How an Online Order Arrives the Next Day

TL;DR. Tapping "buy now" at 9 p.m. feels like one simple action. It's actually the trigger for a coordination system running in milliseconds: an order management system reserves one specific physical unit inside one specific warehouse before another shopper's identical click can claim it, then routes that reservation through a fulfillment center where robots carry shelves to human pickers, a sortation center that groups the package with thousands of others heading the same direction, a network of overnight trucks, and finally the shortest, slowest, most expensive leg of the whole trip: the last few miles to a door. None of this was inevitable. Amazon's decision in 2005 to sell unlimited two-day shipping for $79 a year forced the entire retail industry to rebuild where warehouses sit, how densely they're packed, and how many robots work inside them.

Key takeaways

  • The instant an order is placed, an order management system reserves a specific unit of inventory at a specific facility, a "soft hold" that expires in minutes if payment fails, which is the only reason two shoppers can't both buy the last one.
  • Modern fulfillment centers mostly don't send workers to walk aisles. Amazon's warehouse robots, descended from a company called Kiva Systems bought in 2012, carry entire shelving pods to stationary pickers instead.
  • The last few miles to a customer's door account for roughly 53 percent of total shipping cost, more than warehousing, middle-mile trucking, and inventory management combined, which is why gig-economy drivers and postal partnerships exist at all.
  • Sears built a mail-order empire in the 1880s and 1890s on the back of railroads and, later, a 1913 parcel post law; Amazon Prime's 2005 launch of paid two-day shipping did the same thing to warehouse geography that parcel post did to farmhouses.
  • A shared vocabulary, GS1 barcodes on shipping labels, an EDI standard for carrier status updates, and USPS's ZIP+4 address system, lets a single package's status be tracked consistently as it crosses company boundaries.
  • The system runs with very little slack: the 2013 holiday season, when UPS and FedEx both missed Christmas delivery promises after online sales spiked faster than either network could absorb, is the clearest public example of what happens when a tightly optimized network meets a volume surge.

The moment nobody thinks about

It's 9 p.m., you're on your phone in bed, and you need a phone charger, a child's birthday present, or a replacement part for something that broke that afternoon. You tap "buy now." You don't check where it ships from. You don't wonder whether it exists in a warehouse forty miles away or four hundred. By early afternoon the next day, a box is sitting on the doorstep, and the whole transaction has cost you less thought than deciding what to have for lunch.

That gap, between an instantaneous tap and a physical object crossing hundreds of miles in under eighteen hours, is one of the more astonishing pieces of everyday infrastructure most people never examine. It didn't get built because someone found a faster truck. It got built because a retailer restructured its entire warehouse network around a shipping promise, and then every competitor had to follow.

The immediate mechanism: what happens in the first second

Before any box moves, software has to answer a question that sounds trivial and isn't: which physical copy of this item, sitting in which building, is now yours and not available to anyone else?

That's the job of an order management system (OMS), the software layer that sits between a retailer's website and its warehouses. The instant an order is placed, the OMS checks inventory across the retailer's network of facilities and, if stock exists at a nearby one, creates a temporary reservation, often called a soft hold, on a specific unit. That hold keeps the item out of the available pool while payment is processed, and if checkout fails or the customer abandons the cart, the system releases the hold back to inventory after a short timeout, typically five to fifteen minutes. If payment succeeds, the soft hold converts into a permanent deduction and the order is handed off to a specific fulfillment center for picking.

This reservation step is what prevents the oldest problem in retail: two customers buying the last unit of something at the same moment. Online, it also does a second job invisibly: it usually routes the order to whichever in-stock facility is closest to the delivery address, because shipping distance is the single biggest lever an OMS has over both delivery speed and delivery cost. A single popular item might sit in dozens of fulfillment centers around the country specifically so this routing decision has options to choose from.

Don't be confused: "in stock" is not a shelf count, it's a live balance. The number you see on a product page is total physical inventory minus every unit currently on soft hold for someone else's in-progress checkout. That's why an item can flip from available to "temporarily out of stock" between two page loads with nobody actually buying the last one twice, and why flash sales on high-demand items can show inventory swinging up and down in real time as other shoppers' holds expire and get released back into the pool.

The complete journey: inside a fulfillment center

Once an order is assigned to a facility, it enters a process the industry shorthands as pick, pack, and ship, though a modern fulfillment center (the large warehouse that stores inventory and assembles outgoing orders, as distinct from a store) actually runs four stages.

Stow happens on the inbound side, before any customer order exists. Workers unload delivery trucks and place items, scanned one at a time, into bins on tall, wheeled shelving units. Counterintuitively, items aren't grouped by category the way a store organizes shelves. A phone case might sit next to a bag of dog food and a paperback novel, because in a system where every location is tracked by computer rather than by a human's mental map, random storage lets whichever shelf is nearest be used next, which matters enormously once robots start carrying those shelves around.

Pick is where the robots most people have heard about, without knowing the name, actually work. Amazon acquired a robotics company called Kiva Systems in 2012 and rebuilt its warehouses around Kiva's core idea, now called Amazon Robotics: instead of sending a human employee walking down aisles to retrieve items (the old model, still used in smaller or older warehouses), a fleet of squat, wheeled robots drives underneath a shelving unit called a pod, lifts it a few inches off the floor with a corkscrew motion, and carries the entire pod to a stationary human picker. The robots navigate by reading a grid of barcode stickers on the warehouse floor and carry sensors to avoid colliding with each other. The picker never leaves their station; the shelf comes to them, the system tells them exactly which bin and item to grab, they scan it, and the pod gets carried back into storage while the next one arrives. Amazon has deployed more than a million of these robots across its network, and the company has reported that a later software rewrite of the picking algorithm cut the distance each robot travels per item picked by 62 percent, letting fewer robots do the same work. This is a genuinely different design philosophy from a traditional distribution center, where products stay put and it's the people who walk miles per shift.

Pack comes next: an item arrives at a packing station, gets scanned again to confirm it matches the order, and software recommends a box or mailer size based on the item's dimensions, with a machine cutting exactly enough tape. Ship finishes the on-site process through a step Amazon calls SLAM (scan, label, apply, manifest): a shipping label goes on the package, a final scanner reads it and assigns the package to a specific chute based on its destination, and the package slides down that chute directly into a waiting truck trailer sorted by where it's headed next.

Sortation, linehaul, and the last mile

A single fulfillment center doesn't ship straight to every house in its region; that would mean sending half-empty trucks in every direction. Instead packages typically move from the fulfillment center to a sortation center, a separate facility whose only job is to take in packages from multiple fulfillment centers, re-sort them by destination ZIP code or region, and consolidate them onto fuller trucks headed to a specific area. From there, packages continue to a delivery station, the smaller, local facility that actually loads them onto the vehicles that reach individual doors.

Moving packages between fulfillment centers, sortation centers, and delivery stations, sometimes called the middle mile, runs on a scheduled network of trucks (and, for longer distances, cargo planes) called linehaul. Amazon's internal logistics arm describes routing decisions across this network as a genuinely enormous optimization problem, one so large that a routing team has put the number of possible combinations across its trucking network at greater than 10 to the 88th power, solved continuously by software rather than by any human dispatcher.

Then comes the last mile: the final leg from a delivery station or local post office to an individual address. It is, by a wide margin, the most expensive and least efficient part of the entire journey, and the economics explain why. A linehaul truck can carry tens of thousands of packages at once and burn fuel efficiently per package because it's making one long, mostly highway trip. A last-mile delivery van makes one stop per package (sometimes one stop per several packages, if neighbors both ordered something), idles in traffic, circles back for missed deliveries, and pays a driver by the hour or route regardless of how many parcels get dropped off. Industry cost tracking puts the last mile at roughly 53 percent of total shipping cost as of 2023, up from about 41 percent five years earlier, with labor alone accounting for roughly half of that last-mile expense and fuel adding another 10 to 25 percent.

Retailers have answered the last-mile problem with three overlapping approaches. The first is a dedicated van fleet: companies like UPS and FedEx run their own branded trucks, and Amazon built a parallel network of small, separately owned local businesses called Delivery Service Partners whose drivers wear Amazon uniforms and drive Amazon-branded vans without being Amazon employees. The second is a gig-economy layer: Amazon Flex lets people sign up as independent contractors, use their own car, and pick delivery "blocks" of a few hours through a smartphone app, with pay reported in the range of $18 to $25 an hour before gas and vehicle wear, and a hard cap of 116 hours in any 30-day period. The third is handing the very hardest, most sparsely populated routes to an organization that already reaches every address in the country by law: Amazon and the U.S. Postal Service have partnered on last-mile delivery for roughly three decades, with USPS carrying Amazon packages the final distance in rural areas where running a private delivery van would be uneconomical for the volume involved. That relationship has become a genuine point of financial strain in the mid-2020s, with a renegotiated agreement cutting Amazon's USPS package volume by about a fifth even as USPS opens its delivery network to other shippers through a public bidding process.

Don't be confused: the van outside your house probably isn't driven by an Amazon, UPS, or FedEx employee. UPS and FedEx do employ many of their own drivers directly, but a large share of Amazon's branded delivery vans are operated by independently owned Delivery Service Partner companies working under contract, and a separate slice of deliveries, especially during demand spikes, comes from Amazon Flex contractors using their own personal vehicles. Three different employment relationships, one identical-looking box on the porch.

Who keeps it running

A single overnight delivery passes through more distinct job categories than almost any other everyday transaction. Warehouse pickers and packers move through pick and pack stations at a steady, monitored pace for an entire shift. Robotics and mechatronics technicians keep the pod-carrying robots, conveyor belts, and PLC-controlled sortation machinery running, performing scheduled preventive maintenance and troubleshooting electrical and mechanical failures on equipment whose downtime can back up an entire building within minutes. Linehaul truck drivers run the scheduled overnight routes between fulfillment centers, sortation centers, and delivery stations. Delivery Service Partner drivers and Amazon Flex contractors, along with UPS, FedEx, and USPS carriers, make the final stop at the door. Route planners and logistics engineers build and continuously retune the software that decides which truck carries which package on which road. And customer service staff field the calls and chats that start the moment any of the above goes wrong: a package marked delivered that never arrived, a delayed shipment, a damaged box.

Where this came from

Buying something you can't see and having it delivered to your door is not an internet-era idea. Richard Sears and Alvah Roebuck founded their mail-order company in 1886, and by the mid-1890s their catalog, which grew from a thin pamphlet of watches into a 532-page volume by 1895 generating $750,000 in sales, was reaching customers by way of the country's rapidly expanding rail network. The catalog earned an enduring nickname, the "Wish Book," and became a genuine cultural fixture in rural households that had no other practical way to buy manufactured goods.

Two federal changes made that business truly national. Rural Free Delivery, introduced in 1896, brought mail carriers directly to farmhouses that previously had to send someone into town to collect mail at all. And the Parcel Post Act of 1913 let the Postal Service carry heavier packages (up to 11 pounds at first, eventually 70) at affordable rates; before that, anything over four pounds needed an expensive private express company. Sears and Montgomery Ward had both lobbied hard for parcel post, and it worked exactly as they hoped: Sears reported its order volume grew roughly fivefold in the law's first year, kicking off what's remembered as the golden age of mail order, built on a postal policy change and a rail network, decades before anyone owned a computer.

The internet version of the same idea arrived quickly once the technical pieces existed. Netscape introduced SSL encryption in 1994, making it practical to type a credit card number into a website without it traveling in plain text, and within a year both Amazon (launched in 1995 as an online bookstore, with Jeff Bezos famously packing the company's first sold book from his Seattle garage) and eBay (started the same year as a hobbyist auction site called AuctionWeb) were online. The industry grew steadily through the following decade but still largely assumed ordering online meant waiting a week.

That assumption broke in 2005. Amazon announced Amazon Prime on February 2 and launched it on April 17, offering unlimited free two-day shipping, across the contiguous United States, on more than a million eligible items, for a flat annual fee of $79. There was no minimum order size and no need to bundle items together to make free shipping worthwhile, which had been the standard workaround retailers used to avoid shipping small orders at a loss. Meeting that promise at scale meant Amazon couldn't simply keep operating a handful of enormous, remote warehouses built for cheap land and bulk long-haul shipping. It had to build a denser, more geographically distributed network of fulfillment centers close enough to population centers that a two-day promise was achievable on ordinary ground shipping rather than expensive air freight. Every large retailer that wanted to compete on delivery speed eventually had to make the same investment, and the "two-day" baseline of 2005 has kept compressing since: Amazon later added guaranteed one-day and same-day options, built smaller "Sub-Same-Day" fulfillment centers of around 150,000 square feet specifically positioned to reach nearby customers within hours rather than days, and in the 2020s shifted from a small number of national mega-warehouses toward a regional network with facilities in thousands of smaller cities and towns.

Standards that make it all interoperable

None of this works if a package's barcode, tracking number, or address means something different to each company that touches it. The shipping label on a box, and the pallet or shipment it might travel with, follows a family of standards maintained by GS1, the same organization behind the retail barcodes covered elsewhere in this book. A shipment or pallet gets a Serial Shipping Container Code (SSCC), an 18-digit number that functions as a license plate for that specific physical unit as it moves through a supply chain, encoded in a barcode format (commonly GS1-128) that any GS1-compliant scanner, regardless of which company owns it, can read and log.

Once a package is moving, its status, picked up, in transit, delayed, delivered, needs to update consistently no matter which carrier is handling it or which company's app a customer is checking. That interoperability mostly runs on EDI 214, an electronic data standard (part of the broader ANSI X12 family used across freight) that carriers use to transmit shipment status events using a shared set of codes: a scan at a terminal, a delay notice, a delivery confirmation. It's the reason a UPS truck, a FedEx warehouse, and a regional last-mile courier can all feed status updates into the same tracking timeline a customer sees on one screen, even though the three companies share no other systems.

None of this sorting works, though, without a standardized address to sort toward, which is where the U.S. Postal Service's own infrastructure comes in. ZIP+4, the four-digit extension added to a standard five-digit ZIP code, narrows a delivery location down to a specific block face, building, or floor, precise enough for automated sorting machinery to route mail without a human ever reading the address. Getting messy, human-typed addresses into that exact format is the job of CASS (the Coding Accuracy Support System), a USPS certification program for the address-cleanup software retailers and shippers use to convert "Street" to "ST," fix misspellings against USPS's official address database, and append the correct ZIP+4. Software has to hit at least 98.5 percent accuracy to earn CASS certification, and shippers who want automation-rate postal pricing must run their address lists through certified software first.

Keeping it working

Reliability here is a maintenance calendar, not a fixed property of the machinery. Amazon Robotics technicians run scheduled preventive maintenance on the pod-carrying robots and the conveyor and sortation equipment that moves packages inside a facility, since a single jammed sorter or an offline robot lane can bottleneck an entire building's output within minutes, not hours. Sortation equipment, the optical scanners, cross-belt sorters, and barcode readers that route packages to the correct outbound chute, needs regular recalibration, particularly as packaging materials themselves change and a new plastic mailer or oddly shaped box confuses equipment tuned for yesterday's typical package. Vehicle fleets, whether UPS-owned trucks, Delivery Service Partner vans, or the growing number of electric delivery vehicles Amazon has introduced, run on standard commercial maintenance schedules for brakes, tires, and increasingly battery and charging systems. And the routing software itself is never finished: logistics teams continuously retune the algorithms that decide which truck gets which package and which driver gets which sequence of stops, because a route that was optimal last month can become inefficient as delivery density, traffic patterns, or a new sortation center changes the underlying map.

When it breaks

The clearest, most public failure of this entire system happened in December 2013. UPS normally moved about 15.5 million packages a day and had planned for a 15 to 18 percent jump during the holiday peak; instead, Christmas Eve volume hit roughly 34 million packages. FedEx saw an even sharper swing, from an average day of about 10 million packages to 22 million on Christmas Eve alone, and handled some 275 million shipments in total between Thanksgiving and Christmas that year. Both companies pointed to a combination of winter storms and a surge in last-minute online ordering: Cyber Monday sales that year had jumped more than 20 percent over the prior year, and online sales in the final weekend before Christmas rose 37 percent year over year. Neither company disclosed exactly how many packages missed their promised delivery date, describing it only as a small percentage of overall volume, but a small percentage of tens of millions of packages is still an enormous number of disappointed families on Christmas morning, and it was reported widely enough to become the reference case for what a peak-season overload looks like.

The deeper lesson isn't really about weather or one bad holiday season. It's that a logistics network built for efficiency, trucks scheduled to run close to full, routes optimized to the mile, staffing planned against a forecast, has very little slack by design, because slack is expensive. That's a reasonable trade-off on an ordinary Tuesday and a real liability the moment actual demand diverges sharply from the forecast a network was built around, whether the cause is a holiday surge, a regional weather event, or a global supply shock. The system recovers, but recovery looks like exactly what happened in 2013: public apologies, refunded shipping fees, and packages arriving days after they were promised.

The scale of it

The United States shipped an estimated 22.4 billion parcels in 2024, a 3.4 percent increase over the year before, which works out to roughly 61 million packages moving through the system on an average day. USPS alone handled about 6.9 billion of those parcels for the year, followed by Amazon's own logistics operation at about 6.3 billion, UPS at 4.7 billion, and FedEx at 3.7 billion. On a single average day, USPS ships around 23.9 million packages, which breaks down to nearly 1 million an hour, or about 277 every second.

Behind that volume sits a physical network built specifically to hit fast delivery windows: Amazon operates more than 175 fulfillment centers in the United States, totaling upward of 200 million square feet of warehouse space, plus several hundred additional sortation centers and delivery stations that bring its total logistics footprint past 600 buildings, with California, Texas, and New Jersey hosting the largest concentrations. The company announced roughly $15 billion in new investment in 2025 for about 80 additional sites, many of them the smaller, robotics-heavy same-day facilities described earlier rather than the older mega-warehouse model.

Delivery speed itself has compressed dramatically since Prime's 2005 baseline of two days. By 2023, Amazon reported delivering more than 1.8 billion items to U.S. Prime members on a same-day or next-day basis, nearly four times the volume it had reached by the same point in 2019. By 2025, that figure had grown to more than 8 billion same- or next-day items in the United States alone (over 13 billion globally), with half of those deliveries, roughly 4 billion items, being groceries and everyday essentials rather than the electronics and media that dominated Prime's original 2005 catalog.

Trade-offs and what's next

Speed and density come with costs that don't show up on a receipt. Workplace safety is the sharpest of them: a 2021 analysis of federal injury data found that while Amazon employed roughly a third of U.S. warehouse workers, it accounted for about half of all recorded injuries in the warehouse industry, with a serious-injury rate of 6.8 per 100 workers against an industry average of 3.3. Counterintuitively, the facilities with the most robotics saw the sharpest increases: serious injury rates at Amazon's robot-equipped "sortable" facilities rose 20 percent between 2020 and 2021, ending up 28 percent higher than at otherwise similar facilities without robots, a pattern generally attributed to the faster pace robots enable rather than the robots themselves being dangerous. A 2024 U.S. Senate committee investigation concluded Amazon's own injury statistics likely understated the real rate, citing practices around how injuries get recorded and referred for outside care. Gig-economy delivery work raises a related but distinct concern: because Amazon Flex drivers are independent contractors rather than employees, they cover their own gas, vehicle wear, and insurance, and carry none of the benefits or job security that come with direct employment, even while working under an app that assigns and times their routes.

Automation is also reshaping what jobs exist at all, though not in a simple direction: Amazon frames robotics as shifting workers into higher-skill maintenance and technical roles rather than eliminating jobs outright, and robotics and mechatronics technician positions are themselves a fast-growing job category, but the net effect on total warehouse headcount as a company builds more automated facilities remains a genuinely disputed question between the company and labor researchers.

At the very edge of the delivery network, drone delivery has moved from demo to small-scale reality. Amazon's Prime Air had completed roughly 16,000 deliveries by February 2026 using autonomous MK30 drones across a handful of markets in Texas, Arizona, Florida, and Kansas, with new operations launching in Louisiana, the Chicago suburbs, and Nebraska through the middle of the year, typically completing an order within 60 minutes of purchase. The company has stated a goal of reaching 30 million customers by the end of the year and 500 million drone-delivered packages annually by decade's end, but set against the roughly 22 billion ground parcels already moving through the country every year, drone delivery remains, for now, a rounding error next to the truck-and-van network doing nearly all of the actual work.

The environmental side of fast shipping is the least visible trade-off and arguably the most consequential. E-commerce as a whole generates close to five times more packaging waste per item than the equivalent goods sold in physical stores, according to a 2024 United Nations digital economy report, averaging more than two separate pieces of packaging for a single product. Fast shipping compounds that: meeting one- or two-day delivery windows tends to mean sending out delivery vehicles before they're full and running the same neighborhood's route more than once in a day rather than consolidating everyone's packages into fewer, fuller trips, and multiple research estimates put the resulting emissions increase from rushed delivery well into the double digits compared with standard shipping timelines. A single household making one planned trip to a store will, in most cases, produce a smaller carbon footprint than that same household ordering the same items separately online with fast shipping selected each time, even though the online version feels, to the person clicking "buy now," like the more efficient choice.

Back to the doorstep

That box waiting on the step the next afternoon is the visible tail end of a system that reserved a specific unit of inventory in a specific building the instant you tapped a button, moved it past robots that never learned where anything is stored because a computer always knows, sorted it among thousands of others heading the same direction, trucked it overnight, and handed it to a driver making the least efficient, most expensive leg of the entire trip on purpose, because someone has to. It traces back further than the internet, through parcel post and a Sears catalog nicknamed the Wish Book, to a specific $79-a-year bet Amazon made in 2005 that restructured where warehouses sit across an entire industry. The order looks instant. The building underneath it took two decades, and counting, to build.

The leap: what it replaced, and the work behind it

Picture a farm family in the 1890s who need a new stove. Their choices are two. They can drive a wagon half a day into town to a general store that stocks one model at a price it sets by negotiation, often on credit at steep interest, because the store has a captive customer and slow turnover to cover. Or they can order from a catalog: pick up the Wish Book on one town trip, fill out the form and buy a postal money order on the next trip, mail it, and then wait. The goods came by rail to the nearest depot, and before Rural Free Delivery reached a given route (it only became a permanent nationwide service in 1902), someone still had to ride back into town to collect the package. Weeks could pass between wanting the thing and holding it. Even inside the warehouses of that era, and for most of the century after, nobody knew exactly what was on the shelves: before barcode scanning spread in the late 1970s, stock was counted by hand, roughly once a month, and orders were placed on a manager's rough guess.

The jump from weeks to overnight is larger than it sounds, and it did not happen by finding a faster truck. It happened by rebuilding where inventory sits and adding an enormous, permanent layer of human work underneath the software. The United States shipped an estimated 22.4 billion parcels in 2024, roughly 61 million every single day, and each one is stowed, picked, packed, scanned, sorted, trucked, and carried to a door by people whose shifts are timed against that flow. The last few miles alone, the part that puts a box on your specific step, runs about 53 percent of the whole shipping cost, most of it labor, because a van making one stop per house is the opposite of efficient and always will be.

You feel the leap on an ordinary night. A part breaks at four in the afternoon, you order the replacement at nine, and it is on the porch before lunch the next day, and the whole thing cost you less deliberation than choosing a sandwich. The morning it fails, a winter storm on top of a holiday surge, the kind of collapse UPS and FedEx both had in December 2013 when Christmas Eve volume roughly doubled their normal load, you get a taste of the older world back: the promised box does not come, the gift is late, and there is nothing to do but wait, the way everyone once waited by default. That the wait is now the exception and not the rule is the work of a picker, a mechanic keeping a jammed sorter running, a driver circling back for a missed address, and a routing engineer retuning a map, all of them invisible from the doorstep where the box appears.

Real-world examples and recent developments

Amazon is not the only company reshaping how packages move from warehouse to door; competitors, contractors, and regulators are all pushing on the same problem from different angles.

  • Ocado Group (Hive system, Erith, UK): a British online grocery company that runs its own robotic warehouses instead of licensing Amazon's approach. More than 2,000 robots move across a giant grid at its Erith facility near London, each capable of picking a full order in about five minutes, roughly five times faster than a human picker doing the same job. Ocado Group, Grocery Warehouse Automation
  • GXO Logistics (spun off from XPO Logistics on August 2, 2021): the world's largest pure-play contract logistics company, running warehousing and fulfillment for hundreds of retailers across more than 970 facilities in 27-plus countries, entirely outside Amazon's own network. Wikipedia, GXO Logistics
  • Walmart and Symbotic (deal completed January 28, 2025): Walmart agreed to invest $520 million so robotics company Symbotic could take over and expand Walmart's automated distribution centers, part of a plan to build 400 AI-driven "Accelerated Pickup and Delivery" centers. Symbotic, Symbotic to Acquire Walmart's Advanced Systems and Robotics Business
  • Aurora Innovation (commercial service launched May 2025): a self-driving trucking company that began running fully driverless heavy trucks between Dallas and Houston on public highways, the first company to do so commercially, aiming at the linehaul leg of the journey described earlier in this chapter. TechCrunch, Aurora launches commercial self-driving truck service in Texas

Recent developments

  • Amazon's Vulcan robot (unveiled May 7, 2025): Amazon's first warehouse robot with a genuine sense of touch, using force-sensing fingers to feel whether it has a solid grip on an item rather than relying on cameras alone. It can already handle roughly 75 percent of items at the warehouses where it's deployed, starting in Spokane, Washington, and Hamburg, Germany. About Amazon, Introducing Vulcan
  • The FAA's proposed Part 108 rule (proposed August 7, 2025): the first major rewrite of US commercial drone regulation since 2016, meant to let package-delivery drones fly routinely beyond an operator's direct line of sight instead of needing a special case-by-case waiver. A final rule was expected sometime in 2026 after a public comment period that drew more than 3,000 responses. DLA Piper, FAA's proposed Part 108 BVLOS Rule

Glossary

Fulfillment center. A large warehouse that stores retail inventory and assembles individual customer orders for shipment, as distinct from a storefront.

Order management system (OMS). Software that tracks inventory across a retailer's warehouses in real time and decides which facility fulfills each order, reserving specific units so they can't be sold twice.

Goods-to-person. A warehouse design, pioneered by Kiva Systems and now called Amazon Robotics, in which robots carry entire shelving units to stationary human workers instead of workers walking to fixed shelves.

Sortation center. A facility that receives packages from multiple fulfillment centers, re-sorts them by destination, and consolidates them onto fuller trucks headed to a specific region.

Delivery station. The local facility where packages are loaded onto the vehicles that make final deliveries to individual addresses.

Middle mile / linehaul. The scheduled trucking (and sometimes air) network that moves packages between fulfillment centers, sortation centers, and delivery stations.

Last mile. The final delivery leg from a local facility to an individual address, disproportionately the most expensive and least efficient part of a package's journey.

Delivery Service Partner (DSP). An independently owned small business that operates Amazon-branded delivery vans under contract; its drivers wear Amazon uniforms but aren't Amazon employees.

Amazon Flex. A gig-economy program in which independent contractors use their own vehicles to deliver packages in exchange for pay by the hour or delivery block, arranged through a smartphone app.

SLAM (scan, label, apply, manifest). The final packing-line step where a shipping label is applied and scanned, assigning the package to the correct outbound truck chute.

Serial Shipping Container Code (SSCC). An 18-digit GS1 standard number that uniquely identifies a specific shipment or pallet as it moves through a supply chain, encoded in a scannable barcode.

EDI 214. An electronic data standard that carriers use to transmit shipment status updates, pickup, transit, delay, delivery, using shared codes that let different companies' systems feed one consistent tracking timeline.

ZIP+4 / CASS. ZIP+4 is the four-digit extension that narrows a U.S. postal code down to a specific block or building; CASS is the USPS certification program for the address-cleanup software that appends it accurately.

Sources and notes

Open questions

  • Reported figures for how much faster shipping increases per-package emissions vary widely by study methodology, from roughly 10 to 12 percent in some estimates to much higher in others; treat "rushed delivery raises emissions" as well established and any single percentage as illustrative rather than exact.
  • The long-run employment effect of warehouse automation, whether robotics is primarily displacing warehouse jobs or shifting them into higher-skill maintenance roles, remains genuinely disputed between company statements and independent labor researchers, and the picture may look different in a few years as robotics-heavy facilities make up a larger share of the network.
  • The Amazon-USPS last-mile relationship was actively being renegotiated at the time of writing, so the specific volume-sharing terms described here should be treated as a 2026 snapshot rather than a fixed arrangement.

This closes the "Getting people and packages around" part of the book. Next, a different kind of package, one made of numbers instead of cardboard, moves through its own hidden network every time a card gets swiped. How a payment travels through the banking system 👉