How Supermarkets Keep Thousands of Products Available

TL;DR. Walking down a supermarket aisle and finding the same brand of crackers in the same spot it occupied last week feels unremarkable, which is exactly the point: a store carrying 30,000 to 50,000 distinct products runs a forecasting, ordering, and delivery system precise enough that most gaps get filled before a shopper notices them. The trigger for almost all of it is a single beep at the register, a barcode scan that charges a customer and quietly tells a computer that one fewer unit of that item now exists on the shelf. That one small standard, the Universal Product Code, first scanned on a pack of Wrigley's gum in a Troy, Ohio, supermarket on June 26, 1974, replaced manual shelf-checking and paper order forms with a system spanning forecasting software, electronic purchase orders, regional distribution centers, and, for some products, delivery trucks that belong to the manufacturer rather than the store. It also has real, documented failure points: a pandemic, a ransomware attack, or a bad software update can each empty shelves in ways a shopper actually notices.

Key takeaways

  • Every checkout scan is a data event, not just a payment step. It decrements inventory in real time and feeds the demand forecasts that decide what gets reordered.
  • Two different supply chains run into the same store at once: centrally distributed goods that pass through a retailer's own warehouse, and direct-store-delivery (DSD) goods like bread, chips, and soda that arrive on a manufacturer's own truck and bypass that warehouse entirely.
  • The Universal Product Code (UPC) was first scanned commercially on June 26, 1974, on a pack of Wrigley's Juicy Fruit gum, at a Marsh Supermarket in Troy, Ohio, twenty-two years after the underlying barcode patent was filed.
  • A nonprofit standards body called GS1 assigns the manufacturer codes inside every barcode worldwide, and a separate standard, EDI, carries the purchase orders and shipping notices between retailers and suppliers, mostly invisibly.
  • Retailers actively track "shrinkage," inventory lost to theft, damage, spoilage, and paperwork errors, as an ongoing cost; the industry-wide figure for 2022 was about $112 billion, concentrated most heavily in grocery and pharmacy.
  • Real disruptions, 2020 pandemic panic buying, a 2025 ransomware attack on a major grocery distributor, a 2024 global software outage that knocked out registers, show how much coordination has to work correctly for a shelf to simply look normal.

The moment nobody thinks about

Walk into almost any mid-sized supermarket in North America and it will carry somewhere between 30,000 and 50,000 distinct products: a dozen kinds of mustard, a wall of breakfast cereal, three brands of the same milk in three fat percentages, produce that has to look fresh today and still look fresh in four days. Nearly all of it is where a regular shopper expects it to be. When something does run out, it's usually back within a day, often overnight, and most shoppers never learn why it was gone or how it came back. That quiet reliability is not a natural state. It's the visible tip of a forecasting and logistics system built, tested, and continuously repaired by people whose job titles never appear on a receipt.

The immediate mechanism: what a barcode scan actually does

The visible part of this system is almost too fast to notice: a cashier, or a shopper at self-checkout, passes an item over a scanner, a light and a beep register, and a price appears. What actually happened is that a laser or image sensor read a pattern of black and white bars, decoded it into a string of digits, and sent that string to the store's point-of-sale (POS) system, the combined hardware and software that runs a checkout lane. The POS system looks the number up, pulls the current price and any active promotion, completes the transaction, and, in the same instant, sends a message to the store's inventory database: one less unit of this product exists on this shelf.

The number being scanned is a UPC, a Universal Product Code, almost always printed as a UPC-A symbol: twelve digits under a pattern of bars of varying width. The first digit broadly identifies the product category. The next five identify the manufacturer, a number assigned to that company by GS1, the nonprofit standards body described later in this chapter. The following five identify the specific product, chosen by the manufacturer itself. The last digit is a check digit, calculated from the other eleven, whose only job is to catch a misprint or scanning error: if the math doesn't work out, the scanner rejects the read instead of ringing up the wrong item. None of those twelve digits carry a price; the barcode is only ever a lookup key, and the price, description, and current promotion all live in the store's own database, which is why a store can run a sale without reprinting a single label.

That single scan is also the seed event for almost everything in the rest of this chapter. It updates the store's on-hand inventory count, contributes a data point to that item's sales history, and, once enough data points accumulate, can trigger an automatic reorder. A cash register looks like the end of a shopping trip. For the supply chain behind it, it's closer to a starting gun.

Don't be confused: a barcode, a UPC, and a SKU are three different things. The barcode is the printed pattern of bars, just a way of encoding a number so a machine can read it quickly. The UPC (or, in GS1's broader international terminology, a GTIN, Global Trade Item Number) is the number that barcode encodes, assigned once to a product by its manufacturer and used by every retailer that carries it. A SKU (stock keeping unit) is a retailer's own internal inventory code, which sometimes matches the manufacturer's UPC and sometimes doesn't, especially for store-brand items or products sold by weight. When someone asks how many products a store carries, they're almost always counting SKUs, the store's own internal list, not manufacturer UPCs.

The complete journey: from a scanned gap to a restocked shelf

A single scan is nearly meaningless on its own. What matters is the pattern across thousands of scans a day, at every store in a chain, feeding a demand forecast: a prediction of how much of a given product a specific store will sell over a coming stretch of days. Forecasting software weighs several inputs at once: the item's own sales history at that store, seasonal patterns (charcoal sells differently in July than in January), and any promotion scheduled for that item, since a discount can multiply expected demand several times over for the week it runs. Grocery forecasting is unusually hard compared with other retail categories because so much of the assortment is perishable: overordering fresh strawberries doesn't sit safely in a warehouse, it goes in a dumpster a few days later.

Once a forecast says a store is likely to need more of an item than it has on hand or already scheduled to arrive, an automatic replenishment system can generate a reorder without a human placing it. These systems watch current inventory against a reorder point, a threshold quantity that accounts for how long delivery will take and how much of a safety cushion the retailer wants to hold, and generate a purchase order once stock crosses it. For a large chain, this runs continuously across tens of thousands of SKUs in every store at once, a repetitive, high-volume calculation that was simply impossible before computerized inventory systems existed, back when replenishment meant a store employee walking the aisles with a clipboard, guessing at what looked low, and mailing in a paper order.

Where that reorder actually goes next splits into two very different pipelines. Centrally distributed goods, most shelf-stable packaged groceries, frozen food, and household products, are ordered from the retailer's own distribution center (DC): a large regional warehouse, one of several a major chain operates, that receives full truckloads from manufacturers, stores pallets, and breaks them down into smaller, store-specific loads sent out on the retailer's own trucks, typically several times a week. This is where a chain gets most of its buying leverage: it negotiates prices directly with manufacturers for pallet-scale volumes and decides how much buffer stock to hold in its own warehouses before a shortage at any single store becomes a real problem.

The second pipeline, direct-store-delivery (DSD), skips the retailer's warehouse entirely. For categories with very short shelf life, very high turnover, or brand owners who want tight control over how their product is displayed, bread, salty snacks, soft drinks, beer, and much of dairy are the classic examples, the manufacturer or its regional distributor delivers straight to the store on its own branded truck, driven by a route driver who also stocks the shelf and manages the display personally. A bag of a major snack brand's chips typically never touches the supermarket's own distribution network at all; the snack company's driver brought it in that morning and arranged the shelf himself. This gives the manufacturer more control over freshness and shelf placement, at the cost of the retailer having less centralized visibility over that portion of its own shelves.

Underneath both pipelines sits a genuine trade-off in how much inventory to hold at all. A just-in-time (JIT) approach keeps deliveries small and frequent, matched closely to what a store is expected to sell, minimizing wasted capital, warehouse space, and spoilage, but leaves little slack: a late truck, a supplier shortfall, or an unforecast demand spike can empty a shelf within hours. Buffer stock (safety stock) is the deliberate opposite, holding extra inventory specifically to absorb that kind of surprise; fresh categories run thin, fast-turning buffers because the product can't wait, while shelf-stable goods can afford a steadier cushion. Retailers increasingly manage this as one optimization problem across their whole network, called multi-echelon inventory optimization, rather than setting a fixed buffer at each location independently, since a shortage anywhere can be absorbed more cheaply by moving inventory around than by holding extra of everything everywhere.

None of this reaches the shelf on its own. Once a truck arrives at a store, whether from the chain's own distribution center or from a DSD driver, someone still has to unload it, check it against the delivery paperwork, and carry it to the right aisle, and much of that work happens overnight or in the very early morning, precisely so a store looks fully stocked when it opens. An overnight stocking crew receives freight, verifies quantities against the shipment's manifest, sorts it by department, and physically restocks shelves and displays before the first customers arrive. It's the most physically direct link in the whole chain: a forecast, an order, and a truck all converge on one person moving cases from a pallet onto a shelf, usually while the store is empty and dark.

Don't be confused: not everything on a shelf came from the same place. Two boxes sitting next to each other in the same aisle can have arrived through entirely different systems. The cereal likely moved through the chain's regional distribution center after being ordered by an automatic replenishment system. The soda can beside it may have been carried in that same morning by the beverage company's own driver, who also decided how to arrange the display. A store's supply chain isn't one pipeline; it's several running in parallel, assigned by shelf life and negotiating economics.

Who keeps it running

A shelf staying stocked touches an unusually long chain of job titles, most of them well upstream of the store itself. Category managers decide what a chain stocks in the first place: which brands, sizes, and new products earn a spot on a planogram (the diagram specifying exactly where each item sits), balancing sales data, supplier relationships, and shelf space that is always, by definition, finite. Buyers handle the commercial side underneath that strategy, negotiating price, volume commitments, and delivery terms with suppliers. Supply chain and logistics planners run the demand forecasting models and set replenishment and safety-stock rules chainwide, translating raw sales data into concrete reorder quantities. Warehouse workers at distribution centers receive, store, and pick the pallets and cases those orders generate, increasingly alongside automated systems described later in this chapter. Truck drivers, both the retailer's own fleet and the separate DSD drivers employed by manufacturers, move product from warehouse or factory to store on schedules tight enough that a single missed route can show up as a gap on a shelf the next morning. Overnight stocking crews turn a truckload into a stocked aisle. And at the store level, inventory and loss-prevention managers handle the parts of this system that don't show up in any forecast: shrinkage, damaged goods, and the gap between what the computer says should be on the shelf and what's actually there.

Where this came from

The barcode itself is older than the supermarket scanner by more than two decades. In 1949, according to accounts collected by the Smithsonian and industry historians, engineering graduate student Norman Joseph Woodland, working with Bernard Silver, sketched an early version of the idea while sitting on Miami Beach, dragging his fingers through the sand and extending Morse code's dots and dashes into thin and thick lines. Woodland and Silver were granted US Patent 2,612,994 in 1952, covering both a linear striped pattern and a bullseye-shaped design, but the technology to read it cheaply and reliably at a checkout counter didn't exist yet. It took roughly two more decades of retail industry pressure, driven by grocery chains looking for a faster alternative to a cashier keying in every price by hand, before that changed.

In 1969, U.S. grocery and manufacturing trade groups formed the Ad Hoc Committee for a Uniform Grocery Product Code to settle on one shared numbering standard, since a barcode is only useful if every manufacturer and retailer agrees on the same format. In 1973 the committee selected the Universal Product Code, a design finalized by IBM engineer George Laurer, and the Uniform Code Council was founded the next year to administer it.

The first real commercial test came on the morning of June 26, 1974, at a Marsh Supermarket in Troy, Ohio. At 8:01 a.m., Clyde Dawson, the company's head of research and development, scanned a ten-pack of Wrigley's Juicy Fruit chewing gum across one of the store's new Spectra-Physics scanners, the first UPC-coded product ever scanned at a retail checkout. Dawson later said the choice wasn't random: gum's small package had been one of the more doubtful cases for whether a barcode could even be printed legibly at that size, so scanning it first was a deliberate demonstration that the format worked at its hardest scale, not just its easiest. That pack of gum, along with the receipt from the sale, is preserved today in the Smithsonian's National Museum of American History.

What followed over the next fifteen to twenty years was less a single event than a slow replacement of an entire way of running a store. Before scanning, prices were rung up by a cashier reading a sticker on every item, restocking decisions were made by a clerk walking the aisles and estimating what looked low, and reordering meant mailing a paper form to a supplier, with real lag between noticing a shortage and a truck arriving. Barcode scanning, paired with the computerized inventory systems that grew up around it through the 1980s and 1990s, collapsed both the price lookup and the "what do we need more of" question into something a computer could answer continuously, at a scale no clipboard-carrying employee ever could.

Standards and coordination

None of this works unless competing manufacturers, retailers, and distributors agree on the same numbering and messaging formats, which is the job of a small number of standards bodies operating almost entirely behind the scenes. GS1 is the nonprofit that administers the barcode standard globally: it assigns the manufacturer prefixes at the heart of every UPC and its international counterpart, the EAN (European Article Number), and maintains the technical specification for how those numbers are encoded into scannable symbols. GS1 in its current form dates to 2005, when the U.S.-based Uniform Code Council and the Brussels-based EAN International, which had separately overseen numbering in Europe since 1977, formally merged. Today GS1 operates through more than a hundred member organizations covering roughly 150 countries and serves over two million user companies, which is the practical reason a product barcoded in one country scans correctly at a till on another continent.

Barcodes only carry an identifying number; the purchase orders, invoices, and shipping paperwork that move between a retailer and its suppliers travel over a separate, older standard called EDI, electronic data interchange, the direct computer-to-computer exchange of standardized business documents that predates the modern web and is still the backbone of high-volume business-to-business trade. A grocery retailer's purchase order is commonly sent as an EDI 875 transaction, a grocery-specific variant of the more general EDI 850 used in other industries, and a supplier confirms an outbound shipment with an EDI 856, an advance ship notice (ASN) that tells the receiving distribution center or store exactly what's on a truck before it arrives. Industry estimates suggest automated ASNs can cut receiving time at a distribution center by up to 60 percent compared with checking a shipment against paper documents, saving a 250-store chain on the order of 65,000 labor hours a year.

A third, narrower standard governs the dates printed on perishable packaging. In the United States, outside of infant formula, date labeling on food is not actually required by federal law; where a manufacturer chooses to print one, U.S. Department of Agriculture guidance requires a clear phrase explaining what the date means, such as "Best if Used By," precisely because terms like "sell by" and "use by" mean different things and have historically confused shoppers into discarding food that was still safe to eat. Behind the label, GS1's own specification defines reserved fields called Application Identifiers for encoding a sell-by date, an expiration date, and a batch or lot number directly into a scannable code, which is what lets a manufacturer trace a specific contaminated batch during a recall instead of pulling an entire product line off every shelf.

Keeping it working

Barcodes, forecasts, and delivery schedules only stay trustworthy with constant, unglamorous upkeep. Stores and distribution centers run periodic cycle counts, physically counting a rotating subset of inventory on a regular schedule rather than shutting down for one giant annual count, to catch the gap between recorded and actual inventory. Labels have to stay legible and correctly linked to current pricing, since a barcode that scans to the wrong price or a faded label that won't scan at all both push work back onto a cashier. Behind the distribution network, a warehouse management system (WMS), the software controlling where every pallet is stored, picked, and routed to a truck, needs its own ongoing maintenance as a chain adds distribution centers or new automation.

A large, continuous piece of this maintenance is tracking shrinkage: inventory that vanishes from the books for reasons other than a legitimate sale. The National Retail Federation's 2023 National Retail Security Survey put the industry-wide average shrink rate at 1.6 percent of sales for fiscal year 2022, up from 1.4 percent the year before, roughly $112 billion in losses across the retail sector; grocery, along with pharmacy, was among the categories reporting shrink rates above 2 percent, the highest of any sector. Retailers attributed the losses to a mix of causes: 36 percent to external theft, 29 percent to employee theft, 27 percent to process failures and plain errors, with the remainder unexplained. None of that is a one-time write-off. It's a recurring cost that store-level inventory and loss-prevention teams actively work to reduce, because a percent or two of sales, at the scale of a national chain, is real money.

When it breaks

The clearest illustration of how much coordination sits behind a normal-looking shelf is what happens when a piece of it fails. In March 2020, as the COVID-19 pandemic set off widespread panic buying, researchers tracking retail scanner data found that median stock-out rates for staple grocery items rose by roughly 130 percent in the weeks immediately following March 15, with meat and poultry, frozen foods, baby formula, and carbonated beverages among the hardest-hit categories, following an initial wave of shortages in hand sanitizer, masks, and disinfectants, then toilet paper and shelf-stable staples. The forecasting systems described earlier in this chapter are built to predict normal seasonal and promotional demand variation; they were never designed for a nationwide, simultaneous spike in buying behavior driven by fear rather than ordinary consumption.

More recent disruptions have targeted the computer systems themselves. In June 2025, United Natural Foods (UNFI), one of North America's largest grocery distributors and a primary supplier to Whole Foods Market, suffered a ransomware attack that disabled internal ordering, transportation scheduling, and warehouse logistics; reporting at the time indicated that 18 of the company's 53 U.S. distribution hubs saw operational delays, with shipments halted and stores left waiting on orders that couldn't be processed. A similar case hit Canada directly: in November 2022, Sobeys, which with sister banners Safeway, IGA, and FreshCo operates roughly 1,500 stores under parent company Empire, was hit by Black Basta ransomware. With store computers and handheld order-entry scanners disabled, staff could not place restocking orders, and some stores ran out of items as a direct result; Empire later disclosed a financial hit in the tens of millions of dollars.

Even a failure with nothing to do with groceries specifically can stop a supermarket cold. On July 19, 2024, a faulty software update pushed by the cybersecurity vendor CrowdStrike crashed an estimated 8.5 million Windows-based systems worldwide, including point-of-sale terminals at retailers on several continents. Grocery chains including Waitrose in the UK and Tegut in Germany were reduced to cash-only sales or closed some locations outright, and supermarkets in New Zealand reported widespread payment failures, all because the software layer that runs a modern checkout lane, not the shelves or the trucks behind them, had stopped working. None of the food moved anywhere that day. The register simply couldn't ring it up.

These are dramatic, visible failures, but the ordinary cost of smaller, constant stock-outs is larger in aggregate than any single event. Industry estimates compiled from retail scanner and audit data put the combined cost of out-of-stocks to North American food retailers at close to 6 percent of total retail sales, and separate estimates suggest consumer packaged goods manufacturers lose over $100 billion in sales annually to items that simply weren't on the shelf when a shopper reached for them. A gap on a shelf isn't just one lost sale; a shopper who can't find what they came for often skips the rest of that trip's planned purchases entirely.

The scale of it

A single supermarket's SKU count has actually moved in both directions over the past several decades. Industry data collected by the Food Marketing Institute shows the average U.S. supermarket carried a little over 14,000 distinct items in 1980, a number that climbed as chains chased variety and reached a plateau around 51,000 SKUs by 2008, before falling back to roughly 33,000 by 2018 as chains trimmed slower-selling items and smaller-format stores grew in popularity. Format still drives most of the variation: a discount grocer commonly runs closer to 7,500 SKUs, a tightly curated chain like Trader Joe's around 4,000, and a traditional full-size supermarket somewhere between 30,000 and 50,000.

Multiply that by the size of a national chain and the numbers grow further still. Kroger, one of the largest traditional supermarket operators in the United States, reported operating 2,722 stores as of early 2024. Across the whole country, supermarket and grocery store sales reached roughly $880 billion in 2025, accounting for the large majority of all U.S. food and beverage retail spending, on top of a genuinely enormous number of daily replenishment decisions: tens of thousands of SKUs, at thousands of stores, each forecast, reordered, and delivered on its own schedule, every day of the year.

Trade-offs and what's next

The newest layer being added to this system is physical automation inside the distribution centers themselves. Companies building automated storage and retrieval systems, robotic case pickers, and autonomous mobile robots for warehouse floors, names like AutoStore, Exotec, Swisslog, and Geek+ appear repeatedly in industry coverage, are increasingly standard equipment in new grocery distribution centers, handling picking tasks that used to require a worker walking miles of warehouse aisle per shift. On the forecasting side, machine learning models are replacing older statistical methods because they can weigh far more variables at once, weather, local events, a competitor's promotion, without an analyst manually building each rule; industry estimates suggest AI-driven tools could add well over a hundred billion dollars in value to grocery retail globally by the end of the decade, largely through better-matched inventory.

A more fundamental possible shift is happening at the identification layer itself. RFID (radio-frequency identification) tags, small chips readable by radio signal without needing a direct line of sight the way a barcode scanner does, allow an entire pallet, case, or shelf of individual items to be inventoried automatically, in seconds, without anyone scanning a single unit by hand. Large non-grocery retailers, notably Walmart and Target, have required item-level RFID tagging from suppliers in some categories for years, and grocery-specific adoption is now underway too: Kroger began rolling out RFID-embedded labels in its bakery department in late 2024, part of a broader push toward better traceability for recalls and waste reduction in fresh categories, where a barcode's requirement of a direct scan is a genuine bottleneck. RFID has not replaced the UPC and likely won't for the bulk of packaged groceries anytime soon, since per-item tag cost still matters at grocery margins, but it's expanding fastest in exactly the fresh, perishable categories where item-level tracking has the most value.

Underneath all of this sits a tension that hasn't been resolved so much as managed: leaner, more efficient just-in-time supply chains are cheaper to run and waste less food, but the pandemic, and the ransomware attacks that followed it, made clear that the same leanness that saves money in a normal week can turn a single disrupted supplier or software system into an empty shelf within days. The industry's response has mostly been to get smarter about where to hold buffer, concentrating slack in the categories most exposed to disruption, rather than reversing course on efficiency altogether. Whether that's enough slack for the next disruption nobody has forecast yet is not something any of these systems can answer in advance.

Back to the shelf

The next time a scanner beeps at checkout on something as ordinary as a pack of gum, the transaction on the little screen is a direct, if far more automated, descendant of the one Clyde Dawson ran through a Spectra-Physics scanner in Troy, Ohio, just after eight in the morning on June 26, 1974. What's different now is everything behind that beep: a forecast built from that item's own sales history, an automatic reorder already queued once its shelf position runs low, a truck from either the chain's own distribution center or the manufacturer's direct delivery fleet already scheduled to bring more, and, behind all of it, a shared numbering standard and an electronic messaging format that let a manufacturer, a distributor, and a cashier all agree, instantly, on exactly which product just left the shelf. The gum disappears into a shopping bag. The reorder it triggered is already moving.

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

The self-service shelf is a young idea. For most of the grocery trade's history, a shopper never touched the stock. You handed a clerk a list, and he fetched each item from behind a counter, weighing dry goods out of barrels, while you waited. Those stores were small in every sense: a typical counter-service grocery carried a few hundred items, sometimes as few as 200, in 500 to 600 square feet, and a full week's food meant separate trips to a butcher, a baker, a greengrocer, and a dry-goods store. Because a "small army of clerks" had to be paid to run every order, the labor was baked into the price. Self-service arrived when Clarence Saunders opened the first Piggly Wiggly in Memphis in 1916 and let customers pick from open shelves themselves, and the format took its modern shape on August 4, 1930, when Michael Cullen opened King Kullen in a converted Queens garage stocked with about 1,000 items.

The jump from there is enormous and it runs in two places at once. On the shelf, a full-size supermarket now carries 30,000 to 50,000 distinct products, a hundred times the old counter store's range. Behind it, the "what do we need more of" question that a clerk once answered by walking the aisles with a clipboard and mailing a paper order is now answered continuously by software, one calculation per item per store, triggered by the barcode scan at checkout. The people did not disappear. In U.S. grocery stores alone, stock clerks and order fillers number in the hundreds of thousands, and much of their work happens overnight, turning a pallet into a stocked aisle before the doors open so the morning looks effortless to the first shopper through the door.

For a reader, all of this buys back time and choice. You can walk in expecting the same crackers in the same spot, buy a week of meals in twenty minutes, and find a gap already refilled by morning without ever learning it was empty. Before self-service and computerized reordering, none of that was on offer: you shopped most days, took what the clerk had, waited while he assembled the order by hand, and often made the rounds of several specialty shops to fill a single dinner. The morning it fails is rare and jarring precisely because the system usually hides itself, a ransomware attack on a distributor or a bad software update at the registers, and suddenly the shelves have holes or the checkout can only take cash while pallets of food sit in the back that nobody can ring up. The ordinary reliability of finding what you came for is held up by forecasters, drivers, and overnight crews the receipt never names.

Real-world examples and recent developments

Some of the biggest names in getting a scanned product onto a shelf never appear on a receipt.

  • C&S Wholesale Grocers (1918): founded in Worcester, Massachusetts, by Israel Cohen and Abraham Siegel, C&S is now headquartered in Keene, New Hampshire, and is the largest wholesale grocery distributor in the United States, supplying nearly 137,000 different products to more than 7,700 independent supermarkets, chain stores, and institutions, and reporting $21.7 billion in revenue in 2023. Wikipedia, C&S Wholesale Grocers
  • Kroger and Ocado (since 2018): Kroger partnered with the British online grocery and robotics company Ocado to build automated "customer fulfillment centers," warehouses where robots pick and pack online grocery orders instead of a store employee, with facilities opened in locations including Groveland, Florida; Forest Park, Georgia; Dallas; Frederick, Maryland; Pleasant Prairie, Wisconsin; Romulus, Michigan; and Phoenix. Grocery Dive, Tracking the development of Kroger's automated e-commerce center network
  • Impinj (2000): a Seattle company founded by Caltech researchers Carver Mead and Chris Diorio to commercialize passive RFID chips. Impinj became one of the leading suppliers of the RAIN RFID chips going into the item-level tags now appearing in Kroger's and Walmart's fresh departments, and had shipped its 100 billionth RFID tag chip by 2024. Impinj, Celebrating 25 Years: "We're Just Getting Started"

Recent developments

  • Kroger-Albertsons merger blocked (December 10 to 11, 2024): a federal court in Oregon and a separate Washington state court both blocked Kroger's roughly $25 billion proposed acquisition of Albertsons, and the two chains terminated the deal on December 11, 2024; Albertsons then sued Kroger for breach of contract, seeking at least $6 billion in damages including a $600 million termination fee. NPR, Kroger and Albertsons grocery megamerger blocked by courts
  • Kroger scales back its Ocado warehouses (November 19, 2025): Kroger announced it would close several of its Ocado-powered automated fulfillment centers and shift back toward store-based order fulfillment, taking about $2.6 billion in charges against an expected $400 million annual financial benefit from the change. Grocery Dive, Kroger acknowledges that its bet on robotics went too far
  • Walmart expands RFID into fresh departments (October 2025): Walmart announced it would bring RFID tagging into its meat, bakery, and deli departments to improve freshness tracking and cut unsold food, extending the same item-level approach Kroger began in its own bakery department in late 2024. Impinj, Data Accuracy: Helping Solve Grocery's $473 Billion Food Waste Problem

Glossary

UPC (Universal Product Code). The twelve-digit barcode standard used on most retail products in North America, encoding a manufacturer number and an item number assigned by that manufacturer.

GTIN (Global Trade Item Number). GS1's broader international term for the number a barcode encodes, covering UPC, EAN, and related formats used outside North America.

SKU (stock keeping unit). A retailer's own internal inventory code for a product, which may or may not match the manufacturer's UPC or GTIN.

GS1. The global nonprofit standards body that assigns manufacturer barcode prefixes and maintains the technical specification for how those numbers are encoded into scannable symbols.

Point-of-sale (POS) system. The combined hardware and software at a checkout lane that processes a sale, looks up pricing, and reports the transaction to a store's inventory database.

EDI (electronic data interchange). A standardized format for exchanging business documents, such as purchase orders and shipping notices, directly between retailer and supplier computer systems.

ASN (advance ship notice). An EDI message, commonly transaction type 856, that tells a receiving distribution center or store exactly what's on an incoming shipment before it arrives.

Direct store delivery (DSD). A distribution method where a manufacturer or its distributor delivers straight to a store, bypassing the retailer's own distribution center, common for bread, snacks, soda, and dairy.

Distribution center (DC). A retailer-operated regional warehouse that receives bulk shipments from manufacturers and breaks them into store-specific loads for delivery on the retailer's own trucks.

Just-in-time (JIT) inventory. A replenishment approach that keeps deliveries small and frequent, closely matched to expected demand, to minimize excess stock and waste.

Buffer stock (safety stock). Extra inventory held above forecast demand specifically to absorb unexpected supply delays or demand spikes.

Shrinkage. Inventory lost to theft, damage, spoilage, or administrative error, tracked and actively managed as an ongoing operational cost.

Cycle count. A routine partial inventory count of a rotating subset of stock, used to catch discrepancies between recorded and actual inventory without a full shutdown.

RFID (radio-frequency identification). A tagging technology that allows items to be identified and counted by radio signal without a direct line-of-sight scan, increasingly used for item-level tracking in fresh grocery categories.

Sources and notes

Open questions

  • Exact current SKU counts and store counts shift year to year and by source; treat the figures here (roughly 33,000 SKUs per store, Kroger's 2,722 stores, an $880 billion U.S. market) as recent representative snapshots rather than fixed numbers.
  • The longer-term operational and financial fallout from the 2025 UNFI ransomware attack, and how grocery distributors are changing their cybersecurity practices in response, was still developing at the time of writing.
  • How far RFID item-level tagging spreads into center-store packaged groceries, as opposed to remaining concentrated in fresh categories, is not yet settled and depends heavily on per-tag cost coming down further.

Next, the products that make it onto the shelf still have to survive the trip there without spoiling. How food, medicine, and vaccines stay cold in transit 👉