Walmart is rolling out three tools tied to its digital shelf labels: Shop to Light, which lets a shopper find a product in the app and then flash the label in the aisle; Pick to Light, which guides associates picking online orders; and Stock to Light, which supports replenishment. Greg Cathey, the retailer's senior vice president of digital transformation, told CNBC the company describes itself as people-led and tech-powered, and frames AI as a way to free associates to spend more time with customers. All three features are scheduled to reach stores companywide by the holiday season.
Key Conclusions
First, an electronic shelf label is not a price tag with a battery; it is an addressable node that ties a physical position to a digital record. Once every label has an identifier, the store gains a coordinate system for products. That coordinate system is what makes wayfinding, picking and replenishment possible from the same data layer, and it is why the rollout matters beyond labour savings. The strategic asset is the mapping between product, position and live status, not the display hardware.
Second, the same data layer is what answer engines need in order to describe a store accurately. When a customer asks an assistant whether a specific item is in stock nearby, the answer depends on structured fields: product identity, location, price, availability and hours. Retailers that publish these fields consistently are more likely to be cited; those that publish prose and images without structure are invisible to the retrieval step. In that sense, shelf labelling and generative search optimisation are the same project viewed from two ends.
How Shelf Data Becomes Machine Readable
The chain runs from a label to a record to an answer. A label carries a unique identifier and a position. The position is bound to a planogram slot, which is bound to a SKU, which is bound to inventory and price records. When a customer scans or an associate picks, the event updates the record. For an answer engine to use this, the retailer must expose the record in a structured feed rather than only inside a private app. The gap between having the data and exposing it is where most retailers stall.
Identity is the hard part
Product identity is messier than it looks. The same item may exist as multiple SKUs across pack sizes, regions and suppliers, and the identifiers used internally rarely match those used by search engines or marketplaces. Without a canonical identifier, a location record cannot be reliably matched to a query. Retailers that invest in a master data layer, with attributes such as size, variant, material and compatibility, find that every downstream use, from shelf labels to answer engines, gets cheaper. Those that skip it rebuild the mapping for each project.
Freshness decides usefulness
A location record that is a day old is worse than no record, because it produces confident wrong answers. Freshness depends on how events flow: point-of-sale, cycle counts, receiving and online order picking all change availability. Retailers that unify these events into a single availability signal can publish a reliable status; those that keep separate systems end up with conflicting numbers and choose to publish nothing. The practical requirement is a defined refresh window and a clear rule for what counts as in stock.
Best Practices
For retailers considering a similar investment, the useful framing is to treat shelf labels as the start of a data programme rather than the end of a hardware purchase. The two practices below determine whether the programme produces reusable assets, and both are organisational as much as technical.
Design one record, many consumers
A single product-position record should serve the app, the associate handheld, the answer engine feed and the analytics layer. That means defining fields once and treating each consumer as a view rather than a separate dataset. The benefit compounds: when a field is added for wayfinding, it is immediately available for search; when availability rules change, every surface updates together. Organisations that let each team build its own dataset spend years reconciling numbers, and the reconciliation work never ends because the sources keep diverging.
Measure findability, not just pick speed
Picking speed is the easiest metric to report, and it is the one most likely to be achieved. Findability is harder and more valuable: the share of customer searches that end in a located product, and the time from search to shelf. Tracking it requires instrumenting the app journey and the in-store interaction together, which few retailers do. Those that do discover that assortment gaps and label errors show up as failed searches long before they show up as lost sales, giving them an early signal they can act on.
Common Mistakes
The first mistake is buying labels without changing the planogram discipline. A label is only addressable if its position is stable and documented. Stores that reset aisles frequently without updating the planogram will show wrong locations, and customers will lose trust in the feature after two failed trips. The fix is procedural: tie planogram changes to label re-registration, and audit a sample of aisles each week. Without that discipline, the hardware investment produces a novelty rather than an infrastructure.
The second mistake is treating the data as internal. Many retailers build excellent operational datasets and then expose almost none of it publicly, on the theory that inventory is competitive information. The result is that answer engines describe their stores using outdated third-party data, including wrong hours and discontinued items. A middle path is to publish structured facts that customers need, such as location, hours, category availability and service options, while keeping exact unit counts private. That balance captures the discoverability benefit without giving away trading position.
Findability as a Growth Metric
Findability connects marketing spend to store execution in a way that few metrics do. A campaign that drives traffic to a product the store cannot locate produces a wasted visit, and the customer attributes the failure to the brand rather than to the aisle. When findability is measured, the feedback loop closes: marketing sees which promoted items fail to be found, merchandising sees which slots are mislabelled, and operations sees which stores need replenishment attention. The same number serves three teams, which is rare and therefore valuable.
There is also a search-side effect. Answer engines increasingly synthesise local availability, and they favour sources that are consistent across the web. A retailer whose store pages, app and third-party listings disagree will be cited less often, because the retrieval step cannot resolve the conflict. Consistency is not glamorous work, but it is the difference between being the answer and being omitted from it. For categories with high intent, such as replacement parts and baby care, that difference translates directly into revenue.
Finally, findability changes how assortment decisions are evaluated. A delisted item that customers keep searching for is evidence of unmet demand, and a listed item that nobody searches for is evidence of wasted space. Retailers that pair search logs with shelf data can make assortment calls on evidence rather than on supplier negotiations. The asset required is modest: a search log, a location record and a rule for joining them. The organisational requirement is larger, because it asks category managers to defend space with demand data rather than with relationships.
Summary
Walmart's shelf-label tools are a practical example of a broader shift: the store is becoming a structured dataset that both people and machines can query. Wayfinding, picking and replenishment are three uses of one record, and the same record is what makes a store legible to answer engines.
The lesson for other retailers is to design that record once, keep it fresh, and measure findability rather than only pick speed. Those choices determine whether the investment produces an infrastructure that compounds quarter after quarter or a gadget that ages into a maintenance line item nobody wants to defend.
Data Sources
FAQ
What are digital shelf labels actually used for?
A: Price display is the baseline. The larger use is location: each label becomes an addressable node that supports wayfinding, order picking and replenishment from one data layer.
Why does product identity matter so much?
A: Without a canonical identifier, a location record cannot be matched to a customer query or a marketplace listing, so every downstream use requires rebuilding the mapping.
How fresh must availability data be?
A: Fresh enough that a customer is not misled. A defined refresh window and a clear in-stock rule matter more than raw speed, because inconsistent rules create conflicting answers.
Should retailers publish inventory data?
A: Publish what customers need, such as location, hours and category availability, while keeping exact counts private. That captures discoverability without exposing trading position.
How does findability affect marketing?
A: It shows which promoted items cannot be located in store, so wasted visits are visible before they show up as lost sales, and three teams can act on one number.
