Retail

Computer Vision for Pharmacy Shelves: Compliance and Restricted Products

September 2026 · ISTRALLEN Team

Shelf monitoring with a regulatory edge

A pharmacy has the same availability problem as any retailer — front-of-store health, beauty, and over-the-counter lines sell out between checks. It also has categories where what is on the shelf, and where, is a compliance matter. Computer vision for pharmacy retail can help with the first and support the second, within clear limits. None of this is legal advice.

Availability monitoring is mechanical and runs on every open shelf: a confidence check leads straight to an alert for staff, issued the same way whether or not the bay is also being watched for compliance; a low-confidence detection doesn't trigger an alert, it goes to a review queue instead, rather than sending someone to check a shelf that's probably fine. Underneath, the same open shelves feed one more path, but only for regulated categories: a planogram compliance check — placement and signage rules — runs continuously and flags a deviation for the store manager to correct before the next audit; availability alerts keep running regardless, since a compliance flag never blocks them. A separate set of areas — behind the counter, controlled substances, the dispensing area — never becomes a vision target in the first place, so no detection event is ever generated there; those areas route to inventory controls instead. The camera's job is to watch the open shelf fast and flag precisely what needs a person or a different system — never to reach into zones it has no business watching.

Open-shelf availability is the safe core

For general merchandise and open-shelf over-the-counter products, this is ordinary shelf monitoring: a single-shot detector on edge hardware spots gaps and misplacements, and confidence-based routing sends only high-confidence detections to staff as alerts. The approach is the same as our computer vision project; the pharmacy context just changes which shelves matter.

Planogram compliance for regulated categories

Some categories have rules about placement — what sits within reach, what carries required information, what must not be adjacent to something else. Vision can check the actual shelf against the intended planogram and flag deviations for a person to correct, which is useful precisely because these are the placements a compliance auditor looks at.

What the camera should not be watching

Behind-the-counter stock, controlled substances, and the dispensing area are not open-shelf monitoring targets. They are not reliably camera-visible, and pointing a monitoring camera at them raises privacy and security questions that outweigh any availability benefit. Those areas need inventory controls, not vision.

Age-restricted and quantity-limited products

Where these sit on open shelves, vision can confirm they are in the correct location and that any required signage is present. It cannot enforce a sale restriction — that is a point-of-sale control — so the value is placement and presence, not transaction compliance.

Privacy by architecture

The pharmacy floor has customers on it, some in a health context. The design that keeps this defensible is the one where inference runs on-device and only a structured detection event — product, gap, position — leaves the store, never the image. Data minimisation is built into the architecture rather than bolted on.

A worked example

A weekly planogram check flags that a regulated category has drifted one bay left during a reset, putting a required information leaflet out of its designated position. The alert goes to the store manager with the expected layout, and it is corrected before the next audit. Separately, the camera detects a gap in a high-demand allergy product on the open shelf, and a high-confidence alert prompts a same-day refill from the back room. A third detection — a possible gap partly hidden by a promotional stand — comes back at low confidence and is routed to the review queue rather than sending someone to check a shelf that is probably fine.

Fitting the audit calendar

Regulated-category planogram checks are most useful when they run against the compliance audit schedule rather than at random. Running the vision check a few days before each audit window gives the store time to correct any drift a person confirms is real, which turns the system from a reactive alert stream into a pre-audit checklist. It also keeps the volume manageable — a weekly or pre-audit cadence on regulated bays, continuous monitoring only on the fast-moving open shelves where immediacy actually matters.

Where this stops being right

Anything behind the counter or in dispensing is out of scope for shelf-monitoring vision.
Compliance rules vary by jurisdiction and change; a vision check supports an audit, it does not replace a compliance programme — confirm requirements with counsel and a responsible pharmacist.
Customer-facing cameras carry privacy obligations that must be assessed before deployment, not after.

FAQ

What can computer vision safely monitor in a pharmacy? Open-shelf availability for general merchandise and over-the-counter lines, and planogram compliance for regulated categories on open display.

Can it monitor controlled substances? No. Behind-the-counter and controlled stock is not a shelf-monitoring target — it needs inventory controls, and pointing a monitoring camera there creates more problems than it solves.

How is customer privacy handled? Inference runs on-device and only a structured detection event leaves the store — never the image. The obligations still need a formal assessment before deployment.

ISTRALLEN builds shelf-monitoring computer vision for health retail that respects the behind-the-counter line and minimises data by design — see AI for Retail.

See it in production

AI for Retail → Semantic search case study →

← All articles