Every shelf checked against the planogram on the day of the visit
A merchandiser photographs the shelf in the app, and the model recognises your products by their packaging, counts facings and finds empty spots, competitor products and wrong price tags. The merchandiser sees the issues while still in the store, and the supervisor gets a summary by store instead of a folder of photos.
Planograms, packaging photos, rules for price tags and share of shelf. The model learns to recognise your products and the competitor products that matter for the report.
Following the app prompts: the whole shelf, no steep angle, price tags in frame. A blurry or cropped photo is flagged for a retake while the merchandiser is still at the shelf.
It finds products by packaging, counts facings and marks empty spots, competitor products and price tags.
It compares the actual display with the reference: what is out of place, what is missing, where the tag does not match the product or the price.
Look-alike packs, glare on film wrap, a hidden price tag. The model flags such photos, and a supervisor reviews them.
The merchandiser sees the issues before leaving the store and can fix them on the spot. Managers see a summary by retailer, region and supervisor.
One task, three different projects. Pick the one that looks like you and press “I want this” — your request arrives tagged, and we send an estimate in 1–2 days.
One brand or distributor, a small merchandising team and a few dozen stores.
What we build
Several categories and retail chains, each with its own planogram, and the sales team needs display reports every week.
What we build
A retail chain or a large manufacturer: hundreds of stores, an in-house merchandising team and audits done by store staff.
What we build
Pilot from RUB 500,000. Beyond that we price by task and hours.
If the packs look different on a photo, yes, once trained on your range. Near-identical packs in different flavours are the hardest case: we check them by the price tag or send them to a person to confirm.
The model will not recognise new packaging on its own. You send photos, we retrain. Under a support contract this is routine work, not a separate project.
We start with them: during the assessment we check how many are usable. Most often the capture rules need adjusting (angle, distance, lighting), plus a photo quality check built into the app.
Photos can be taken offline and sent once the connection is back. Running the model on the phone itself, with no server, is a separate task we discuss during the assessment.
Name the process that eats time. You get a pilot range, not a 40-slide deck.