Self-Checkouts: Retail’s Double-Edged Sword of Innovation and Irritation

This article is written in times of great sorrow and polarization around the world, so I hope we can at least unite around a simple fact. The fact that self-service checkouts are pure evil. 

If you agree, I suggest you keep reading as I will feed your confirmation bias. If you disagree, I also suggest you keep reading as, hopefully, you have a falsification bias bone in you, and I can convince you. And believe me, if I can’t convince you here, I never will. 

What prompted me to revisit this issue? Walmart dropped a bombshell recently by saying goodbye to many of their self-checkout lanes! This move caught the attention of The Atlantic, which decided to look deeper into this. Turns out, Walmart isn’t alone; more stores around town are putting the brakes on these DIY checkout machines:

“Walmart is pulling self-checkout lanes from at least three stores, requiring shoppers to have an employee ring up their orders. Two stores in Albuquerque, New Mexico, were recently renovated to replace the checkout lanes, and changes to a third area store are slated to be finished in October, Walmart spokesperson Josh Havens told Insider. ‘We continually look at ways to provide our customers with the best shopping experience and that includes adjusting the checkout area in stores,’ he said.”

Of course, it’s worth mentioning that I’ve always harbored reservations about these machines as I’ve already critiqued their efficiency (or lack thereof). However, it’s essential to approach this topic with more thoroughness and precision. To genuinely comprehend why firms like Walmart are moving away from self-checkout registers, it’s important to delve deeper into the intricacies and underlying reasons behind this decision.

Theft

One prominent reason behind the decline of self-checkouts is theft:

“Theft, as well as the losses from the unscanned and mis-scanned items that poorly designed kiosks create, is a trade-off of which retailers are well aware. (Companies contacted for this story, including Walmart, CVS, Costco, and Kroger, either did not respond or declined to comment on self-checkout’s relationship to theft.)”

First, there’s no denying the escalating incidents of theft in retail—both organized and sporadic. This surge has resulted in a noticeable inventory shrinkage, which is unquestionably a pain point for retailers. According to Business Insider, U.S. retailers lost more than $112 billion to inventory shrinkage, including over $41 billion due to theft.

The introduction of self-checkout lanes in retail stores has been identified as a contributing factor to increased retail theft, due to reduced human oversight compared to traditional checkout lanes. Shoplifters may feel emboldened by the lack of security, and some might even morally justify stealing small items if they perceive the self-service option as a way for the retailer to cut costs at their expense​. At self-checkouts, consumers might engage in fraudulent activities such as switching barcodes on items, hiding smaller items inside larger ones (the “honest mistake”), or covering up an item’s barcode while scanning it (the “machine did it”)​.

The rate of theft at self-checkouts is reportedly 5 times higher than at traditional cashiers, with retailers experiencing a 50% higher rate of loss when using self-checkout. This underscores a significant downside to the convenience and cost-saving potential offered by self-checkout technologies.

The financial toll of theft at self-checkouts is significant. In the U.S. alone, over $3 billion worth of products are estimated to be stolen annually via self-checkout terminals, highlighting the challenge retailers face in maintaining the integrity of these systems​​. In a particular survey of $21 million in sales, almost $850,000 worth of items were never scanned and purchased at self-checkouts​​.

The statistics and insights gathered indicate a trade-off between the potential cost reductions gained through self-checkouts and the increased vulnerability to theft. While a human cashier isn’t a foolproof deterrent, it undoubtedly provides an additional layer of security. Self-checkouts have struggled with theft prevention for a while —for example, they weigh the bag to identify whether an item is missing and then flag a supervisor. But flagging a supervisor is a daunting task, which gives me little confidence that stores are quick to respond when there’s an alert about theft. This all contributes to a perception of reduced safety. 

Cost Cutting, but…

Self-checkout systems were often introduced as a cost-cutting measure, but I would argue that they’ve been overestimated. 

Traditionally, during off-peak hours cashiers were tasked with restocking shelves, organizing the back room, and cleaning aisles. When supermarkets decided to introduce self-checkout kiosks, the main trade-off was the fixed costs of these machines vs the reduced labor costs. 

However, the first glaring issue in many of these stores was that there were no cashiers available during off-peak to stock shelves. Consequently, stores with more self-checkouts often present a more disorganized front. The realization started to set in for retailers: they needed the personnel. Not just for checkouts but to ensure smooth store operations, from stocking shelves to general store maintenance, to attending to customers that need assistance, or to a kiosk alerting for theft. So, when self-checkouts took over, many retailers found themselves in a catch-22 situation, needing to rehire staff for these tasks and, as a result, witnessing the promise of cost reductions evaporate.

Furthermore, many of these machines are quite buggy and need constant maintenance or attention:

“While the idea of being able to use a self-checkout machine to complete one’s shopping quickly and efficiently may seem like a pleasing concept for shoppers looking to get in and out of stores as quickly as possible, this expectation doesn’t always line up with reality. CNN reported that 67% of surveyed shoppers who’ve used self-checkout machines experienced problems in doing so.”

This is true with almost any self-service technology, as it requires more human attention than initially estimated.

Why Customers Should Hate Self-Checkouts

Among people between 18-34 y.o, more than 67% prefer self-checkouts. 

I get it, you don’t like people, and the idea of human interaction scares you. You also probably don’t know how to cook (so you don’t buy fresh produce), and you probably don’t drink much alcohol either, so you don’t require much assistance at the self-checkout kiosks. Also, the perceived service time is less when you’re busy beeping products rather than waiting in line.

Ok, I’ll stop with my millennial stereotypes... 

The truth is that the underlying notion we must address is the efficiency of self-checkouts. 

In ideal conditions, self-checkouts may be marginally less efficient than human cashiers. On a day-to-day basis, and for someone well-acquainted with the self-checkout process, this marginal delay might not even be noticeable. Experienced users, for instance, might instinctively avoid the self-checkout when they have items like alcohol, which requires age verification (hence assistance from a store associate), or fresh produce —many self-checkout systems rely on weight-sensitive platforms which may not always register certain fruits or vegetables correctly, requiring multiple scanning attempts or staff assistance.

However, the true challenge arises when the system becomes congested. In periods of high demand, the self-checkout lanes also attract less experienced users. These novices may have items in their basket that require additional time and attention (like alcohol or fresh produce), leading to hold-ups.

But the main problem with self-checkouts isn’t the prolonged waiting times, it’s their inherent unpredictability. From a queueing theory standpoint (which studies waiting times, stoppages, and queues of various processes), this unpredictable nature introduces high variance, which is particularly troublesome in operational systems. Ideally, these systems desire consistency and predictability, ensuring a smooth flow of users and efficient utilization of resources.

Allow me to illustrate:

Imagine each shopper holds a bag of 12 items, on average. For simplicity's sake, let’s assume that scanning each item takes around 30 seconds. This means that, on average, a consumer would spend roughly six minutes at checkout.

But when there are items that could unexpectedly increase the checkout time (like alcohol and fresh produce, as mentioned above), things become tricky. 

So, let’s consider an M/M/2 queueing system (where 1 server is a human, and server 2 is a kiosk) with an arrival rate of 16 customers per hour, and each server having a service rate of 10 customers per hour.

I know some of you think you’re faster than a cashier, but… Let’s look at the waiting time:

“There’s little evidence that self-checkout is reliably faster than the old cashier system, and that feeling of convenience has always been largely a trick of perception. ‘Trained cashiers can scan and bag goods faster than even the most aggressive or enthusiastic shopper,’ conceded a 2002 New York Times story about the machines’ growing popularity.”

I simulated the system, and the assumption is that in off-peak hours, people will choose the right server for them —people who don’t have produce or alcohol will choose the self-checkout or cashier, and those who do will consistently choose the cashier. 

This is what the waiting time distribution looks like: 

And here is a summary of the waiting times:

Average waiting time: 0.135 hours (or about 8.1 minutes)

The standard deviation of waiting times: 0.162 hours (or about 9.7 minutes)

These are not meant to be correctly assessed by any particular supermarket and are made for illustration purposes only. 

But what if during peak time, people chose not the right queue, but the one that’s available?

Let’s run a similar queueing system with a slight deviation: 

With a 90% probability, the service time is 6 minutes.

With a 10% probability, the service time is 18 minutes (indicating the need to wait for the supervisor to check what’s going on and enable the transaction). 

This does two things: it simulates the exceptions that we get when people don’t choose the right queue and must wait for assistance to continue scanning, but also the slightly additional congestion (which will prompt customers to choose the available queue, rather than the “right queue for them”). 

Note that this is a small increase in server utilization. The impact on the waiting time is quite substantial, but even more so, the impact on how the waiting time looks like:

Average waiting time: 0.318 hours (or about 19.1 minutes)

The standard deviation of waiting times: 0.381 hours (or about 22.9 minutes)

The main issue is evident: if you’re unlucky and get stuck behind someone that encounters a problem, it becomes your problem too. The main takeaway is the unpredictability of wait times, which not only increase, but also spike erratically, making the self-checkout experience more time-consuming than anticipated.

You may say, “But I like self-checkouts.” Well, we know that people tend to ignore their impact on others behind them in the queue. 

So you may wonder whether we’re doomed to have human cashiers forever. The answer is “Not necessarily,” which brings us to potential future solutions.

Amazon Go, for example, promises a seamless shopping experience without traditional checkouts. However, it has its own set of challenges. The system demands minimal product variance and optimal lighting conditions for accurate item detection. Any deviation from these ideal conditions can hamper its effectiveness. Thus, despite the technological advancements, at this point, the human touch remains irreplaceable in the checkout process. Interestingly, this scenario provides an employment opportunity, especially for individuals with disabilities or those needing flexible work conditions. It’s a testament that while automation has its place, human roles in certain sectors continue to be indispensable.

In a world racing toward automation, self-checkouts stand as a slightly awkward middle position —not quite the seasoned, reliable worker nor the futuristic, flawless machine. They were retail’s first flirtation with reducing human interaction —a test if you will— and like a poorly constructed IKEA table, they’re wobbling under the weight of reality. The thefts, the unexpected item in the bagging area, the long lines as one solitary employee rushes between six malfunctioning machines… it’s all a far cry from the swift, efficient future we were promised. So, as we stand, somewhat irritated in a not-so-quick self-checkout queue, we’re reminded that perhaps, just perhaps, the future isn’t as self-served as we imagined. As for those human cashiers, well, they aren’t just good at spotting a shoplifter from a mile off, they can also spot an overly ripe avocado, a task no machine has managed thus far. So next time you find yourself stuck at self-checkout, remember, it’s not just checking out your groceries; it’s checking out the limits of automation.