Retail shelf monitoring uses cameras and software to spot empty or low facings and send a task to the person who refills them, ideally before a shopper finds the gap. The shelf images come from one of four places: fixed shelf cameras, a robot that drives the aisles, the ceiling cameras you already have, or staff taking phone photos. For an independent grocer or a small chain, the right choice comes down to which aisles you need to see, how alerts reach your staff, and how much accuracy each aisle really needs.
This guide is for the operator with a few stores, security cameras already on the ceiling and a busy team.
Quick answer
- What it does: turns each gap on the shelf into a refill task for a named person.
- What decides success: alert routing. A correct alert sent to nobody fixes nothing.
- What to expect on accuracy: shoppers, glare and product pushed to the back cause misses and false alarms. Test in your own aisles before you trust a vendor number.
Why empty shelves are mostly a store problem
The most cited study on out-of-stocks is old but still useful. Researchers working for the Food Marketing Institute, the Grocery Manufacturers of America and CIES surveyed 71,000 shoppers in 29 countries and found a world average out-of-stock rate of 8.3 percent (FMI, 2002).
Two findings from that study matter for a store operator:
- 70 to 75 percent of out-of-stocks start at the store. About 47 percent came from store ordering and forecasting, and about 25 percent from shelf restocking.
- Shoppers don't wait. When an item was missing, 31 percent went to another store, 26 percent switched brands and 9 percent bought nothing.
Shelf monitoring fixes the restocking share directly. The product is often in the back room and nobody has noticed the gap yet. Ordering needs a different fix, but shelf data points at it. When the same item goes empty every Tuesday, look at the order.
What vendor pages cover, and what they leave out
The big shelf monitoring vendors describe their products clearly. Focal Systems says its shelf cameras "continuously scan your shelves," flag out-of-stocks, low stock and planogram issues, and rank stocking tasks by lost-sales impact (Focal Systems). Simbe's Tally robot "autonomously scans every aisle," runs up to 12 hours per charge, and sends prioritized task lists to associates in its mobile app (Simbe).
Both are built for chain-wide rollouts. Simbe says pricing is a subscription that "varies based on store count, store size, scanning frequency, and selected capabilities." Neither page covers using a store's existing security cameras or where accuracy breaks down. That's fair for a product page, but it leaves the smaller operator's questions open.
Four ways to capture the shelf
| Option | How it sees the shelf | How often | Good at | Trade-offs |
|---|---|---|---|---|
| Fixed shelf cameras | Small cameras mounted facing the shelf, often across the aisle | Continuous or on a timer | Item-level reads on the aisles they cover | New hardware per aisle, install work, power or batteries to manage |
| Shelf-scanning robot | Cameras on a robot that drives every aisle | A few passes a day (Simbe says "multiple times per day") | Whole-store coverage, price tags, misplaced items | Subscription priced per store, aisle space, gaps between passes |
| Existing ceiling cameras | Your security cameras, with analytics added | Continuous | Chillers, endcaps, produce, front-of-store displays you already see | Angled views, limited reach into mid-aisle shelves, no fine SKU reads |
| Phone photo walks | Staff photograph shelves on a set route | Once or twice a day | Cheap start, good for planogram checks | Only as fresh as the last walk, relies on staff doing it |
Two public examples show the range. Wakefern, a cooperative whose members mainly run ShopRite stores, piloted Focal's shelf cameras in four supermarkets for over a year before planning a 50-store pilot (Supermarket News, 2019). Nisa, a UK network of independently owned stores, trialled Shelfie software on Axis network cameras, with "a real-time alert" sent "to a dashboard or mobile app informing staff of the need to re-stock specific items" (Axis).
Starting with the cameras you already have
Most grocery stores already have cameras over the places where gaps cost the most: dairy doors, beverage coolers, endcaps, produce and front-of-store displays. Before you buy anything, pull one still from every camera and ask three questions:
- Can you see the shelf face? A camera pointed down the aisle at an angle can tell you a cooler door looks half empty. It usually can't read which yogurt is missing on the bottom shelf.
- Is anything in the way? Hanging signs, a pillar, a display that moves every week.
- Is the light steady? Glare on glass doors and dark corners both cause misses.
Mark each camera "full view", "partial view" or "no view". That list tells you which areas the cameras can cover on day one. You add a shelf camera only where an area matters and your current view can't support it.
Out-of-stock vs planogram compliance
These get lumped together, but they answer different questions and need different setups.
| Out-of-stock detection | Planogram compliance | |
|---|---|---|
| Question it answers | Is there a gap where product should be? | Is the right product in the right spot, in the right number of facings? |
| What the system needs | A view of the shelf and a sense of "empty" | A current digital planogram and item-level recognition |
| Camera needs | Moderate. A clear view of the facing | High. Close, sharp images that can tell similar packs apart |
| Who usually cares most | Store managers and department leads | Brands, vendors who pay for space, merchandising teams |
| How fast it needs to be | Within the hour, while a sale is still at stake | Daily or weekly is usually fine |
The MAssist blog puts the difference well: out-of-stock means an item is missing where it should be, while planogram compliance means products are placed per the agreed layout (MAssist). A shelf can be fully stocked and still fail the planogram.
For most independents, out-of-stock comes first because it touches today's sales. Planogram checks only work if your planograms are current and digital. If yours live in a binder, start with gaps.
How alerts reach staff
This is where shelf monitoring projects succeed or stall. Detection gets the attention, but the alert is what gets the shelf filled.
Every alert needs an owner. Route by area and shift, with one escalation step:
| Area | Day shift | Evening | If not cleared in 30 min |
|---|---|---|---|
| Dairy and eggs | Dairy lead | Closing lead | Store manager |
| Beverage coolers | Grocery stocker on shift | Closing lead | Store manager |
| Produce tables | Produce lead | Closing lead | Store manager |
| Endcaps and promo displays | Grocery lead | Closing lead | Assistant manager |
(Example routing only. Your roles and times will differ.)
A few rules keep alerts useful:
- One gap, one alert. If a facing stays empty, escalate once instead of resending every five minutes.
- Send it where staff already look. A text, a handheld or the team's existing app. Nobody opens a dashboard mid-shift.
- Make "done" easy. One tap to clear, then the camera confirms the facing is full.
- Mute during known work. While a pallet is being worked in an aisle, hold alerts for that area.
- Track time from alert to refill. By store and by shift, it tells you whether the system is working.
Focal ranks tasks by lost-sales impact and Simbe sends prioritized lists to its app. Whatever you use, write the routing table down before go-live, or the alerts end up with whoever set up the account.
Accuracy limits to plan for
Every vision system misses some gaps and flags some shelves that are fine. The question is where, and how often, in your stores.
Common causes:
- Shoppers and carts. Someone standing in front of the cooler blocks the view. Good systems wait for a clear frame before deciding.
- Product pushed to the back. The front looks empty but there's stock behind it. It's a top source of false alarms, and the task is to face up the shelf.
- Glare and lighting. Cooler doors, fogged glass after a restock, and lights on a timer.
- Look-alike packs. Flavours of the same brand can look almost identical from across an aisle. This mostly affects planogram checks.
- Angle and distance. The further and steeper the view, the less detail. Bottom shelves suffer most.
Vendor accuracy numbers are worth reading carefully. Simbe reports "99% shelf label identification accuracy." Focal says its AI is "trained on over 2 billion images" and claims "unmatched shelf accuracy" without giving a method. A Net Solutions write-up of their own build reports over 90 percent detection accuracy, measured in controlled settings (Net Solutions). None of those numbers tell you how a system will do on your dairy doors at 6 p.m.
So test it. Run the system alongside a manual walk for two weeks and compare.
Worked example (made-up numbers, not real store data)
One store, one aisle group (dairy and beverage coolers), one weekday. Every number is an illustration of the math.
| Line | Example value |
|---|---|
| Alerts sent | 40 |
| Real gaps a person confirmed | 31 |
| False alarms (shopper in view, product pushed back) | 6 |
| Duplicates of an earlier alert | 3 |
| Share of alerts that were real (31 ÷ 40) | 78% |
| Gaps found on the manual walk that the system missed | 4 |
| Share of real gaps the system caught (31 ÷ 35) | 89% |
The six false alarms point to two fixes: wait for a clear frame, and add a "face up" task type for product pushed back. The three duplicates mean the repeat rule needs tightening. The four misses were all on bottom shelves, so that's where a closer camera might earn its cost.
Remote grocery store monitoring across several stores
Once each store is running, the owner's view changes. You stop looking at single gaps and start looking at patterns:
- Alert-to-refill time by store and shift. If a store clears gaps fast on days and slowly on evenings, look at evening staffing.
- Repeat items. The same products going empty on the same days point back to ordering, which is the 47 percent from the FMI study.
- Areas with the most open time. Where the shelf sits empty longest, that's where sales are lost.
Focal offers a remote "Virtual Store Inspection" so head office can look at aisles in any store. A simpler version works on existing cameras too: a daily summary per store, with a snapshot attached to each gap that stayed open too long.
In our grocery case study, a demo build on sample data, a camera view flags an empty facing and messages the right aisle manager, who can send a merchandiser or mark it done. Time from alert to refill shows across every store on one screen.
How to start
- Pick one store and three to five areas, such as dairy, beverage coolers and your best endcap.
- Audit the camera views and fix angles before adding hardware.
- Write the routing table.
- Run it next to a manual walk for two weeks.
- Decide area by area. Add a shelf camera only where the test shows a gap worth closing.
Where Trilix fits
We build camera analytics on the cameras most stores already have, with alerts routed to named people and a daily summary for owners. For grocery, that means gap alerts on the areas your cameras can see, alert-to-refill time by store, and an honest list of the areas where a camera view can't support a reliable read. If you run a few grocery stores and want to talk it through, get in touch.
Want to see this on your stores?
We’ll look at your calls, bays and tickets with you and show what we’d build first.