An empty shelf = a sale that went to a competitor
Picture a shopper who came in for your yogurt. It's not on the shelf. What do they do? There are only a few options, and none of them are good for you: grab a competitor's product, go to another store, postpone the purchase β or not buy at all.
The classic Corsten & Gruen (ECR) study, still an industry benchmark, found that on average around 70% of shoppers "punish" the brand when their product is missing β they switch to a competitor, change stores, or buy nothing. Only a minority come back later specifically for your product.
average out-of-stock rate in retail
That means roughly every 13th item on a shopper's list is missing at the moment of purchase. During promotions it climbs to 12β15% β exactly when you're paying for traffic and advertising.
The core trap is that this loss is invisible. There's no line in your sales report that says "not sold because the shelf was empty." You only see what sold β never what would have sold if the product had been there. That makes out-of-stock the single biggest loss item almost nobody measures.
How much OOS you actually have (and why you can't see it)
Ask your team: "What's our on-shelf availability?" The most common answers:
Manual OOS control fails on three fronts. First, delay: if a merchandiser visits once a week, an item can be missing for six days before you know. Second, subjectivity: a person eyeballs "yes / no" with no recorded proof. Third, no data: even when a problem is spotted, it rarely enters your systems in an analyzable form β where, when, which SKUs, how often.
A product can be in the store's back room but not on the shelf. It can be in the accounting system but physically gone (mis-picks, theft, spoilage). It can be in the wrong place. On-shelf availability (OSA) and inventory records are two different worlds β and the shopper lives only in the first one.
OOS and OSA: definitions and formulas
To manage a metric, you first need to define it unambiguously. Two key terms:
OSA β On-Shelf Availability
The share of checks in which the product was physically present on the shelf and available to the shopper. It's a "positive" metric β higher is better.
OOS β Out-of-Stock
The mirror metric: the share of cases where the product was missing from the shelf. In essence OOS = 100% β OSA. A "negative" metric β lower is better.
The key detail: measure at the level of a specific SKU in a specific store, not "on average across the chain." An average OSA of 95% can hide a situation where your top SKU is missing in 20% of your premium stores β and that's your main cash flow.
The 4 real causes of out-of-stock
An empty shelf almost never happens because "the product doesn't exist." The causes are almost always operational β and, crucially, manageable:
1. The shelf isn't replenished (even though the product is in the store)
The most common cause. The product sits in the store's back room, but staff didn't put it out in time. Industry research suggests a large share of OOS is a merchandising problem, not a supply problem. It's solved with merchandising discipline and control.
2. Ordering error
Ordered too little, ordered the wrong thing, ignored promo or seasonality. Especially painful during promotions, when demand spikes several times over but the order stays "as usual."
3. Supply / logistics failure
Delayed delivery, short shipment, product stuck at the DC. It's a real cause, but it usually explains a smaller share of cases than people assume.
4. Phantom inventory
The sneakiest β the system confidently shows availability, but the shelf is empty. More on this below.
Phantom inventory β the sneakiest cause
Phantom inventory is when the accounting system is sure the product is there, but physically it's not on the shelf. Causes: mis-picks at receiving, theft, spoilage, POS scanning errors, product "lost" in the store's back room.
The danger is that the system won't automatically reorder something it thinks it "already has." This creates a vicious loop: the shelf is empty β the system thinks stock exists β no new order is placed β the shelf stays empty for weeks. And there's only one way to detect it β physically look at the shelf.
No accounting system sees the real shelf β it only sees transactions. The gap between "data in the system" and "fact on the shelf" is closed only by regularly recording that fact in-store. Without it, phantom OOS can live for months.
How photo monitoring and Vision AI catch empty shelves
The only reliable way to know your true OSA is to record the shelf's state in-store, regularly. This used to mean manual checklists: a merchandiser walks the shelf and ticks boxes. Slow, subjective, no proof.
The modern approach is a shelf photo + image recognition (Vision AI). The merchandiser takes one photo of the shelf on a smartphone. The system then automatically:
Instead of "the merchandiser said it's fine," you get structured data for every store, every SKU, every visit. And you can react not at month-end, but the same day β while the sale can still be recovered.
OSA stops being a "feeling" and becomes a managed number. You can see which SKUs disappear from which shelves most often, why, and whether the merchandiser fixed it during the visit. That's the foundation of managing availability instead of reacting to complaints.
How much you can win back: the math
Let's run a simple example. Say your annual sales in a controlled channel are $2,000,000 and your average OOS is 8%.
Even if systematic shelf control halves OOS β from 8% to 4% β you recover roughly $80,000 a year. And you spend nothing extra on traffic or advertising: the product is simply there, where the shopper is already ready to buy it.
the cheapest growth is removing your own empty shelves
Acquiring a new shopper costs money. Selling to someone already standing at the shelf, intending to buy, costs only availability. Cutting OOS is growth with no marketing budget.
How to start measuring OSA: step by step
Step 1: Define your control list of SKUs and stores
Don't try to control everything at once. Start with the top 20% of SKUs that drive 80% of revenue, and with your key stores. That's where OOS costs the most.
Step 2: Introduce a single recording standard
Every visit = a shelf photo with time and GPS. Not "by eye," but a recorded fact. This removes subjectivity and gives you a base for analytics.
Step 3: Measure OSA by SKU and store, not "on average"
Averages hide problems. Look at exactly where and on which items you're failing β that's where the biggest recovery potential is.
Step 4: Close the "detect β fix β verify" loop
A detected OOS must turn into action: replenish the shelf, adjust the order, deal with phantom inventory. And the next visit should confirm the problem is resolved.
Out-of-stock isn't "force majeure" β it's a manageable operational metric. The moment you start measuring it honestly (photo + Vision AI, not tick-boxes), it almost always turns out higher than you thought β which is exactly why it hides the easiest resource for sales growth.
An empty shelf doesn't shout for attention in your reports. It just quietly hands your sales to a competitor, every single day. The only way to stop it is to start seeing the shelf the way the shopper sees it.