Is the expected range visible?
Compare observed SKUs with the intended assortment. Flag products that were not observed so teams can check shelf coverage and investigate the gap.
AI-POWERED RETAIL MEASUREMENT
SKU counting software.
Know what’s
on the shelf.
Move from manual product tallies to evidence your team can use. Aislemetry applies AI to shelf photos to identify visible SKUs, count their facings, and help teams investigate assortment gaps.
For retail brands, category teams, and field audit programs.
PHOTO → PRODUCT RECORDILLUSTRATIVE
VISIBLE FACINGS05
IDENTIFIED SKUs02
Harvest Original · 400 g4 facings
Harvest Honey · 400 g1 facing
Fictional products and sample counts. Diagram of the intended output, not a live AI analysis.
QUICK ANSWER
What is AI SKU counting?
AI SKU counting identifies distinct products in shelf photos and counts their visible front-facing packs. Aislemetry detects packs, matches packaging to a reference catalogue, and groups counts by product, variant, and size. Teams review uncertain matches before using the results. Photo coverage and image quality limit what can be confirmed; hidden inventory is outside the count.
- Inputs
- Shelf photos, category scope, and a product catalogue with pack images.
- AI process
- Detect visible packs, match product identities, and group facings by SKU.
- Outputs
- Identified SKUs, visible facing counts, and uncertain matches for review.
THE USE CASE, EXPLAINED
What is SKU counting
in a retail audit?
SKU counting identifies how many distinct products are visible within a defined shelf or category. A stock keeping unit, or SKU, represents a particular product, variant, and pack size. A 400 g original cereal and a 400 g honey cereal are two different SKUs.
Facing counts answer a related question: how many front-facing packs of each SKU can you see? Retail teams need both measures to understand the breadth of assortment and the prominence of each product.
ONE PRODUCT ≠ ONE FACING4 identical packs
1 unique SKU4 visible facings
A different flavour or pack size adds a separate SKU. Stock hidden behind the front row is outside the visible-facing count.
THE PROBLEM ON THE GROUND
A shelf photo is useful.
A product-level answer is better.
Manual shelf counting means recognizing similar packs, recording variants, and reconciling results across visits. Small packs and repeated designs make that work harder.
Which products have more facings?
Count repeated packs by SKU to understand how shelf presence is distributed across products, stores, and visits.
Can the team check the result?
Keep product identities and counts linked to the source photo so disputed or uncertain observations can be reviewed.
HOW AISLEMETRY USES AI
From shelf image
to structured SKU counts.
Aislemetry’s Visual Intelligence is designed to turn visible shelf evidence into product observations. Exact SKU identification depends on reference data and what the image actually shows.
Define the catalogue and audit scope
Set the category, stores, and expected range. Reference product names, variants, pack sizes, and pack images give AI a catalogue to match against.
Detect visible product facings
AI locates visible packs in shelf photos. Detection establishes where products appear; it does not, on its own, establish their exact SKU identity.
Match products and group counts
AI compares visible packaging cues with reference products, groups matching facings by SKU, and separates distinct product variants and sizes where the evidence supports it.
Review uncertain identities and act
Unclear matches remain available for review. Teams can check the linked photo, confirm identities, and compare the observed assortment with the expected range before deciding on follow-up.
Built around evidence, with room for review.
Blur, glare, occlusion, similar packaging, and small packs affect what can be identified. A pilot should validate counting and exact SKU matching on representative shelves, with small packs assessed separately.
WHAT THE OUTPUT CAN LOOK LIKE
One record per SKU.
The evidence stays attached.
ILLUSTRATIVE OUTPUT · FICTIONAL PRODUCTS · NOT CUSTOMER RESULTS
| Product / SKU | Visible facings | Identity status |
|---|---|---|
| Harvest Original · 400 g | 4 | Identified in sample |
| Harvest Honey · 400 g | 1 | Identified in sample |
| Unresolved product group | 2 | Needs review |
This wider example includes two additional unresolved facings: seven visible facings in total, two identified SKUs, and one product group requiring review. Do not add an unresolved group to the confirmed SKU count.
Explore the interactive shelf example ↗SKU COUNTING FAQ
Questions from
retail teams.
What is the difference between SKU counting and facing counting?
SKU counting measures distinct product identities. Facing counting measures the visible front-facing packs of those products. Five identical packs represent one unique SKU and five visible facings.
Can AI distinguish similar variants and small packs?
AI can support variant identification when visible packaging cues and reference products are sufficient. Similar variants, small text, and small packs can be difficult to resolve. Image quality and catalogue coverage matter, and ambiguous matches need review. A pilot establishes performance for your products and capture conditions.
Does SKU counting from a photo measure total inventory?
No. It measures the visible products within the captured shelf scope. It cannot establish the quantity behind the front row, backroom stock, or products outside the image.
Does a missing SKU mean the store is out of stock?
No. A product that is not observed is a gap to investigate, rather than proof of a stockout. The photo may not cover its location, or the product may be obscured. Review the evidence and store context before drawing a conclusion.
What should we provide for a SKU counting pilot?
Start with representative shelf photos and a reference catalogue containing product names, variants, pack sizes, and pack images. Agree how the shelves will be manually labelled, then assess facing counts, exact SKU identification, and the share requiring review.
Reviewed · Aislemetry