I was wondering how all those packages get delivered in the shortest time. Mail carriers have a specified route they drive every day but it seems an Amazon driver’s work load would be random. Do they have an application that shows them the next delivery location after they complete each drop?
Almost certainly they don’t try to get them all delivered in the shortest time. That, infamously, is an extremely hard problem to solve, even with as much computing resources as Amazon has available. You always just settle for a solution that’s probably reasonably close to optimal.
Yes, drivers are provided with a route, but they can deviate from it if necessary. Traffic or an event, for instance, may cause snarls that the system may or may not be aware of.
I live in a large apartment building. We have about 450 apartments. Many people receive Amazon deliveries every day. You cannot walk along my hallway of 20-ish apartments without seeing an Amazon box by someone’s door. It is not uncommon to see 4 of their vans outside our loading dock. It’s not uncommon to ride the elevator with a couple of different Amazon delivery folks who are not working as a team chatting about their day.
Whatever Amazon HQ is doing about routing drivers, it is clearly NOT about minimizing trips to my address.
We of course also see FedEx, UPS, and USPS package deliveries as well. I’m just talking about official Amazon drivers in their spiffy blue shirts and their blue smily vans and carts. If only they could be trained to take pix of their deliveries which include the apartment number in the field of view.
My condo campus is not quite as big as @LSLGuy 's apartment building, but my experience is similar. Although I might not see all the trucks at the same time, I probably see at least 3 to 4 trucks a day. I’m in the northwest corner of the property, so that numberwould probably be at least quadrupled for the entire complex.
There are a few YouTube channels hosted by folks who work for Amazon (as well as Fedex and UPS) and they bring you along as they do their shift.
It’s actually kind of cool to watch these videos and see how they set up their truck and so on.
I’m always surprised that they are allowed to film these videos.
The multiple drivers could be coming from different facilities. There are literally dozens of fulfillment centers and warehouses within 40 miles of our house (as the crow flies).
They don’t all carry the same products.
The mail carriers who deliver letters/magazines/small packages have a specified route that they drive or walk every day. But the USPS workers who deliver the larger packages or priority mail probably have a workload more similar to an Amazon/Fed Ex/UPS driver. I mean , UPS/FedEx drivers probably work in the same general area every day but it’s way more likely that no one on my block has a UPS delivery today than that nobody has a single piece of mail so the mail carrier always has to come down the street but the UPS driver may be able to skip it.
Or even all the variations of the same product. I once ordered four cycling jerseys from the same page on Amazon. The only difference between them was their colors. They came in three different packages. That puzzled me until I figured they were found in three different warehouses.
I did some seasonal work for UPS; I was given a load of packages and list of their addresses; we also had everything shown on a digital map (basically Google maps with the addresses shown). It was rural Wyoming, and I was very familiar with the area, so I chose my own routes based on my knowledge of the area (I had done survey work for 10+ years in my delivery area).
Not Amazon, but thought I’d share my perspective.
UPS trucks almost never make left turns. Right turns as much as possible (obviously left turns cannot be completely avoided). Supposedly this saves time and gas:
I’ve seen a few patent applications from Amazon and other such companies about systems and methods for optimizing deliveries. Route planning is definitely a part of that, so I’m sure they do have at least a general plan for the route in question. I’ve even seen some applications for changing the route on the fly, if something comes up, or if a new high-priority delivery order comes in.
I mean, when you think about it, to them, time and distance are both money. So they’d want the most efficient route possible. You don’t want your drivers wasting time going back and forth over the same road, if you can avoid it.
As @Chronos alluded to above this is actually a really difficult problem. Usually called the Travelling Salesman Problem. It turns out even really advanced computers have trouble optimizing routes (and no chance a human will do better). But, again as mentioned, they can get kinda close to a most efficient route, albeit not perfect.
Hence all the patent applications!
The point is, they absolutely do care about it, and so do use systems to come as close to perfect as they can. They create a plan based on known orders, track the vehicles with GPS, and send updates if something changes, so as to avoid as much downtime as they can.
Out here in the boonies we don’t have smiley vans nor do I think we have drivers in spiffy blue shirts, mainly because Amazon deliveries are outsourced to contract drivers. At least I presume so because I’ve seen a variety of different vehicles making deliveries, all of them unmarked. What remains impressive, though, is that on the day of the scheduled delivery the ETA gradually narrows down to a two-hour window. So whatever scheduling and routing gadgetry Amazon uses, the contractors use it, too. The two-hour delivery window on the tracking information is particularly impressive compared to the wonders of Canada Post regular package tracking, which if you’re lucky might show as “delivered” maybe three days after the fact (or not) and is equally useless at all points prior.
As to pix, however, the contractors are inconsistent – at least mine are. They all universally adhere to the Strict Code of Delivery Conduct, which is “throw the package on the porch and never, ever ring the doorbell”. The delivery status always says that the package was left on the front porch, but there may or may not be a pic. It’s very cool when I’m anxiously waiting for something and check the tracking and see a picture of my own front porch with my package, but this doesn’t always happen.
I drove a UPS truck for a few years (early 80s).
Not sure how Amazon does it, but when I was driving we were responsible for a defined, bounded area. The company engineers had defined the best route to follow through the area, and the truck was loaded* in this order. Supposedly, all we needed to do was drive to each successive address and only memorize the “next” package’s address before setting off. In reality, we soon learned our routes and customers’ habits better than the company and could greatly optimize this at the driver level.
We rarely ventured into another driver’s region, except for misloaded or traded** packages.
-
Almost every truck had one or more large buildings with lots of packages and it paid to deliver these first (example: it could be impossible to climb over the mall’s shipments to reach packages in the planned/loaded order).
-
Some businesses had atypical open/close times and these had to be accommodated, often shifting the delivery order for addresses nearby.
-
Some small business owners had odd schedules with multiple locations, and it was faster to drop their shipments where we expected them to be (ie. Jim managed the bowling alley on Mondays, so drop his boxes there, rather than at his repair shop).
-
We had to pickup packages in the afternoons from businesses that shipped from their own load areas. This required re-schedulign some delivery areas to avoid leaving halfway through a neighborhood and then returning to finish.
-
It was very common (in my day) for drivers to rid themselves of the large deliveries early in the day, then park for a while to examine the remainder and re-arrange to suit their knowledge of the area.
*loading order: In UPS trucks there are 8 defined sections on the four shelves behind the driver. These are selected/delivered in the order: front half of left lower shelf, front half of left higher – repeat for right side. Then scoot all the packages from the back “half” of the shelves forward, and repeat the original order. Oversized packages were located immediately below their “normal” location on the shelves above.
**Occasionally drivers would trade stops before leaving, due to area boundaries that caused extra miles for both.
I’m surprised Amazon doesn’t mandate the pictures. It’s a nice convenience for the customer (“Oh, I see, it’s tucked over behind the rocking chair”), but it also protects the company from liability (“Yes, we did in fact deliver it, and here’s the proof”). And big companies really like things that protect them from liability.
While you can’t optimize routes in the worst case in reasonable time, the question is whether you can optimize the routes you see in polynomial time. I worked on something that reduced to the bin packing problem (also NP-hard) in grad school, and while we could find small examples where fast heuristics failed, they were optimal in tens of thousands of manufactured cases.
There is also the question of how non-optimal the solution is. 1% of 50%?
We published in IEEE Transactions on Computers.
Right, like I said, you can find solutions that are probably close to optimal. You just get problems arising if you end up with a pointy-haired boss who insists that you be absolutely certain of finding the absolute best path.
The press gets this wrong all the time. As do the introductions to lots of papers I’ve reviewed that make it sound like a better solution is essential to problems where good enough solutions exist already. One of my pet peeves, papers who say the entire industry will crash if our 1% better solution isn’t adopted.