Stockouts cost the retail industry an estimated $1 trillion in lost sales every year. A large share of that is product that made it to the store but never reached the shelf.
Most CPG brands track on-shelf availability (OSA) using inventory system data or periodic manual audits. Both report what stock exists somewhere in the store, but neither tells a category manager whether that stock is on the shelf, in the right position, right now—which is the only measurement that matters to a shopper.
On-shelf availability (OSA) is the percentage of a brand's listed SKUs that are physically present on the shelf, in the correct position, at the moment a shopper arrives—measured from the shopper's perspective, not the inventory system's.
That definition draws a hard line between three different things that brands often treat as equivalent:
|
Availability type |
What it actually measures |
|
Supply chain availability |
Stock exists somewhere in the supply chain |
|
In-store availability |
Stock is physically inside the building |
|
On-shelf availability (true OSA) |
Stock is on the shelf, in position, shoppable right now |
On-shelf availability measures the third layer only.
A product sitting in the backroom, a product in the wrong bay, or one placed at floor level when the planogram puts it at eye level, they all fail on-shelf availability, even if the inventory system says the SKU is available.
That distinction is key because the consequences of an on-shelf availability failure are lost sale.
The answer isn't 'they wait for a restock. Research consistently shows:
That 70% brand-switch rate is why poor on-shelf availability isn't just a compliance metric. For an established brand, on-shelf availability is a direct revenue leak.
For a brand still building awareness, it's worse—a shopper who never found you in the first place has no reason to come back.
|
"At a previous CPG brand I worked at, a third of our first-time consumers only engaged with the brand because they saw it on a display in a store. Off-shelf execution was directly driving trials—not advertising, not pricing. If the display wasn't there, that first interaction never happened." — Steven Bussiere, VP Customer Success, Vision Group |
Understanding what on-shelf availability measures and why it matters is step one. The next question is how to calculate it—and where most teams get the number wrong before they've even started.
Example: A brand has agreed 10 must-have SKUs with a grocery banner. On a field visit, 7 are present and in the right shelf position. OSA = 70%. The industry target is 98%+. That 28-point gap costs revenue in every store where it exists, every day it persists.
Home shelf OSA only counts availability at the primary display position. In-store OSA includes secondary placements—endcaps, promotional displays, checkout areas.
A brand running a promotional display campaign needs in-store OSA. Measuring only the home shelf misses every display execution gap.
Some brands count an interchangeable SKU as available when the primary is out—the same beer in a bottle if the can is missing. Others require the exact SKU. This choice changes the reported number significantly and must be agreed with the retailer before targets are set.
If the numerator in your on-shelf availability formula comes from your inventory management system, you're measuring whether stock is in the building—not whether it's on the shelf. The result is a reported OSA that consistently overstates actual availability.
This is the phantom inventory problem.
|
What is phantom inventory? Phantom inventory is stock that the inventory system records as 'in stock' but that a shopper can't find on the shelf. It's in the backroom, placed in the wrong bay, or left over from a reset that was never fully executed. The inventory system and the shelf tell two different stories—and your on-shelf availability score reflects the system's story, not the shopper's. |
Once you're calculating on-shelf availability from actual shelf data rather than inventory records, the next question is what number you're aiming for—and how realistic that target is given where most brands actually operate.
ECR (Efficient Consumer Response) research sets 98%+ as the threshold below which stockout frequency creates measurable shopper switching. At 98%, a brand provides reliable shelf availability. Below it, gaps appear often enough that shoppers start encountering them—and switching when they do.
Average grocery on-shelf availability sits between 92–95% in practice. During promotional periods—the exact windows when trade investment is active and conversion is highest—it frequently drops to 85–90%.
A Quri study found on-shelf availability dropped by up to 50 percentage points in the days immediately following major promotional events. Brands that funded the promotion, negotiated the shelf space, and activated the campaign were seeing availability scores in the 50s at the moment the campaign was designed to convert.
|
OSA change |
Sales impact |
|
3% increase in OSA |
≈ 1% increase in sales for a CPG manufacturer |
|
2% increase in OSA |
≈ 1% increase in sales for a retailer |
|
2-digit drop in OSA |
≈ 1-digit drop in sales |
|
OSA consistently below 80% |
Significant, persistent revenue loss across the store network |
For a flagship SKU doing $50,000 in weekly sales at a single high-traffic grocery account, a 10-point OSA gap is $5,000 per week in that one store. Across 500 stores, closing that gap becomes a significant annual recovery—which is why field execution directors at large CPG brands treat on-shelf availability improvement as a commercial priority, not a compliance program.
|
"If I look at on-shelf availability as a use case and I'm a warehouse brand relying on a third-party collector going in once a month—if my OSA is regularly in the 70–80% range, that's a problem. I need to be north of 90% every time an auditor takes a picture of that shelf, so I know I'm present and shoppable from the consumer's perspective." — Steven Bussiere, VP Customer Success, Vision Group |
The gap between a 98% target and a 92–95% reality is too consistent across brands and retailers to be explained by supply chain variance alone. There are specific, repeatable causes—and half of them happen inside the store, after the product has already arrived.
On-shelf availability failures come from two distinct layers: upstream supply chain causes and in-store execution causes.
Most CPG brands have invested in systems to detect and fix the first layer, but most have very little visibility into the second—which is where a large proportion of failures actually occur.
|
"When you're a warehouse brand relying on scan data, you're working from a projection based on maybe 10% of the market. Syndicated data can't tell you what's actually on the shelf right now across thousands of stores. That's where the measurement gap lives—and that's where image recognition can be transformative. It gives you eyes and ears at retail at any given moment, so you can confirm your merchandising is actually in place and your trade investment is validated." — Steven Bussiere, VP Customer Success, Vision Group (formerly Nestlé Waters, Body Armor) |
Notice what causes 4, 5, and 6 have in common: none of them show up in an inventory system, and none of them are reliably caught by a field rep doing a visual pass during a store visit. That's not a coincidence—it's a measurement problem. The methods most CPG brands use to track OSA are designed to catch supply chain failures, not execution failures.
A field rep physically checks each shelf position during a store visit. The most direct method available to CPG brand teams—but with three structural limits that make it unreliable at scale:
Uses sales velocity and inventory records to infer whether a product should be on shelf. Faster and broader than manual audits—but measures supply chain performance, not shelf reality.
Phantom inventory makes this structurally unreliable for on-shelf availability specifically. A product in the backroom shows as in-stock. A product placed in the wrong bay shows as available. Both register as good on-shelf availability in the system while a shopper looks at an empty or wrong shelf position.
The system's OSA and the shopper's OSA are two different numbers, and the gap between them is exactly the execution-layer failure that most brands aren't measuring.
Weight sensors or RFID tags embedded in shelving units detect product removal in real time and trigger restocking alerts. The most accurate continuous monitoring method available—and largely inaccessible to CPG brand teams.
To deploy sensors, the retailer has to install and maintain hardware on their shelves. A CPG brand managing execution across Kroger, Walmart, and Target cannot install IoT infrastructure in those stores. This method is viable for retailers managing their own estate. For CPG brands auditing retail partners, it's not an option.
A field rep photographs a shelf section during a store visit. The AI identifies every visible SKU, maps each product to its shelf position, counts facings, reads price tags, and compares the full picture against the planogram for that specific store.
Within 90 seconds, the rep receives an on-shelf availability score and a prioritized gap list on their phone—before they leave the aisle.
This is the only commonly available method that captures both whether a product is present and whether it's in the correct position, from the shopper's perspective, during the visit itself. It operates at the execution layer—the exact place where manual audits and inventory systems consistently fail.
It doesn't require retailer hardware. It runs on the device a rep already carries. And it catches planogram non-compliance and backroom-to-shelf failures—the execution causes that POS data and inventory systems are structurally blind to.
With a clear picture of where on-shelf availability gaps come from and why they're hard to see, the question becomes practical: what does a category manager or field execution director actually change to move the number?
A forecasting fix won't help a backroom-to-shelf problem. A field execution fix won't help a replenishment lag. The first step is diagnosing which layer is generating the failures.
Pull a sample of your lowest-OSA stores and check whether they have normal delivery frequency. If delivery is on schedule but on-shelf availability is still low, the gap is execution—product is arriving but not reaching the shelf correctly. That determines whether you invest upstream or downstream.
If your on-shelf availability number comes from your WMS or inventory system, you're measuring phantom inventory as availability. Shelf-level verification—through image recognition or physical audits—gives the ground-truth read that inventory data can't.
Start with your highest-velocity SKUs. Getting an accurate shelf-level on-shelf availability on your top 20 SKUs is more valuable than an estimated number across your full catalog, because that's where the revenue impact of any gap is largest.
The top 20% of problem stores and SKUs typically drive more than 80% of on-shelf availability failures. Rank stores by OSA score. Rank SKUs by the product of revenue velocity multiplied by on-shelf availability gap. Work from the top of both lists first.
Uniform improvement across a 5,000-store network costs significantly more and produces less revenue recovery than concentrated improvement in the 200 stores where your hero SKUs are consistently failing.
When on-shelf availability measurement runs as a separate audit program on its own schedule, it happens infrequently and the data is already outdated by the time it reaches a field manager. When a rep captures shelf data as part of every standard visit—through a 60-second photo—the measurement cadence matches the visit cadence, and corrections can happen immediately rather than in a follow-up cycle days later.
A trade marketing manager's most common on-shelf availability failure is a display that's correct on setup day and gone by day four, with no one catching it until the post-campaign review, after the budget has been spent.
Mid-campaign on-shelf availability tracking is where trade investment gets protected. A promotional display that fails on Thursday doesn't need a new visit next week—it needs a flag on Thursday, when a rep can still fix it within the campaign window.
An on-shelf availability gap found during a store visit can be corrected in minutes. The same gap routed to a reporting dashboard for a manager to review on Thursday requires a new store visit to fix—and costs the brand days of lost sales on that SKU in the meantime.
For a field execution director, this is the lever with the most immediate revenue impact. Getting gap data to the rep's phone before they leave the aisle means same-visit correction. Getting it to a dashboard means next-visit correction. That's a 5–7 day difference in correction timing on every gap found.
Of the six levers above, the last one—same-visit correction—has the most direct impact on recovered revenue because it eliminates the correction lag entirely. But it requires a measurement method that gets gap data to the rep during the visit, not after it. That's what makes image recognition the enabling technology for this lever specifically.
Most on-shelf availability programs find gaps after the visit and fix them on the next visit. Image recognition changes the timing so that detection and correction happen during the same visit.
|
Without image recognition |
With image recognition |
|
Rep visits Monday, submits report |
Rep photographs shelf section—60 seconds |
|
Manager reviews report on Wednesday |
AI returns OSA score and gap list—90 seconds |
|
Correction visit scheduled for Friday |
Rep corrects gaps before leaving the aisle |
|
Shelf wrong for 5–7 days minimum |
Shelf corrected during the same visit |
For a high-velocity SKU, that 5–7 day correction lag is the difference between a minor compliance note and a material revenue loss. Closing it doesn't require more store visits—it requires a different method of capturing and acting on shelf data during the visits that already happen.
Without image recognition technology, a rep covering 10 stores per day spends significant time on manual shelf analysis—checking positions, counting facings, comparing against a planogram they're holding in their head or on a printed sheet. That's time spent on work the system can do more accurately.
With image recognition, the rep takes one photo per shelf section. The system does the SKU-level comparison. The rep receives a prioritized list showing which gaps have the highest commercial impact and spends their time on what requires judgment: fixing what's wrong, managing the store manager conversation, escalating what they can't resolve in that visit. The visit produces more corrections per hour of field time.
|
"Image recognition gives you line of sight—whether you're a DSD manufacturer or a warehouse brand—on what your in-stock or out-of-stock ratio actually is on a frequency basis. And when you think about the investment at retail, whether it's promotional frequency, promotional vehicles, or placements, you want to ensure the return on those investments. That's what real-time shelf data gives you." — Steven Bussiere, Vision Group, from a retail execution webinar |
Image recognition closes the correction loop. But the platforms that do it differ significantly on three points that determine whether you can actually deploy it and get on-shelf availability data across your full store network.
That's where Store360's design choices matter.
Most IR platforms measure compliance. Store360 is built specifically to close the correction loop during the visit—and to generate OSA data across a full store network, not just the stores where planogram files are complete and product data has been loaded.
Three design differences that matter for on-shelf availability specifically:
1. No product data handoff before going live
Most IR platforms require the brand to supply product images, dimensions, and UPC codes before the model can recognize their SKUs. That onboarding process takes 8–16 weeks—which means 8–16 weeks of on-shelf availability data the brand simply doesn't have.
Store360 runs on a pre-trained library of over 1.3 million SKUs across CPG categories. For most clients, their products are already in the library before deployment starts. Most clients go live in under 30 days.
2. On-shelf availability data even when there's no planogram on file
Most IR tools require an official planogram to benchmark against. No planogram file means no compliance score and no OSA data for that store—a structural blind spot in any network where planogram coverage is incomplete or outdated.
Store360 benchmarks shelf presence against category norms and competitor positions even without a planogram. Every store in the network generates actionable on-shelf availability data, regardless of planogram file coverage.
3. The gap list reaches the rep before they leave the aisle
Store360 returns an on-shelf availability score, a gap list prioritized by commercial impact, and a correction task to the rep's phone within 90 seconds of the shelf photo. The rep corrects the shelf before moving to the next section. No dashboard review later. No follow-up visit.
Success story:
L'Oréal deployed Store360 at Walmart locations where out-of-stocks were a persistent problem. Their reps had been working from audit data that was 2–4 weeks old. With Store360, reps saw exactly which SKUs were trending toward out-of-stock during each visit and had the data to drive store manager action on the spot.
Result: $50,000+ in replenishment orders across 10 Walmart stores in two weeks.
"Inventory levels are going back up, sales are going back up on these out-of-stock items, and it's really moving the needle."
— Barbara Kline, Head of Retail Execution, L'Oréal
Store360 is live in 55+ countries. It runs on the device a field rep already carries—no new hardware, no retailer permission required to deploy. Trusted by teams at Coca-Cola, Nestlé, L'Oréal, Mars, and Revlon.
→ Book a 20-minute walkthrough: shelf photo to OSA score to correction task list in under 90 seconds.
On-shelf availability is the percentage of a brand's listed SKUs that are physically present on the shelf, in the correct position, when a shopper arrives. It measures what the shopper actually encounters—not what the inventory system believes is in stock.
OSA (%) = (SKUs on shelf in correct position ÷ Total listed SKUs) × 100. A brand with 10 agreed SKUs in a store, 7 of which are present and in position, has a 70% on-shelf availability. The industry target is 98%+.
98%+ is the standard target. Average grocery on-shelf availability sits between 92–95% in practice and drops to 85–90% during promotional periods. Any brand consistently below 90% is losing sales daily across its store network.
They describe the same gap from opposite directions. If 7 of 10 SKUs are on shelf, on-shelf availability is 70% and OOS rate is 30%. The distinction: OOS technically includes product missing from both shelf and backroom. OSA is specifically a shelf metric—product in the backroom but not on the shelf fails on-shelf availability even if it's not technically out-of-stock in the inventory system.
Supply chain causes: inaccurate demand forecasting, replenishment lag, and promotional volume spikes the supply plan didn't model. In-store execution causes: backroom stock not reaching the shelf (the largest single cause in many networks), planogram non-compliance where product is present but in the wrong position, and shelf reset gaps where new planograms aren't fully implemented.
Phantom inventory is stock the inventory system records as 'in stock' that a shopper can't find on the shelf—it's in the backroom, in the wrong location, or misplaced after a reset. It's the primary reason inventory-based on-shelf availability overstates actual availability. Brands measuring on-shelf availability from their WMS are counting phantom inventory as availability.
Home shelf OSA tracks availability only at the product's primary display position. In-store OSA includes secondary placements—endcaps, promotional displays, checkout positions. A brand running a promotional display campaign needs in-store OSA; tracking only the home shelf misses every display gap.
A trade promotion drives a 30–40% volume lift the replenishment plan wasn't sized for. Stock depletes faster than the normal replenishment cycle replaces it. At the same time, promotional displays add execution complexity—builds that deviate from the brief, displays moved by store staff, missing POS materials. OSA falls at the exact moment conversion is highest.
A rep photographs a shelf section; AI returns an on-shelf availability score and gap list to their phone within 90 seconds. The rep corrects gaps before leaving the aisle. The correction loop closes during the visit instead of 5–7 days later after a manager reviews a report and schedules a follow-up. For high-velocity SKUs, that timing difference is a material revenue recovery.
By changing what a rep does during each existing visit. Image recognition replaces the manual shelf-checking portion—scanning positions, counting facings, comparing against a planogram—with a 60-second photo. The rep's time shifts from analysis to correction. The same visit cadence produces more corrections per visit.
Most IR tools can't—no planogram means no benchmark, which means no data for that store. Platforms like Store360 benchmark against category norms and competitor positions even without a planogram file, so every store generates actionable on-shelf availability data regardless of planogram coverage.
Research shows around 70% switch brands and around 30% switch stores. Very few wait for a restock. For a brand building trial, that brand-switch is often permanent—a shopper who finds a competitor's product in your empty shelf position has no immediate reason to seek yours out on the next trip.
Inventory system data tells you whether stock is in the building. Manual audits tell you what one rep saw at one moment during one visit. Neither one tells a category manager or field execution director whether a shopper arriving right now will find their product in the right position on the shelf.
That's the execution-layer gap—the distance between stock arriving at the store and product landing in position, correctly placed and visible. It's where the majority of on-shelf availability failures in a well-run CPG network actually occur, and it's the layer that traditional measurement methods are structurally blind to.
Closing that gap requires measuring on-shelf availability at the shelf, during the store visit, and getting the correction into the rep's hands before they walk out the door. That's what image recognition makes possible.
→ Book a walkthrough of Vision Group's Store360 here.