I continue to be puzzled at the high percent of my purchases which are flown to the Seattle area. I am talking about everyday sorts of things. Potato Chips, jams, spices, denims, food grater. If they are not two-day delivery they sit back East (or Texas or Ohio) for a couple days, then suddenly on a plane and to my house. The algorithms to do all of this are simply mind bending. Obviously those thousands of highly paid Amazon workers in Seattle figure out the cheapest way to send things to particular addresses, and planes are often the answer.
I actually heard somewhere the average cost per package for Amazon Air is around $6 vs. about $6.75 for UPS Ground and $8 for UPS 2nd Day Air at Amazon's rates, however, it didn't specify whether the last mile carrier was AMZL or USPS.
BTW I heard the ground linehaul cost per package with USPS as the last mile carrier is about $4.55. Also, the old FedEx rates were about $7.25 for Ground/Home Delivery and $10 for Express 2Day.
I'm a little skeptical that the numbers you're quoting ($6 via Amazon Air, $4.55 via Amazon ground) are apples to apples for the following reason. An Amazon Air delivery is truck from FC to Sort Center then Truck to USPS DDU or maybe a Delivery Station. Then USPS or Amazon Flex (or other Amazon last mile) to Customer.
Amazon Air is all that, plus at least one air flight (possibly two, with a sort in the middle) plus another truck arc - it's truck from FC to Gateway, Air (at least one air segment, possibly two plus a sort in between), the Truck from Gateway to Sort Center.
In other words, the air delivery is much more elaborate, far more handling, many more chances for something to go wrong.
So, it might be the numbers you quote are correct, but the air packages are, say, materially more smaller/lighter.
To do this right requires all kinds of very very careful accounting. For instance, suppose you are committed to flying a certain flight. So long as you can ensure that the stuff that must be on the flight (time sensitive) is on the flight, then, once you have committed to flying the flight you may as well fill it up. Packages are, on average, pretty light, the marginal cost of adding an additional package to a flight you are already committed to flying, is pretty close to zero. I don't know whether Amazon is taking advantage of such phenomena to put more air packages on board, but economics would suggest that such moves should be considered at least.
It's possible there are calculations like that ongoing. However, I would say it's odd, overall, that fairly ordinary items are being supplied from so far away. It's an undesirable outcome perhaps reflecting Amazon's ongoing adaptation to Covid (which is one of the biggest disruptions ever to their network of FCs, trucks, aircraft, etc).
The planning problem for Amazon is more complicated than probably any other on earth. It's far more complicated than any passenger airline planning problem.
It's important to understand that, overall, delivery by air is a failure in the context of Amazon. This will almost certainly raise the hackles of some reading this - Amazon Air, a failure?! Well, no, obviously not. But looked at the right way, it's undeniably true that delivery by air is undesirable relative to other alternatives.
The cheapest form of delivery for Amazon is by truck from a nearby warehouse. Only when this is not possible is air used. In that respect, delivery by air occurs only when inventory placement has failed to stock the right amount of whatever near enough to the customer. And everytime that happens, it's a planning failure.
For sure there are always going to be rare items that can only be stocked in a few (or even one) place and for which air delivery will be pretty routine. But overall, if you're ordering something pretty normal and you live near a reasonable number of Amazon warehouses, the best solution for Amazon (in terms of cost of fulfillment, including cost of delivery) is if it's delivered by a truck from a nearby warehouse.
Given that the number of Amazon warehouses is still increasing, the reality is that the average distance of a customer to the nearest Amazon warehouse continues to decrease. That is to say, on average, the number of truckable deliveries ought to be increasing over time.
That assumes, however, that the product mix stays static (obviously not true) and that average promise (days from when you click to delivery) is static (also not true - when Prime went from nominally two days to nominally one day, that obviously changed the equation for Amazon quite a bit). But at this point, there's little that Amazon doesn't stock - Amazon must be way down the "long tail" of retail in terms of increasing product mix (in other words, most things added to the site at this point are things that, in the grand scheme of things, are not ordered that often on a relative basis - meaning they should have little impact on ability of Amazon to deliver the vast majority of orders from a local FC).
But if those two conditions were met (or largely met) then so long as the number of FCs continues to increase across the nation, you'd expect the *proportion* of Amazon Air deliveries to drop relative to truck deliveries. Overall, Amazon is still growing, so even if that proportion were dropping, it might be overwhelmed by the increase in overall flow. Also, of course, currently life is anything but normal - Covid has put supply chains in general and clearly Amazon in particular under enormous strain. So, as Amazon continues to adapt to life in the age of Covid, for sure we could expect a bit more than average weirdness in terms of a delivery "failure" like delivery by air when delivery by truck would be the more desired (by Amazon) outcome.
To change this up just a bit - moving from two day to one day Prime likely diminished the importance of the existing daytime Amazon sort at CVG quite a bit. That sort only supports a two-day delivery, because you need a nighttime sort to support one-day delivery. That nighttime CVG sort will come on line when Amazon's own hub is built.
But, in the meantime, until that hub is built, One Day Prime likely meant a big increase in the relevance of point-to-point flights and, perhaps, a nighttime cross-dock operation in ILM - I assume that's what it is (cross dock is a more primitive container-level hub). Is that right? Perhaps there are other nighttime cross docks?
The importance of the new regional air sort buildings (AFW, etc) becomes even greater to support those outbound point-to-point flights until that CVG hub comes on line.
That's my guess anyway.