Category permission

For every brand and each of 45 subcategories, two models score three things from 0 to 100, using only what they already know. Web search is off. The models see the brand's name and domain, never the categories it declares.

• Today alignment — how strongly the subcategory belongs to the brand as it is today.
• Category permission — how credibly the brand could offer products or experiences there, whether or not it does today.
• Knowledge confidence — how well the model knows the brand.

Each model answers three times and keeps its median. OpenAI and Gemini are read separately. Where the page shows both together, it is an ensemble summary: their average, with the range between them kept visible. The conservative reading takes the lower permission and the higher alignment of the two.

The cutoffs are provisional screens, not business rules. An expansion candidate has permission of 70 or more and today alignment of 30 or less; the permission bar can be moved to 60 or 80 to test sensitivity. Today alignment of 70 or more means the brand is strongly present.

Market labels describe the compared set, not this brand. A subcategory is presently crowded when at least 15% of compared brands (and at least two) are strongly present today; potentially crowded when it is not presently crowded but as many could credibly enter (permission at the bar, alignment 30 or less); and whitespace when neither bar is met. Those three labels — whitespace, potentially crowded, presently crowded — are the only market statuses used on this page.

Distinctive permission is the brand's permission minus the average permission of the 10 brands whose alignment profile across all 45 subcategories is closest to its own.

The universe view plots every brand's established breadth against its permission breadth. Established is how many subcategories have today alignment at the chosen bar; permission is how many sit at that same bar, whether or not the brand is there today. Each count is scaled by the mean score of the subcategories that qualify, so eight at 90 sit further out than eight at 70. The line is an ordinary least-squares fit of permission breadth on established breadth. Brands above it have more permission than that fit predicts.