It'd only be a key question if one assumes the technology isn't going to radically improve over the next 5 or 10 years. It's going to. Deploying to consumers and dealing with such problems, learning from how millions of people use something, is a requirement of that process of improvement. It's similar to people questioning the GUI in the very early days of the PC, because working directly with text operating systems was still superior. It was extremely widely argued that GUIs were junk, were never going to catch on, were too slow, too glitchy, didn't add enough value etc - it's an obvious and common failure to look beyond the tip of one's own nose.
I looked into it a while back. The conclusion I reached was that it might be worth it if driving to the grocery store/Walmart, shopping, and bringing your supplies home was a hassle for some reason. But Walmart tended to be cheaper than even Prime Pantry and you have to be somewhat selective in shopping in Amazon. (You'll find some items are randomly pricing.)
In my case, there's a Walmart I can drive to in about 5 minutes so stocking up on paper towels and dishwashing detergent every few months (and/or when I'm at the grocery store anyway) is pretty easy.
Dash seems limited by being tied to one item. I don't get the appeal of Dash either because items I know I need a reoccurring supply of, I get a subscription to.
how many doll houses is a person willing to return in order to make their soap purchase easier?