One photo, three answers.
The capabilities, stated plainly.
Every row ships today and traces to a capability described on this page.
| Capability | xMatix Sense Vision |
|---|---|
| Share-of-shelf measurement from a photo | Native |
| Planogram compliance scoring | Native |
| Planograms per outlet, class or tenant | Configurable |
| Capture inside the offline visit flow | Offline |
| New SKU from reference photos, no retraining | Native |
| Manager nudges on out-of-line shelves | Native |
| Model version and thresholds stored per result | Native |
| Self-hosted vision pipeline | Native |
| Metered on the same Sense credits | Native |
The questions evaluators ask.
What does Sense Vision measure from a shelf photo?
Three things: share of shelf, facing counts and planogram compliance. From one photograph it detects the products on the shelf, counts your facings against competitors', and compares the arrangement to the planogram that applies at that outlet — including shelf-band placement, so a product sitting below eye level is flagged, not just a product that is missing. The result is a score, not an opinion.
How is share of shelf calculated?
From detected facings — your brand's facings as a share of all facings visible in the photo, measured by the vision pipeline and never eyeballed by a language model. The same detection produces the facing count and feeds compliance scoring, and because the measurement is mechanical, two photos of the same shelf produce the same number regardless of who took them.
What is planogram compliance and how is it scored?
Planogram compliance is expected versus actual, shelf by shelf. You define the layout an outlet should carry — scoped per outlet, per outlet class or tenant-wide — and the score compares expected facings and shelf bands against what the photo shows. The misses are named specifically: off the eye-level band, two facings short — so the fix is actionable, not a general scolding.
Does it work when the rep is offline?
Yes — shelf capture is a merchandising activity inside the same offline visit flow reps already use, so the photo is taken with no signal at all. Measurement runs on sync: when the device reconnects, the pipeline scores the shelf, writes the result, and nudges the manager if the shelf is out of line. No new app, no new habit.
What happens with a product it has not seen before?
You teach it with photos, not a project. Brand identity comes from a reference-image gallery — add five to ten photos of the new SKU, reindex, and it is recognized, typically within a day. There is no model retraining and no vendor ticket, which matters in categories where the pack changes every season.
Can results be audited?
Yes — every result stores the model version and the thresholds that produced it, permanently. When a score is challenged months later, you can state exactly what measured it and by what rules. The vision pipeline is self-hosted, so your shelf images stay on your platform, and usage is metered on the same credits and budgets as the rest of Sense.