It probably has to do with traveling distances which vary greatly depending on the area. I don't want my frozen stuff to thaw while sitting around waiting for the next driver or during transit. I don't want my stuff squashed so vehicles need to be bigger to handle multiple loads, which also sort of implies inefficiency while traveling without a load.

I could see it working extremely well in big compact cities since they already have many delivery services. However, in the spread out areas it wouldn't work so well without many drivers which increases costs and the price of goods.

Also, most people want to inspect their fruit and veggies before they buy them. The interfaces could take input on the state but that would just slow things down IMO while a human would have to sort them. Unless there are systems out there for ripeness detection. I might buy something for tonight's dinner, in which I want something ready, or I might be preparing for a weekend dinner and want something that still needs time.

I believe all of these issues can be overcome, but I don't see how all that labor, vehicles, and tech could be a marginal cost. Well, I can see the tech cost approaching zero/order since all it needs is the existing catalog/POS system with a simple web interface. Most people I know either shop on their way home from work, a trip they are already paying for, or wait for the weekend to go as a group.