Why Real-Time Inventory in Retail Has Never Worked Well

Shoppers find it frustrating to check whether an item is available and then try to purchase it. But they are not nearly as frustrated as retail IT leaders.

The cruel truth is that real-time inventory systems typically break down when they are needed the most. 

This problem has haunted retailers for decades. Yet today’s technology — including digital analytics and even agentic AI — are unable to cost-effectively address the issue.

A typical example: On December 20, a parent is frantically trying to purchase one of the season’s hottest toys for her four-year-old. Stores across the region are sold out. The mother finally finds a site that says that one of its stores — 56 minutes away — has five in stock. 

When the parent cannot reach anyone at that store by phone, to beg them to put the toy aside at the customer service desk, she speeds to that location. At every stop light, she checks her phone to ensure that the site still says that the store still has five toys left for sale.

She arrives at the store, runs in, and finds none on the shelf. Wondering whether she simply took too long, the would-be customer checks her phone again and, yep, it still says that five are available.

What happened? What went wrong? 

The Availability Reality Gap

This is where the purchasable inventory slams into theoretical inventory. Let’s call it the availability reality gap. 

Plenty of things could have gone wrong. Store availability is a simple math equation: Take the number of that specific product — a.k.a. an SKU — and subtract how many of that SKU were sold. In other words, if the store received 50 of that item and sold 45, the system concludes that five are still available.

Unfortunately, that arithmetic assumes perfection along the way. That’s not how reality works. Here are common scenarios that cause discrepancies.

A Counting Problem

The store logged that it received 50 items from a truck delivering 50 of that specific SKU. 

During the holidays, stores add lots of seasonal workers who might not precisely count boxes or verify that every single product is what it claims to be. The temp workers place more attention on unpacking and forklifting products to their destinations. 

Maybe the store never received 50 items. It might have only received 45. The five supposedly remaining in stock never existed.

Dropped Packages

The truck might indeed deliver 50 items, but maybe in the insanity of a major retailer’s backroom in mid-December, four or five boxes fell off the forklift. They won’t be discovered until mid-January.

Those five items never make it out to the shelves. Again, the items are not actually available for purchase.

Shoplifting

Despite the loss prevention team’s best efforts, thieves still manage to grab some product. The season’s hottest toy is likely a top theft target.

Typically, the only way the number of available items is reduced in the system is when the point-of-sale system registers that someone paid for it. A stolen item shows up as available for purchase even though it’s not.

Misplaced Products 

It is not unusual for shoppers to grab a product and later decide that they no longer want it. Maybe the shopper’s spouse calls to report that they just grabbed the toy at another store at half the price. When a shopper changes their mind, they might eject the item from their shopping cart wherever in the store they happen to be. Result: A child’s stuffed animal is abandoned in the middle of the microwave ovens department. 

That SKU actually is still available for purchase, but there is no realistic way to find it.

The Shopping Cart Black Hole

It’s entirely possible that the five SKUs that the inventory application shows as available for purchase are in the shopping carts of five people walking the aisles right now. The items might even be in one of the long lines for checkout.

There are a handful of other incidents that can fuel the availability reality gap, such as a cashier accidentally scanning an item twice and thereby telling the inventory management system that it has fewer items than it really has. And here’s a fun fact: If the store associate overcharges and the customer later calls to complain and that customer gets a credit, many systems don’t automatically fix the inventory level. However, the five reasons above are the most common.

Are there ways to fix this? Absolutely. But each approach has pros and cons. 

Solutions That Aren’t

Retail IT has toyed with various solutions but has generally failed to find anything that is cost-effective at volume.

Years ago, item-level RFID was seen as the fix for real-time inventory. It ultimately stalled, only being used for especially expensive, high-theft-risk items. Those RFID tags plateaued at about five cents per tag, and retailers concluded that RFID solutions wouldn’t make financial sense until the cost was lower — perhaps sub-one-cent per tag. 

This became a chicken-and-egg headache. Chip suppliers told retailers that if they used RFID for everything, the per-unit cost would be the price they wanted because of volume discounts. The retailers, however, wouldn’t buy until they saw the much lower prices. Stalemate.

A different approach is one that Amazon Go and a few other niche retailers are pushing: digital analytics. In this scenario, a large number of high-res cameras capture everything going on in the store. That system ostensibly knows where every product is and, if someone touches the item, what happens next. 

That doesn’t eliminate the shopping cart black hole, but the store becomes aware of the change and has the option of removing that SKU from purchasable inventory. And it does effectively negate most other inventory problems. 

However, it is expensive, and it works best in very small-footprint stores. Good luck getting it to work in a large Costco or Target. And even Amazon announced in January that it was rethinking the camera concept — and, in fact, that it will shut down all Amazon Go stores. 

The changing nature of retail is another factor in the difficulty of real-time inventory. Mobile commerce and e-commerce rewrote many of the rules, with exceptions in grocery, restaurants, and some apparel segments. Stores are no longer the center of the retail universe. Online retailers can house inventory in as many warehouses as they can afford, in facilities often supported by robots. That sidesteps many of the real-time inventory challenges listed above.

But those industry changes also create new real-time inventory holes. Consider Amazon Marketplace, a group of small merchants around the world that sell through Amazon. The Marketplace creates the impression of close-to-infinite products in inventory. 

Those small merchants, however, deliver little to no visibility into quality control and certainly not into shipment details. A marketplace merchant can claim that a product will arrive on March 20, but Amazon has no solid idea whether it actually will.

This brings us back to our panicked parents shopping on December 20. They can go to Amazon and purchase that sought-after toy, but with no guarantee that the toy will actually arrive in time. And by the time the parents conclude that they won’t receive the item — perhaps on the afternoon of December 24 — they have few if any other viable options left.

The Best Path Forward

Retailers must embrace a combination of store design (high tech and low tech) and commonsense tactics to bring reality to inventory.

It seems that the best way to truly fix the availability reality gap is some version of digital analytics. Stores must be designed, or redesigned, with that in mind. For example, though it’s often not cost-effective to add new cameras to a store, every location — from aisles to checkout — must be reconsidered. 

There was an interesting effort some years ago to design stores without checkout areas. The rationale was to move entirely to mobile checkout and to repurpose the checkout lane space to showcase more products. POS would mostly shift to the cloud. JCPenney tried this plan, though it didn’t get sufficient support. 

But that seems to be the approach most likely to succeed. A high-tech store — in effect, one that truly embraces an intelligence edge infrastructure — could bring confidence back to brick-and-mortars.

Or perhaps a less radical approach would be to better handle all the data currently available. Store-level platforms would ingest sensor data, computer vision events, RFID reads, and POS transactions in real time, not in overnight batches. A store wouldn’t necessarily need RFID to note when a SKU leaves the shelf. If that action reduces — even temporarily — the available inventory by one, that would go a long way toward delivering meaningful data. 

Some merchants have opted for an ultra-low-tech approach: When a an online purchase is made for pickup in the store, the shopper is told to wait for a confirmation message. A human — remember those? — runs out to the aisle, grabs the product, and brings it back to the reserved area. And only then is a message sent to the shopper that they are clear to pick up the item. It may be low tech, but it usually works.

The problem is, yet again, Murphy’s law: These high-confidence confirmations are most needed when they are the most difficult to achieve (say, December 20–December 24).

Solving Inventory Fixes So Much

The wonderful part is that fixing inventory reality can address dozens of other problems and unlock much more revenue.

For example, what if product details could be far more precise? Consider: A parent needs to find a stuffed animal for a child. But not any stuffed animal. For a school project, it needs to be, let’s say, a lamb. 

Merchants usually list stuffed animals as “assorted.” Instead of visiting a dozen stores, what if a web or AI search could identify the stores that have a stuffed-animal lamb in stock? The retailer gains a definite sale and perhaps at a higher price. Requiring far more details from suppliers could bring more money to both. And yet, older systems do not support it.

The typical way to fix that particular shortcoming is to move from classic barcodes to QR codes. Barcodes are limited to 35 characters, whereas barcodes can support thousands of characters. And because QR codes can include lengthy URLs, with the data housed on a web server somewhere, the amount of data that they can present is almost unlimited. 

The QR code move would allow for far more details, and it also opens the door to capturing all kinds of different information. Retailers could access everything from shipping method to date of manufacture and how long the item can sit on the shelf. It could even track the product’s complete history of temperatures, which is critical when deciding whether to accept pallets of perishable items.

What about it, retailers? Maybe fill your cart in 2026 with some modernization?