What is OSA (on-shelf availability) and why most chains measure it wrong
OSA, on-shelf availability, is the percentage of SKUs the customer finds on the shelf at the moment of the purchase decision. It is not the same as shelf-out. OSA is the product’s presence; shelf-out is its absence.
It is not the same as service level either. Service level measures whether the distribution center delivered. OSA measures whether the product reached the shelf and is there, visible, now. Stock in the warehouse and an empty shelf? High service level, low OSA.
How the operation measures it
There are two ways, and they disagree. One is a physical audit: someone walks the store and counts what is on the shelf. Expensive, slow, true.
The other is to derive it from the system: if ERP stock is greater than zero, it counts as available. Cheap, fast, and frequently false. Most chains measure OSA this way, and this is exactly where the number lies.
What the number hides
The system says there are three units. On the shelf, zero. It could be unrecorded theft, damage, an item in the back but not restocked, a count error. That is phantom inventory, and it inflates system-measured OSA.
The result: the dashboard shows 97% OSA and the customer, standing in front of the empty shelf, sees 88%. The difference is lost sales nobody books, because the system swore the product was there.
Why this changes an AI project
If you are about to install computer vision or sensors to measure OSA, the first question is ground truth. Training the system against ERP stock is teaching the machine to repeat the lie.
OSA measured at the shelf, with physical truth, becomes the acceptance criterion of any replenishment or loss-prevention project. Measured from the ERP, it becomes dashboard theater. It is the same problem as phantom inventory and a cousin of shrinkage.
Think about the term each area of your operation uses but measures differently. “Availability” is usually that term: logistics measures it at the DC, the store measures it at the shelf, and the two report numbers that do not talk. Tell me yours. In one hour I will send back a one-page card: the formula, the measurement points, what the number actually means, and where it misleads. If the card shows the way you measure OSA has to be rebuilt, the two-week Diagnóstico is the next step.