I always thought that self checkouts were faster than traditional cashiers, but this article made me aware of the unexpected costs. Great read and it’s awesome to see how human cashiers are still important!
I personally would always choose a cashier over a self checkout (perhaps my age is playing here) just because I find the self checkouts annoying when there is no bar code (loose grocery items) or the items fails to scan. Finding a Sweet Potato as an example is a pain. Its all the way buried in a page in nearly 6-7 pages that start with S. Then when you do get there, you have to find the variant - organic vs non organic vs ... and that just slows me down.
I also feel bad for being slow if there is line behind me. I do not feel that bad when there is cashier because I am not slow, its the cashier. I know, I know --- but that is how brain functions.
Interesting, and a great (more nerdy) addition to the article in The Atlantic!
However, in both cases I’m missing the distinction between the horribly designed self-checkout kiosk machines placed at the exit of the store compare to the scan-as-you-shop solutions.
There are different types, either dedicated handheld devices or an app on your phone using the camera to scan barcodes.
Both have their pros and cons, the hardware being much faster to scan and the app enables payment with just one tap. But in my experience this offers a much smoother experience overall compared to the machines.
Is this not a commonly used solution in the US, does it have some negative effect I’m not aware of, or why is this not mentioned? Just curious.
I know of several experiments with these in the US, but most were unsuccessful. I think the issue of theft is even more prominent there (easier to hide things and harder to verify what was put in the cart). I agree that this will be a better slution and a more scalable one, if this can be addressed.
In Sweden I would say self scanning is the default for pretty much all larger stores (big-box and shopping centers), while smaller convenience style stores in the middle of the city prefer kiosk style machines.
When using self scanning you can be selected for a manual inspection at checkout, ranging from a few randomly selected items to going through your entire cart. The result of this builds your reputation for the store membership, so with less issues found the probability of a manual inspection decreases (though never to 0).
There is obviously the risk of theft, but combining manual inspection with a high-trust society this seems to be working well enough to have been in use for the past 20+ years.
I would think there is some breakage percentage that the retailer's model has built in, and that perhaps does not breach in high trust societies OR in high trust areas of US.
Having read your article, I realize that self-checkouts pose many inefficiencies for customers and store owners. However, I’d like to point out that self-checkouts occupy much less room than traditional registers, a vital difference for stores that have limited floor space. Stores therefore can have multiple self-checkouts in place of cashier kiosks. If we factor in a ratio that represents this, I wonder how it will impact your analysis. While I agree that the waiting time distribution seems significantly worse when the comparison is one-to-one, I think self-checkouts may gain an advantage (or at least decrease their disadvantage) when we account for a higher self-checkout to kiosk relationship.
Prof. Allon, whether a store has human cashiers or self-check-outs, having a single line should reduce both the wait time variability and customer anxiety from picking the "wrong line." Most of the stores I visit have banks of self-check-outs that are fed from single lines. So more like a M/M/4 or M/M/8 setup. Also, there are often human employees assigned to each bank of self-check-outs, reducing the delay in getting help with produce or age-restricted items, and also likely watching out for theft.
I would be interested to see what your #'s look like if you run as M/M/4 and M/M/8 with either all human cashiers or all self-check-outs.
Furthermore, a 3rd option is scan-and-go, where the customer scans each item as they put it into the cart. This is being pushed by Sam's Club (sometimes with special discounts from using it), and then everyone gets their receipt checked when exiting, no matter how they checked-out. Having the customer scan the item as they are picking it probably reduces perceived time spent scanning and also reduces the likelihood of mis-pricing surprises at the check-out line.
A few supermarkets tried these. I think the issue of theft if even bigger there, but maybe checking everyone on the way out can solve some of these issues.
I always thought that self checkouts were faster than traditional cashiers, but this article made me aware of the unexpected costs. Great read and it’s awesome to see how human cashiers are still important!
I personally would always choose a cashier over a self checkout (perhaps my age is playing here) just because I find the self checkouts annoying when there is no bar code (loose grocery items) or the items fails to scan. Finding a Sweet Potato as an example is a pain. Its all the way buried in a page in nearly 6-7 pages that start with S. Then when you do get there, you have to find the variant - organic vs non organic vs ... and that just slows me down.
I also feel bad for being slow if there is line behind me. I do not feel that bad when there is cashier because I am not slow, its the cashier. I know, I know --- but that is how brain functions.
Interesting, and a great (more nerdy) addition to the article in The Atlantic!
However, in both cases I’m missing the distinction between the horribly designed self-checkout kiosk machines placed at the exit of the store compare to the scan-as-you-shop solutions.
There are different types, either dedicated handheld devices or an app on your phone using the camera to scan barcodes.
Both have their pros and cons, the hardware being much faster to scan and the app enables payment with just one tap. But in my experience this offers a much smoother experience overall compared to the machines.
Is this not a commonly used solution in the US, does it have some negative effect I’m not aware of, or why is this not mentioned? Just curious.
I know of several experiments with these in the US, but most were unsuccessful. I think the issue of theft is even more prominent there (easier to hide things and harder to verify what was put in the cart). I agree that this will be a better slution and a more scalable one, if this can be addressed.
Thanks for the reply!
In Sweden I would say self scanning is the default for pretty much all larger stores (big-box and shopping centers), while smaller convenience style stores in the middle of the city prefer kiosk style machines.
When using self scanning you can be selected for a manual inspection at checkout, ranging from a few randomly selected items to going through your entire cart. The result of this builds your reputation for the store membership, so with less issues found the probability of a manual inspection decreases (though never to 0).
There is obviously the risk of theft, but combining manual inspection with a high-trust society this seems to be working well enough to have been in use for the past 20+ years.
I would think there is some breakage percentage that the retailer's model has built in, and that perhaps does not breach in high trust societies OR in high trust areas of US.
Professor Allon,
Having read your article, I realize that self-checkouts pose many inefficiencies for customers and store owners. However, I’d like to point out that self-checkouts occupy much less room than traditional registers, a vital difference for stores that have limited floor space. Stores therefore can have multiple self-checkouts in place of cashier kiosks. If we factor in a ratio that represents this, I wonder how it will impact your analysis. While I agree that the waiting time distribution seems significantly worse when the comparison is one-to-one, I think self-checkouts may gain an advantage (or at least decrease their disadvantage) when we account for a higher self-checkout to kiosk relationship.
Good point. In the next iteration, I will try to factor this as well.
Prof. Allon, whether a store has human cashiers or self-check-outs, having a single line should reduce both the wait time variability and customer anxiety from picking the "wrong line." Most of the stores I visit have banks of self-check-outs that are fed from single lines. So more like a M/M/4 or M/M/8 setup. Also, there are often human employees assigned to each bank of self-check-outs, reducing the delay in getting help with produce or age-restricted items, and also likely watching out for theft.
I would be interested to see what your #'s look like if you run as M/M/4 and M/M/8 with either all human cashiers or all self-check-outs.
Furthermore, a 3rd option is scan-and-go, where the customer scans each item as they put it into the cart. This is being pushed by Sam's Club (sometimes with special discounts from using it), and then everyone gets their receipt checked when exiting, no matter how they checked-out. Having the customer scan the item as they are picking it probably reduces perceived time spent scanning and also reduces the likelihood of mis-pricing surprises at the check-out line.
A few supermarkets tried these. I think the issue of theft if even bigger there, but maybe checking everyone on the way out can solve some of these issues.