted the way we purchase items, this is just another instance of them creating convenience for a slightly different buying experience and will further entrench themselves into the lives of 21st century daily habits.

  2. I’m not convinced this technology will spread out over our consumer life. I see it more as something to create a Buzz on Amazon and boost a customer experience for the niche early adopters eager for innovation and technologies in the market. The value proposition of time saving for me looks weak as most supermarkets already offer a self-checkout option and consumers are migrating to the online space. Online sales already represent 10% of retail total sales in US [1] and are growing at fast pace. On that sense, I question myself until what point will humans cope with this growing technology dependency/fixation and wake up one day desiring a warming good morning from the Trader Joe’s real cashier.

    1. Customers may one day desire a warm greeting from a customer service rep or cashier, however, consumers will have to pay for this greeting. Are you willing to pay another dollar for your Yogurt each time you checkout? My opinion is no. Human interaction will start to come at a premium but only in “luxury” cases. You will pay for the service during a luxury dining experience but not checking out in the grocery store.

  3. I think Amazon’s boldness in disrupting industries is remarkable. But, this time, I think that Amazon Go is a display of novelty that won’t really change how the industry operates. I wonder how much more efficient such a system can be, when taken to scale. As you mention, collecting the necessary information in a Whole Foods can be a daunting task, very error-prone. Mistakes could hurt greatly if users complain of over-charging.
    Regarding your question on how generalizable this can be; I think it might transform other ‘queues’. But there’s a human element in all this. Amazon Go may be a great proof of concept, but do we really need to eliminate one of the few human touch-points that are left in our shopping experience?

  4. Amazon has truly mastered the “invisible payment” with this technology, something we’ve seen succeed with Uber/Lyft. There will certainly be kinks to smooth out as Amazon Go scales up, such as incorrectly charging a customer or dealing with multiple guests with the same account shopping together. I’m less concerned about these glitches, which I think Amazon could quickly solve…rather, I think the value of this technology hinges on consumers’ preferences for physical grocery/retail stores. I wonder if we’ll see increased physical stores with technology like Amazon’s, leveraging ML, cameras, sensors and advanced data analytics; or instead, if we’ll see fewer retail stores as consumers move to on-demand grocery delivery services (already offered through Amazon Fresh or PrimeNow).

  5. Customers may one day desire a warm greeting from a customer service rep or cashier, however, consumers will have to pay for this greeting. Are you willing to pay another dollar for your Yogurt each time you checkout? My opinion is no. Human interaction will start to come at a premium but only in “luxury” cases. You will pay for the service during a luxury dining experience but not checking out in the grocery store.

  6. I believe that once this technology is available that it will be scaled to as many industries especially places like retails scores including big box, fashion and other such areas. It would be interesting however to look at what percentage of people are coming into stores for the ability to buy something with convenience and who is coming in for the customer service experience and how this would evolve the role of retail workers.

  7. I do think that this way of shopping will scale with time and the improvement in technology. People place a premium on efficiency and with how busy modern life has become, there is a lot of value in not having to wait in line. While you are right that in the near term the popularity of Amazon Go might lead to queues at the turnstiles, I am convinced that this is not a particularly difficult problem to solve even with technology we have today. There is an argument to be made for human interaction but grocery stores are not necessarily the place for that.

  8. Par for the course, I think Amazon is ahead of the game here–and expect this innovation to spread far and wide once they’ve demonstrated it can be achieved at scale. That may be a long time coming, however–the change-over costs will be enormous for big-box grocery (and other retailers), so I’d expect them to drag their feet on implementation until the market demands the service. For a parallel, consider automation in food distribution: though the robotic technology exists and is commercially available to eliminate “pickers” and reduce errors, the labor used is inexpensive enough and the build-out costs of automation high enough that virtually no distributor is yet fully automated.

Leave a comment