Psst… We received a tip…

Walmart has rolled out digital shelf labels across thousands of U.S. stores. Around the same time, Walmart also picked up patents tied to algorithmic pricing, which was enough for a lot of people to connect the dots and assume they had patented prices that change on the shelf whenever it feels like it.

You can see why people freaked out. One minute your pasta is $3.98, the next minute it is $4.41 because too many people wanted spaghetti after work... and the all-seeing algorithm says so.

But are these events and Walmart's recent patents actually connected? We read them so you don't have to. Here's our take.

Walmart did get two pricing patents, and they do deal with algorithms. But they are not the same thing as a digital shelf label flickering up and down in front of you in aisle seven. Both patents were assigned to Walmart Apollo, LLC, and both are much more about how Walmart decides prices in software than about a tag on a shelf changing by itself.

The first patent is the more direct one. It is called “System and method for dynamically and automatically updating item prices on e-commerce platform.” It is aimed at online pricing, and more specifically at markdowns. The patent says this is for situations where Walmart wants an item to hit a target inventory by a certain date, which is saying: we need to clear stock, but we do not want to cut the price more than necessary. 

The second patent is the forecasting engine sitting behind that kind of decision. It is called “Methods and apparatus for determining item demand and pricing using machine learning processes.” This one is broader. It looks at past sales, builds features from that history, uses a time-series machine learning model to predict future demand, and then generates a recommended price for the item.

One patent tries to answer what will happen next, and the other tries to answer what price should we put on the site right now. They are related because one can feed the other. 

Start with the forecasting patent. The system pulls in sales data from an earlier period, including how many units sold and what price they sold for. It turns that history into features, feeds those features into a trained model, and gets back a prediction for demand in a future period like next week, next month, or next season. Then it uses that prediction to recommend a price and stores that result for later use. It can also factor in category budgets and substitute products, which means it is not just guessing demand for one lonely item in a vacuum. 

Then comes the markdown patent. A merchant identifies an item that needs to be marked down and gives the system a target, like clearing inventory by a certain date. The system checks what data it has. If it has both price elasticity data and predicted demand data, it uses a first model to calculate a markdown price. If it does not have both, it falls back to a second model based on a decay rate, the current price, and the data that is available. Then it applies upper and lower bounds and sends the final price to the website.

The patent also makes clear what kinds of signals can feed this machine. It mentions user session data from browsing activity, purchase data, search data, catalog data, markdown history, price elasticity, predicted demand, and item-related features. That does not mean Walmart patented secret one-to-one prices for each shopper. It means the system can use a lot of behavioral and sales data to learn how fast an item is moving and how sensitive demand seems to be to price.

That is why the shelf-label panic missed the real mechanism. Walmart says its digital shelf labels are a closed system, do not collect shopper information, and are mainly there to make store operations faster and more accurate. The patents, by contrast, are about the software logic behind pricing decisions, especially online markdown decisions.

The patents say markdowns matter when a retailer needs to clear unsold, seasonal, or perishable goods by a specific date, but most older markdown systems were built for physical stores, where each price change costs time and labor. E-commerce is different. You can change prices more often, react to daily demand shifts, and use fresher data, but that also means the old blunt way of doing markdowns stops looking good fast.

Walmart's global eCommerce grew another 27% in its fiscal 2026 third quarter. One patent describes the goal as increasing sales velocity while controlling contribution profit loss. The other says retailers struggle at both ends of the problem: price too high and goods do not move, price too low and revenue suffers or stock runs out too quickly.

Walmart’s own FY23 earnings materials said increased general merchandise markdowns pressured gross margin. More broadly, McKinsey estimated U.S. retailers were sitting on roughly $740 billion in unsold goods in 2022 and said markdowns were squeezing margins.

Walmart is not alone here. This is already a real software category, not some weird one-off experiment buried in a patent filing. IDC says it assessed 21 key vendors in retail price optimization in 2025, and found that price optimization was the second-most important merchandising IT investment, with 56.02% of retailers saying they planned to increase spending in the next 12 months. 

There are already companies selling almost this exact promise. Blue Yonder has a markdown product that recommends the timing and depth of price cuts to hit sell-through goals. Oracle’s retail pricing software explicitly covers regular pricing, promotions, markdowns, and even targeted offers across the product life cycle. Aptos, through Revionics, pitches centralized pricing across stores, channels, and banners. RELEX says its markdown tools use demand forecasts, inventory balances, and price elasticity to identify the best markdowns by store and timeline.

Market-size estimates vary, but one recent report puts the broader global price-optimization software market at about $1.4 billion in 2025, growing to roughly $2.67 billion by 2030. (zebra.com)

The more interesting opportunity is what happens when the same pricing logic moves into giant service markets where capacity still expires, pricing is still messy, and software penetration looks thinner.

Look at home services. Harvard’s Joint Center for Housing Studies says the U.S. spends over $600 billion a year on home maintenance and improvement, and that spending is expected to remain above $600 billion through 2025. The whitespace here is using pricing software to fill underused technician time, smooth route gaps, and price standardized jobs with more precision than today’s quote-by-quote playbook.

Logistics may be even bigger. Grand View Research estimates the global warehousing market at about $1.08 trillion in 2024, growing to roughly $1.73 trillion by 2030. Inside that, cold storage alone is estimated at $185.75 billion in 2025, with forecasts to more than double by 2033. If retail patents are about not wasting inventory, logistics pricing is about not wasting space, routes, and time windows.

EV charging is smaller today, but that is exactly why it is interesting. The IEA says electric car sales topped 17 million globally in 2024, more than 20% of all new car sales, and public chargers have now passed 5 million worldwide after adding more than 1.3 million in 2024 alone. One recent estimate puts the EV charging management software market at $3.4 billion in 2025 and growing to $31.5 billion by 2035. So the hardware build-out is already happening. The whitespace is the software layer that decides who gets access, when, and at what price without turning the experience into a mess.

RealPage offers a lesson for those looking to build in price optimization. In November 2025, the U.S. Department of Justice filed a proposed settlement with RealPage after alleging its software relied on nonpublic, competitively sensitive information from competing landlords and included features that helped align prices across rivals. That is the line you do not want to drift across. Using your own data to optimise your own prices is one thing. Using a shared system that nudges competitors toward the same answer is something else entirely. (justice.gov)

Regulators are going to care about where your data comes from, who can see what, and whether the product makes competitors act less independently. 

Then there is the consumer side. The moment you start feeding personal data into price decisions, you are stepping into a much hotter political and regulatory zone. (ftc.gov)

There is also a more basic risk that has nothing to do with AI and everything to do with timing. If a customer sees one price, adds the item to a cart, and then gets hit with a different effective price later without a clear explanation, you are inviting trouble. The FTC has been very clear that online sales practices still sit under the same ban on unfair or deceptive acts or practices, and its fee rule now explicitly bans bait-and-switch pricing and other tactics that obscure or misrepresent total prices in covered sectors.

The clearest signal is Blue Yonder, which sells software that helps retailers and brands plan inventory, demand, fulfilment, and markdowns. Panasonic bought the company at an $8.5 billion enterprise value in 2021. That matters because it shows pricing and markdown software can become strategic enough for a giant buyer to treat it as core infrastructure.

Then you have Revionics, which is even closer to Walmart’s patent problem. Revionics built software for retail pricing, promotions, and markdown optimisation, and Aptos bought it to add price optimisation into its wider merchandise lifecycle stack. That is a useful clue for founders. The exit path can involve becoming a big standalone company and can also be the pricing layer that a larger retail software platform needs to own.

The strongest scaled independent example is RELEX. RELEX helps retailers and manufacturers plan demand, inventory, replenishment, and markdowns, which makes it relevant because price optimisation works best when it is tied to stock levels and sell-through. Blackstone and TCV increased their investment in RELEX in 2024, and the company said it closed 2025 with 30% subscription revenue growth and 28% ARR growth. (blackstone.com)

If you want the earlier-stage version of the story, look at 7Learnings. It is a Berlin company focused on predictive pricing and retail optimisation, and it raised more than €10 million in a 2025 Series B. That is the more realistic startup signal. Big platforms are being bought or scaled up, but investors will still fund younger pricing players if they can show a clean wedge, measurable margin impact, and a product that fits directly into merchant decisions.

So no, the scary version of the story does not quite hold up. Walmart says its digital shelf labels are a closed system, do not collect shopper data, and are used to keep shelf prices accurate and consistent, with updates typically pushed outside shopping hours. That makes them a different thing from the patents, which are aimed at software-driven pricing decisions, especially around online markdowns and demand forecasting.

But you should not dismiss the bigger signal either. Even if the shelf tags and the patents are not the same system, they point in the same direction of more flexible pricing, more operational control, and more software sitting in the middle of decisions that used to be slower, messier, and more manual.

And that is why this one is worth reading for yourself. Start with US 12,524,776 B2 and US 12,572,954 B2.