Retail Intelligence
From Shelf to Category Structure
Products, brands, claims and categories, read across retail channels.
Retail · August 2026
A shelf is a completed argument about what a category believes shoppers want. Structuring products, SKUs, ingredient lists and on-pack claims turns that argument into comparable data — and makes it possible to see which positions are crowded and which are merely assumed.
01Key findings
01
InferredClaims are a competitive map
On-pack claims are structured as their own layer rather than treated as marketing text. Read across a category they show which arguments brands are actually making, and where several brands are making the same one.
Source — Structured retail product layer — claim extraction
02
MethodIngredient lists connect retail to foodservice
The same normalised ingredient vocabulary is applied to retail ingredient lists and to restaurant menus. That shared spine is what allows a single ingredient to be traced across both channels.
Source — Shared ingredient vocabulary across retail and menu layers
03
MethodThe SKU is the unit that matters
Brand-level analysis hides format, size and variant decisions. Structuring at SKU level keeps those decisions visible, because they are usually where margin and manufacturing complexity actually live.
Source — Structured retail product layer — SKU granularity
02Data snapshot
- MethodRetail structuring
- Retailer → brand → product → SKU
- MethodClaim layer
- Extracted per SKU
- MethodCross-channel link
- Shared ingredient vocabulary
03Methodology
Retail coverage is reported as structural capability rather than as a coverage count. Product, SKU, ingredient and claim layers are built from retail observation, and the ingredient layer uses the same controlled vocabulary as the menu layer so that the two channels remain comparable. Published coverage counts follow once a channel's extraction is complete and audited.
04Commercial implication
Reading claims and ingredient lists together shows which positions are already crowded, which are defensible, and which categories are arguing about the wrong attribute.