First, you need more POS stations, because customers are not as fast as trained/experienced cashiers/baggers. So you need more machines to serve the same volume. That cost might be worth the lower labor costs.
But those extra stations also consume space. Picture a Walmart that dedicates one entire edge of each store to a row of checkout lines. It's easy to imagine that 10% of their public floor space is devoted to that function. If it takes 50% longer to self-checkout, you might need 50% more floor space to handle the same volume of purchases. That's not completely accurate because stores don't convert all their checkout lines. But you get the point.
I have noticed that most places compensate for the space requirements by drastically compressing the space for a station. No more conveyor belts. No more funneling customers throught a gauntlet of magazines and chewing gum. They might replace 4 checkout registers with 10 self serve stations. In many ways this can be better for the customer but worse for the business. There's a lot of money to be made in all that merchandise immediately surrounding checkout stands. Even the endcaps on the aisles closest to the checkout lines are worth special attention for maximizing sales. Losing this hurts the business.
I'd be curious to see the results of rigorous studies. In the meantime, as a customer, I like the self checkout stations.
I'd be super curious to see things completely flipped on their head. I'd like to see a grocery store where you walk through aisles that have only one item per SKU on the shelves and you point your phone at the qr code to "buy" one. The store app lets you indicate how many of that item you want. In the back room you have robots feeding your items up to a boxing station. In the time it takes you to walk to your car after pushing "Finalize and Pay" your order is waiting at the will call loading dock as you drive through.
No shoplifting losses, dramatically lower security costs, better and cheaper inventory management... A long list of benefits. And of course an offsetting list of increased costs. For example: smaller sales floor requirements, huge loading dock requirements.
I mean, this is HN, why do people still go to the grocery store instead of using a delivery app? Price? Browsing an app is slow? Item availability? Time? The app sucks? Orders always come wrong?
The grocery store offers an old fashioned produce section so I can hand pick a pear that isn't bruised by a robot.
And all the reasons you list.
Part of the beauty, though, is that it becomes obvious to offer both. Developing in-store system puts you 90% of the way to having shop-from-anywhere abilities. Developing shop-from-anywhere puts you 85% of the way to having in-store shopping without shopping carts.
In the US, there used to be a store like this back in the 80s-90s called Service Merchandise
They weren't supposed to. They were supposed to divert money away from checkout staff and towards the company making and servicing the terminals.
(Is my cynical streak showing? ;) )
modern self-checkout machines and touchscreen POSes are perfect examples of heaping more layers of abstraction on top of abstraction on top of abstraction on top of a simple underlying system that really didn't need all those abstractions, but they sure look fancier and apparently that's all that matters, productivity be damned.
The stupid self-checkout things... clunky, inaccurate, awful touchscreen calibrations, flag needlessly for "help" automatically. Whoever wrote those systems should feel bad for what they produced. I can't imagine that in the long term, they've helped the store save money or save anyone time.
As a young kid, I learned to use an AS/400 system and got fast at it quickly. I did a ton of data entry for my father's company. It felt so accurate and fast in comparison to the UI garbage today.