The Case Against Self-Checkouts

The Associated Press News reported on an interesting poll this week showing that 70% of people prefer a hybrid store that allows you to choose between self-checkout and traditional human-operated cash registers.

Personally, I found this outcome a little surprising.  

I was even more surprised to see that only 13% preferred fully autonomous stores. Maybe not that surprising if you consider the fact that only 3% have actually shopped in one. 

Given that more and more supermarkets and stores are starting to install these self-checkout kiosks, I became a little worried.  

I will begin with a bold (no pun) statement: For supermarkets, I think self-checkouts are not that great. The way I see it, they don’t make matters better, they actually make matters worse.

Why?

Imagine you are at your typical Target or CVS where there are some cashiers and a few self-checkout stations.

When the store is not congested, you have the option. If you are one of those who are slower at self-checkout or have complex items (alcohol or fresh produce, that need weighing) you can choose the traditional option (cashier). If you are not high on human interaction that day, or just have primarily pre-packaged items, you can choose to self-checkout. People of each group become better at doing the work needed to check out in their respective choices. 

But when there is congestion, lines behind the cashier become too long and people start drifting to the self-checkout kiosks. The issue: they are not well "trained". Neither in terms of their speed nor their choice of items. The combination of congestion and slow processing makes things worse; the lines become longer and everyone becomes more annoyed. 

In fact, we know that when service systems (or congestion-prone systems, like roads) give people too many options, it makes things worse. 

Technology is not the limiting factor here. It’s the humans operating these technologies, and in this case, it’s the customers.  

But the reality is that more and more stores are adding self-checkouts. Walmart is adding more, Taco Bell just announced the “store of the future” with 3 of its 4 lanes dedicated to mobile pickups, and McDonald’s has been installing self-service kiosks for several years. 

In fact, both Forbes and The Washington Post report on a productivity boom driven by better technology and higher wages. 

So let me try to abstract from my own bias and analyze the situation more thoroughly. 

Self-Service vs Self-Checkout

First, I would like to distinguish between two types of services where technology is replacing humans. 

• Self-Service Kiosk (McDonald's): The product is Made-to-Order. The cashier's job is done prior to assembling the order (making the burger). 

• Self-Checkout (Supermarket): The picking (from available inventory) is done by the customer, prior to the cashier’s work.  

I would argue that the role of technology is quite different in each model. It is, therefore, important to first understand the process and the customer journey in absence of the new technology, and only then explore the role of technology. 

So, for example, when shopping at the supermarket, the process is as follows:

Note that the customer repeats steps 1 - 5 until all products are found.

This is clearly very cumbersome and has quite a few non-value-added activities. One would even argue that all of them are non-value added. And the reality is that self-checkout only replaces step (8), the difference being that the customers do the work rather than the cashier. 

What's the tradeoff? The firm doesn’t have to pay a cashier, reducing its labor costs. 

I can see the benefits, but they don't come without a cost since the processing times become much slower. In fact, I would argue that we underestimate the cost of additional congestion driven by these self-checkouts, as I described at the beginning of this post. 

If the only option is to automate, I would choose to automate all cashiers and just have helpers that assist with difficult products, moving between the cashiers. This is the idea behind Walmart’s “hosts”.

 Let's now look at a similar analysis of McDonald’s:

The role of technology is quite different in this model. The self-service kiosk essentially replaces steps (1) to (5). 

What's the tradeoff? Lower labor costs, while also allowing the customer more time to look through more options, driving more sales. As Forbes reports, on average, consumers spend 12%-20% more when ordering with their eyes and by touch from a self-service kiosk than when ordering from a cashier. What's the downside? Again, lower processing speeds. The customer is just not as fast as the cashier and the time consumed making choices that were previously made while standing in line are now made while occupying a resource. 

As we can see, one important benefit of the kiosk in the Make-to-Order model is that it can create upsell opportunities that do not exist in the supermarket model. But in fact, the cashier in the supermarket is probably better at suggesting deals when they see you browsing through the aisles and will definitely hand you your discount coupons and the weekly sales flyer at checkout!

Note that in both cases, technology (when operated by untrained humans, i.e., the customers) is slower than trained humans, i.e., the cashiers. 

I would argue that the self-operated options are, in fact, worse. It’s not just the average processing times that are slower, it’s also the variance in processing times that becomes worse. 

Variance significantly increases congestion, exactly at the wrong time...When the system is overloaded.  

People add significant variance to the process since they are just not as efficient as cashiers who do the same actions over and over again. Given the limited menu options in self-service kiosks, such as McDonald’s, the additional variance is quite low. For self-checkout at supermarkets, given the number and complexity of items, as well as the edge cases (alcohol, fresh produce, etc.), the variance is quite high. 

To summarize: Self-service kiosks at Make-to-Order restaurants can be quite valuable. However, when implementing self-service in supermarkets, the challenges may dominate the benefits, especially when you leave customers with options but without support. 

Luckily, this is not the end game. 

So, what is the End Game?

Three new innovations emerged over the last few years to address these challenges. 

Amazon Go:

Amazon Go is a chain of convenience stores operated by Amazon. The stores are almost fully automated. Customers in these stores are able to purchase products without being checked out by a cashier or using a self-checkout station. You just pick an item, the product is added directly into your real and virtual basket, and you are charged upon exiting the store. 

If we can scroll back to my description above, these stores remove steps (7) to (11), which are all non-value-added steps. Labor cost is clearly lower (although, you still need people to stock the shelves), there are no upsell opportunities, but just like any self-service model, there are plenty of opportunities for theft. 

Dark Stores and Delivery Models

Many of the current e-grocers pick and pack orders from physical stores. However, many of them are considering automated warehouses (dark stores). For example, Instacart plans to develop larger, stand-alone, roboticized warehouses. This way, firms can remove steps (7) to (11), as well as step (1): no shelf stocking is necessary and you can optimize for automatic pickup, instead of optimizing customer purchase behavior. The additional cost, of course, is step (12); driving the products to the customer’s home, which is now incurred by the firm. 

Mobile Orders and Self-Pickup

Customers order online on their mobile app, but pick up at the store. This is what Taco Bell is describing in its future stores. This is already available in many places. When I need a quick coffee on the go, Dunkin’ Donuts has one of the best app-store integration (no pun). I make my choice when I leave my parking garage (while walking), and pick up the coffee on my way to the classroom without even a single moment of waiting. Maybe not the best coffee out there, but definitely the fastest. 

The advantages are clear: you are removing steps (1) to (5), in the Make-to-Order model. You remove the need to use a kiosk; it’s always going to be expensive to buy and install hardware and even more expensive to maintain it. 

Newer technologies are going to emerge, so the question is: How can we judge the validity of such new innovations?

The three questions you should ask yourself is whether this new technology can:

• Reduce costs (labor, rent, etc)

• Improve efficiency (speed, waiting times)

• Improve value to the end customer (sell more, better selection, upsell better products)

In particular, can you simplify the process and make the process more predictable, so you gain these benefits? 

It is clear why Taco Bell is going to the “Taco of the future”: you get all three. But even this is clearly just a phase. In a future of self-driving cars, for example, why can’t my taco or coffee trace me and meet me on my path? In a future where all of us will have AR/VR goggles, we won’t need to use our mobiles. We will only need to look at something and “think” that we want it, and it will show up. We are only a few years away from this reality, and I am eager to experience it. 

I am curious to hear about new innovations you have seen that add value and that doesn’t just focus on reducing costs.