OSA is the most direct execution-to-sales link: roughly every 3% of availability equals 1% of sales. The three OSA metrics, the data, and how AI detects gaps.
On-shelf availability (OSA) measures whether the right products are physically on the shelf, in the right quantity and variant, when shoppers look for them. It is the most direct link between execution and revenue in all of consumer goods: across categories, roughly every 3% improvement in OSA translates into 1% of sales growth: because an absent product is an unmade sale, and in most categories an unmade sale is handed straight to a competitor or, worse for the retailer, to another store.
This data guide covers the three levels of OSA precision, why availability gaps stay invisible to conventional systems, what shopper research says happens at an empty facing, how AI detection changes the operating rhythm, the math for your own category, and the implementation pattern that turns availability from a monthly statistic into a daily control loop driving revenue growth.
The three levels of OSA precision
- OSA Shelf: is the product physically present on the shelf at all? The binary version most programs start with.
- OSA Facing: are the agreed number of facings visible? Presence with one lonely facing where five were agreed is technically "available" and commercially eroding.
- OSA SKU: is the exact variant displayed, size, flavor, format? The 500ml present while the 1L is gone reads as availability at shelf level and as a stockout to the shopper who wanted the 1L.
Programs that track only the binary level systematically overstate availability, usually by several points. The facing and SKU levels are where quiet erosion happens, and where the recoverable revenue concentrates, because those gaps never trigger anyone's alarm.
What happens at an empty facing: the shopper's three exits
Decades of shopper research converge on the same uncomfortable arithmetic. Facing a gap, shoppers take one of three exits: substitute (often to a competitor, your loss, the category's wash), postpone (a coin-flip whether the purchase ever happens), or switch stores (the retailer's nightmare and the reason availability is one conversation where brand and retailer incentives genuinely align). The brand share of the loss varies by loyalty and category, but the direction never does: every empty facing taxes someone, every day, silently.
That silence is the strategic point. Availability failures don't announce themselves, no system beeps, no shopper files a complaint. The product simply sells less, and the cause hides inside a hundred plausible explanations until someone measures the shelf.
Why OSA gaps stay invisible
- Sell-out lag: By the time a stockout appears in sell-out data, the damage is weeks old and the cause is cold.
- Sampling: Manual checks catch a sample of SKUs in a sample of stores on a sample of days. The gaps live in the unsampled cells.
- Timing: The gap that opens an hour after the rep leaves runs unmeasured until the next visit.
- Virtual stock: The most insidious case: inventory systems say the product is in the store, and it is, in the back room, while the shelf sits empty. Every dashboard reports availability; every shopper sees a gap.
Detection speed is therefore the whole game. An availability program is not a measurement project; it is a latency-reduction project.
How AI detection changes the rhythm
With image recognition, one shelf photo produces the complete OSA picture instantly: which planned SKUs are present, at how many facings, in which variants, checked against multiple assortment matrices (TOP, must-have, promo) simultaneously, with results in under 20 seconds and real-time detection accuracy around 96–97%.
The operational loop that follows:
- In-visit: the rep receives the exact missing-SKU list with corrective actions, restock from the back room, flag to store staff, log the reason for absence.
- Reason codes: each absence is classified, out of stock, blocked, not on shelf, virtual stock, giving supply chain the data to fix causes rather than symptoms. A store with chronic "virtual stock" codes has a replenishment process problem, not a demand problem.
- Alerts: availability drops on hero SKUs in key accounts notify owners the same day.
- History: SKU-by-visit statistics across the entire usage period, so chronic offenders, stores, SKUs, day-of-week patterns, surface from the data instead of anecdotes.
OSA and the retailer: the one shared KPI
Most execution KPIs put brand and retailer in mild tension, your share of shelf is someone else's, your block is their layout constraint. Availability is the exception: an empty facing costs the brand the sale and costs the retailer the basket, and the store-switching exit costs the retailer most of all. That alignment makes OSA the strongest opening move in collaborative execution programs. Brands that share availability data with retail partners, gaps by store, reason codes, recovery times, consistently report faster escalation paths, joint replenishment fixes, and a different tone in trade reviews, because the conversation starts from a number both sides want to move in the same direction.
The practical artifact: a monthly one-pager per key account showing OSA trend, top recurring gaps with reason codes, and the recovered-sales estimate, positioned as shared upside rather than supplier complaint. Several teams credit that single document with unlocking back-room access and priority replenishment that years of asking never produced.
From detection to prevention: reading the patterns
Once availability history accumulates, the program's second act begins: prediction. Chronic gaps cluster, by day of week (weekend sell-through outrunning Monday replenishment), by promo cycle (feature volume draining shelf stock faster than the plan assumed), by store cluster (small back rooms, single-shift staffing), by SKU (slow movers that fall off the reorder radar). Reading these patterns converts the OSA program from a correction loop into a prevention system: visit schedules shift toward high-risk windows, promo plans carry shelf-stock buffers sized from last cycle's drain rate, and assortment reviews use absence-reason history to separate true demand failures from supply process failures.
This is also where the data earns its seat in S&OP: a rising trend of "virtual stock" codes in one region is a replenishment process alarm; a rising trend of true out-of-stocks on a growing SKU is a forecast revision. The shelf, photographed at every visit, becomes the earliest reliable sensor in the demand chain.
The math for your category
The 3-to-1 ratio makes the business case unusually easy to localize. Take annual category revenue through monitored channels, your measured OSA, and a realistic target:
- A brand doing 50M with OSA at 88% that climbs to 94% is recovering roughly 1M of sales that was silently leaking, every year, compounding with distribution growth.
- Programs that instrument availability typically gain +5 to 15 percentage points, because the starting point is almost always lower than believed: the binary-only measurement and the virtual-stock blind spot both flatter the baseline.
- The cost side moves in the same direction: availability checks via photo add zero field time, and the audit they replace was costing 30–50% of each visit.
A note on measurement integrity: availability numbers feed bonuses and retailer conversations, so they need the same evidence discipline as any commercial KPI, photo quality control to reject blurry or partial shots, GPS-verified visits, and drill-down from every dashboard figure to its source image. An OSA number nobody can dispute changes the meeting; an OSA number built on self-reported checklists restarts the dispute every month.
And a note on scope: start with the must-have matrix, not the full catalog. A 40-SKU must-have list measured at facing level in every store beats a 400-SKU list measured nowhere, and the discipline of choosing the list forces the assortment clarity that half of availability management actually is.
OSA in the traditional channel
In markets like Mexico and Brazil, a large share of volume flows through traditional stores where availability failures are most common and least measured, small back rooms, informal replenishment, distributor-dependent supply. Two capabilities make traditional-trade OSA feasible: offline recognition (photos processed on-device without connectivity, syncing later) and simplified must-have matrices per store type. Brands extending OSA measurement into this channel consistently find their largest availability gaps there, and the cheapest recovered revenue in the portfolio, because the distribution already exists; only the shelf is failing.
Finally, sequence the audience. The field hears "shorter visits, clearer tasks"; supply chain hears "reason codes that separate our failures from the store's"; finance hears "the 3-to-1 bridge"; sales leadership hears "recovered revenue by account." OSA is the rare program with a genuine win for every function, but only if each one is told its own version of the story.
Common mistakes
- Measuring only the binary level: Facing and SKU-level gaps are most of the leak.
- Skipping reason codes: Without causes, every gap becomes the field's fault and nothing structural gets fixed.
- Averages without distributions: Network OSA of 93% with a bottom decile at 78% is a targeting opportunity disguised as a good number.
- Trusting inventory systems: Virtual stock means the system's "yes" requires the shelf's confirmation.
- Treating OSA as a field KPI only: Half the chronic gaps are supply, assortment, or retailer process issues wearing a field costume.
