What problem does it solve?
A shelf that looks fine from the end of the aisle can still be missing the one size or flavour a shopper wants. Store associates and field sales representatives have traditionally found this the slow way: an occasional manual walk of the aisles with a clipboard or a phone, covering only a fraction of the floor. By the time a gap is logged, the sale is already lost, and a walk that only reaches part of the shelf misses far more out of stocks than it finds.
The problem sits on both sides of the shelf. Retailers lose sales and see slower, less accurate replenishment signals when the shelf state is not known. Consumer goods manufacturers lose distribution and shelf share they have paid for in trade terms, and field reps can end up spending a large share of a store visit on manual counts instead of selling: Trax's case study on Henkel reports that its reps spent only ten minutes of every hour in a store on active selling before Henkel started using its shelf image recognition.
How does it work?
- Capture the shelf. An autonomous robot that traverses the aisles on a schedule photographs every shelf section several times a day; a fixed camera, a shelf edge sensor or a field rep's own phone camera can capture the same view without a robot.
- Read the image. Computer vision matches what is on the shelf, and what is missing, against the planogram and the product catalog: out of stocks, low stock, misplaced items, wrong facings and price tag mismatches.
- Prioritize the findings. A well designed system ranks findings by expected sales impact, rather than listing them in shelf order, so the highest value gap gets fixed first.
- Push a task, not a dashboard. A prioritized, store or route specific task list reaches the associate's handheld or the field rep's phone, with the aisle, the SKU and the action needed.
- Close the loop. Completed tasks, and the shelf state after restocking, feed back into inventory and replenishment systems and into the KPI dashboards category and account managers use to negotiate with retail partners.
- Audience
- Back office
- Autonomy
- Assist
- Adoption
- Early adopters
- Channels
- Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Out of stock reduction | 35% | 4.3% to 60% | 4 | 2 organization, 2 vendor |
| Revenue uplift | Too few to pool | 2.1% to 7% | 2 | 2 vendor |
| Detection improvement | Too few to pool | 14x | 1 | 1 organization |
Value drivers: Lower cost to serve, Revenue growth, Employee productivity.
Indicative value
A regional grocery chain with 100 stores and USD 25 million in annual sales per store
USD 2 million to USD 42 million
Sales recovered from fewer out of stocks per year
How this is calculated
Formula: stores * salesPerStore * oosSalesAtRisk * oosImprovement. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Stores in the chain stores, stores | 50 | 150 | Editorial assumption for an illustrative regional grocery chain, a range around the size of Schnucks' 112 store chain referenced on this page. Replace with your own store count. |
| Annual sales per store salesPerStore, USD per store per year | 20,000,000 | 35,000,000 | Editorial assumption for a regional grocery chain. Replace with your own store level revenue. |
| Share of sales at risk from out of stocks before the change oosSalesAtRisk, fraction of sales | 0.02 | 0.04 | Editorial assumption, replace with your own on shelf availability audit. |
| Share of that risk removed by shelf monitoring oosImprovement, fraction of the at risk sales | 0.1 | 0.2 | Conservative against Schnucks' reported at least 20% reduction in out of stocks in stores using the robot, reported at the September 2020 expansion announcement, kept lower here because that figure covers one retailer's deployment, not a cross industry average. |
What it leaves out: Gross sales recovered only. It leaves out the cost of the cameras, robots or platform, the labor used to act on the task list, and any change in markdowns or waste from tighter replenishment signals.
Who already uses it?
8 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
B&R Stores
United States · Retail and ecommerce · 2026
B&R Stores, the family owned grocery operator behind banners including Russ's Market, Allen's, Super Saver and Cash Saver in Nebraska, Iowa and Missouri, introduced Simbe's Tally autonomous shelf scanning robot to select locations in Lincoln, Nebraska in June 2026. Tally captures real time, shelf level data on product availability, pricing and placement so store teams can respond faster. Simbe's press release quotes B&R's Director of Marketing on the goal of freeing associates from manual audits, and notes store associates typically spend up to 30 hours a week on manual inventory tasks before the robot takes them over.
No outcome disclosed.
Walmart de México y Centroamérica
Mexico · Retail and ecommerce · 2026
Walmart de México y Centroamérica, the country's largest retailer, is expanding its partnership with Vusion to deploy the EdgeSense connected store platform, which combines smart shelf infrastructure, computer vision, AI, electronic shelf labels and real time retail data into one operating system for physical stores, across its Walmart Express chain and into its Supercenters. The initial rollout equips Walmart Express stores with more than 1.7 million electronic labels and more than 180,000 EdgeSense smart rails, with a separate pilot in the Bodega format, aimed at automating in store tasks such as price updates and inventory management and freeing associates to focus on customers.
No outcome disclosed.
BJ's Wholesale Club
United States · Retail and ecommerce · 2025
BJ's Wholesale Club Holdings, which operates nearly 300 membership clubs across 18 states in the Eastern United States, deployed Simbe Robotics' Club Solution, built on the Tally autonomous robot, across packaged goods, meat, dairy, frozen, beauty and alcohol sections to check on shelf availability, price and promotion accuracy, and pick path efficiency on the taller, warehouse grade racks a club format uses. Simbe's case study states 90% fewer pricing and promotion errors, 60% fewer out of stock items and 50% shorter ecommerce fulfillment, and carries a BJ's quote on how Tally's imagery now creates "a digital twin" of each of its buildings.
- Out of stock reduction: 60%
"60% less out-of-stock items"
Claimed by: vendor
SPAR Austria
Austria · Retail and ecommerce · 2025
SPAR Austria says it was the first food retailer in Austria to test an inventory robot in regular store operations, running Simbe's Tally at a EUROSPAR supermarket in Vienna and an INTERSPAR hypermarket in Eisenstadt from July 2025 for a five month pilot. Tally moves through the store several times a day, scanning shelves, identifying gaps and changes in product placement, and sending the findings to employees' mobile devices. In May 2026, SPAR announced it would expand the pilot to six EUROSPAR and INTERSPAR stores after the trial showed automated shelf scans could significantly cut the time staff spent on manual inspection rounds and searching for missing stock.
No outcome disclosed.
Wakefern Food Corp.
United States · Retail and ecommerce · 2025
Wakefern Food Corp., the largest retailer owned supermarket cooperative in the United States, expanded Simbe Robotics' Tally shelf scanning robot across its network after a proof of value for inventory tracking, deploying it to Inserra Supermarkets' 23 ShopRite stores and to ShopRite Supermarkets, Inc. (SRS), which operates 37 ShopRite locations in New York and New Jersey. Both operators run Tally across all of their stores. SRS's Manager of Shrink Analytics and Inventory Control, Nick DeSetto, says Tally cut the out of stock rate by 50% in one of the company's highest volume stores by providing a prioritized pick list for every category, letting the team cover every aisle every day for the first time.
- Out of stock reduction: 50%, one of the highest volume stores of ShopRite Supermarkets, Inc.
"Tally has helped my team reduce our OOS rate by 50% in one of our highest volume stores by providing prioritized pick lists for every category."
Claimed by: organization
GoGo SqueeZ
United States · Manufacturing · 2021
GoGo SqueeZ, described by Trax as one of the fastest growing international healthy snack food companies, with 68% of the market share in the pouch category, used Trax Image Recognition to get SKU level visibility into shelf placement, price and display across its retail accounts. Its own analytics team turned the shelf data into scorecards and store level recommendations for its field force, and Trax's case study reports outlets that adopted the recommendations grew three times faster than outlets that did not.
- Revenue uplift: 7%, versus the period before using Trax
"GoGo SqueeZ used Trax IR and saw 11% increase in total points of distribution, 26% sales growth YoY, and 7% sales growth versus the period prior to using Trax."
Claimed by: vendor
Henkel
Germany · Manufacturing · 2021
Henkel, the German consumer goods manufacturer, used Trax Image Recognition to monitor more than 900 SKUs across 2,500 stores in Germany. Trax's case study reports that before the rollout, Henkel's sales reps used manual methods to measure in store distribution and shelf share and spent only about ten minutes of every store hour on active selling, and that with Trax, Henkel monitored close to 500,000 products in three and a half months and identified more than 20,000 that were regularly missing from shelves.
- Revenue uplift: 2.1%
"Trax enabled Henkel to reduce OOS by 4.3% and see sales uplift of 2.1%."
Claimed by: vendor - Out of stock reduction: 4.3%
"Trax enabled Henkel to reduce OOS by 4.3% and see sales uplift of 2.1%."
Claimed by: vendor
Schnuck Markets
United States · Retail and ecommerce · 2017
Schnuck Markets, a St Louis based regional grocer, piloted Simbe Robotics' Tally shelf scanning robot at several stores from July 2017 and deployed it to 15 stores in fall 2018. In September 2020, Schnucks and Simbe announced an expansion to 46 more supermarkets, bringing the rollout to 62 of the company's 112 stores. The robots travel the store floor two to three times a day and scan roughly 35,000 products each pass; with the expanded deployment, the companies said Tally would scan an average of more than 4.2 million products daily, feeding out of stock and shelf condition data into Schnucks' inventory and replenishment systems.
- Detection improvement: 14x, in stores using the robot, reported at the September 2020 expansion announcement
"Dave Steck, vice president of IT infrastructure and development at Schnucks, reported that Tally has provided 14 times more out-of-stock detection than manual auditing and at least a 20% reduction in out-of-stocks in stores using the robot."
Claimed by: organization - Out of stock reduction: at least 20%, stores using the robot
"Dave Steck, vice president of IT infrastructure and development at Schnucks, reported that Tally has provided 14 times more out-of-stock detection than manual auditing and at least a 20% reduction in out-of-stocks in stores using the robot."
Claimed by: organization
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A current planogram or shelf plan per store or store cluster
- A product catalog with images, matched to barcodes and SKUs
- A task or work order system that store associates or field reps already use
Systems to integrate
- The shelf camera, sensor or robot platform's own detection API
- Store inventory and replenishment system
- Field force or task management tool used by associates and sales reps
- Point of sale or category management reporting, for the resulting KPIs
Complexity: Medium
The computer vision itself is usually bought, not built: the work is fitting it to the retailer's own planogram and product catalog, routing the resulting tasks to the right device, and closing the loop back into replenishment and field sales systems.
- 1
Start with the highest value gaps, not full coverage
Begin with the categories and store clusters where out of stocks cost the most, rather than trying to cover every shelf on day one.
- 2
Match detections to the current planogram
Load the planogram and product catalog the vision system checks against, and keep both current, or every detection is measured against a shelf plan that no longer matches reality.
- 3
Turn detections into one task, not a report
Route each confirmed gap as a single, prioritized task to the associate or rep responsible, with the aisle and SKU, instead of a dashboard someone has to remember to check.
- 4
Feed the loop back into replenishment
Send confirmed out of stocks and shelf state changes into the store's inventory and ordering system, so the detection improves replenishment accuracy as well as visibility.
- 5
Measure detection against a manual baseline first
Before trusting the system, compare its findings against a manual audit on a sample of shelves, so false positives and blind spots are known before staff are asked to act on them.
Guardrails
- A confidence threshold below which a detection is queued for human review, not auto pushed as a task
- Store staff and field reps can mark a task as wrong, and that feedback retrains the matching
- No biometric identification of shoppers who appear incidentally in shelf images
- Clear ownership of the planogram data feeding the comparison, with a review date
KPIs to instrument
- Out of stock detections versus a manual audit baseline, on the same shelves
- Time from detection to a completed task
- Task completion rate and false positive rate reported back by staff
- Sales or share of shelf in the categories covered, before and after
Human in the loop
People still do the physical work: restocking, fixing a misplaced item, checking a low confidence detection before acting on it. Category and account managers review the resulting KPI trends, not individual detections, and decide what changes in the shelf plan or the order.
Common failure modes
- A stale planogram makes every comparison wrong
- The system flags gaps against a shelf plan nobody updated. Give the planogram an owner and a review date, the same as any other approved content.
- Detection without action
- Gaps are found but no one restocks them, so the dashboard looks busy and the shelf does not change. Route a single task to a named owner, not a report to a queue.
- Cameras that see people, not just shelves
- A shelf camera's field of view can capture shoppers or staff. Angle and mask the feed so only the shelf is analyzed, and document that in the privacy notice.
- Treating a pilot's numbers as the chainwide number
- Early pilot stores are often the easiest cases. Track results separately by rollout wave before quoting a single improvement figure company wide.
What are the risks and rules?
EU AI Act
Minimal risk
The system analyzes shelves and products, not natural persons, and makes no decision about a person's access to a service, a job, credit or benefits, so it falls outside Annex III. If camera placement or software is changed to identify or track individual shoppers or staff, the analysis changes and biometric identification rules would apply.
Controls to put in place
- Camera and robot placement and processing scoped to shelves and products, not people
- A named owner for the planogram and product catalog the system checks against
- A manual audit sample kept running after go live, to catch drift in detection accuracy
Frequently asked questions
- Does AI shelf monitoring replace store associates?
No. It replaces the manual walk that finds a gap, not the physical work of restocking it. Trax's case study on Henkel reports that reps' active selling time rose after Henkel adopted shelf image recognition, and Schnucks' VP of IT infrastructure said the wider Tally rollout would free store teams from tedious inventory tasks to focus more on service, a stated aim rather than a separately measured result.
- How much of an improvement in out of stocks is realistic?
It depends on the baseline and how much of the store is covered. Schnucks reported at least a 20% reduction in out of stocks and 14 times more out of stock detection than manual auditing, in stores using the robot, at its September 2020 expansion announcement, and Trax reports a 4.3% reduction in out of stocks for Henkel from shelf image recognition across its distribution. Treat any single figure as tied to that organization's starting point, category mix and who is making the claim.
- Does this need robots, or do cameras work too?
The evidence on this page covers two modes without a fixed camera. Autonomous robots such as Simbe Robotics' Tally scan aisles on a schedule, as at Schnucks, and vendor image recognition such as Trax, fed by images the field force captures, lets a manufacturer's sales reps check shelf compliance without a robot: GoGo SqueeZ's case study describes reps uploading a shelf photo, and Henkel's describes reps capturing SKU level information without naming the capture method. Fixed shelf edge cameras are also sold for this job, but no evidence record on this page covers one.
How to cite this page
Blits.ai AI Use Case Library, "AI computer vision for store shelf and stock monitoring", last verified 29 September 2026, https://www.blits.ai/ai-use-cases/store-and-shelf-monitoring. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 12 November 2026: Added the out of stock reduction KPI, new in the taxonomy, so the out of stock results of Schnucks, Henkel, Wakefern (ShopRite Supermarkets, Inc.) and BJ's Wholesale Club count as benchmarks.
- 30 September 2026: Added deployments: Walmart de México y Centroamérica.
- 29 September 2026: First published