It morphed quite logically, actually. Remember the original point was to use bike couriers to reduce the number of delivery vans. That was solved by having the vans pick up parcels in crates already organized by zip+4.
Then the only question becomes how to make the connection between the van's dropoff location and the delivery address. Amazon is planning to have lockers by people's front door, but they could also have a single locker station in each zip+4 area. The most the van driver would have to do then is sort the parcels into the lockers by address, which would go much faster than driving around doing door-to-door delivery.
Then you say people don't want to go pick up their own parcel in their zip+4 area. Ok, for those who don't they can pay an extra $1 to have it delivered by a bike courier. You say that will add extra cost to shipping, but how much less will shipping cost by having the van drivers only shuttle packages to a single dropoff location in each zip+4 area instead of wasting time going door to door? Every van you cut out of the system saves money UNLESS the vans are being sold at $10k for a profit, in which case the more vans they use, the more profit they make from selling the vans. If that's the case, there's not much hope in using efficiency to save money, because they're making extra money by being more inefficient and thus selling more vans.
edit: I found this article that lays out what the business idea behind Amazon's plan: It seems that Mercedes vans are going to be sold and they're trying to find entrepreneurs to buy 20-40 vans, so apparently there's an overstock of these vans and they want to boost their sales numbers and thus sell higher sales to investors. So I'm sure they wouldn't want any bike-couriers or other efficiency tweaks to get in their way of maximizing van sales by billing it as an opportunity for entrepreneurs to buy the vans in fleets.
https://www.fastcompany.com/40590799...r-its-packages