Restaurant Intelligence
Menus to Market Structure
How ingredients and categories appear across restaurant menus — and which patterns that structure makes visible.
Restaurant · September 2026
A menu is a document written in public. Read one at a time it describes a kitchen; read across a city it describes a supply chain. This report sets out how menu evidence is structured into ingredients, applications and restaurant segments, and what becomes legible once it is.
01Key findings
221
InferredVaried menu wording resolves into one shared vocabulary
93,479 menu items from 2,040 restaurant listings across 11 cities resolve into 221 ingredient and product concepts through a 222-rule extraction framework. However differently kitchens word their menus, the result is one shared vocabulary — which is what makes restaurants, segments and cities comparable on the same terms.
Source — Research snapshot, 10 Sep 2026 — 222-rule extraction framework
02
InferredSegment signatures are stable enough to plan against
Within a segment, a small set of ingredients recurs across most operators while a longer tail separates them. The recurring component set reveals common product patterns within the segment; the tail describes where positioning happens. Both are extracted from the same structured menu evidence.
Source — Structured menu extraction — normalised ingredient layer
03
InferredApplication matters more than presence
The same ingredient behaves as a different commercial opportunity depending on where it sits: a sauce base, a topping, a marinade, a side. Applications are captured separately from presence so a supplier can address the use, not just the word.
Source — Structured menu extraction — application layer
02Data snapshot
- InferredRestaurant listings
- 2,040
- ObservedMenu items analyzed
- 93,479
- InferredCities covered
- 11
- InferredIngredient & product concepts
- 221
Food Intelligence menu research corpus · 93,479 menu items across 11 cities · snapshot 10 Sep 2026
03Methodology
Menus are read from public sources and captured faithfully to the source text. A 222-rule extraction framework resolves each menu item's name and description into ingredient and product concepts. Presence is counted once per restaurant listing, not once per mention, so a term repeated across a menu does not inflate its weight. Figures on this page come from a dated research snapshot of the menu research corpus (10 Sep 2026); they describe that corpus, not the whole market.
04Commercial implication
Penetration tells you the size of the room. Application tells you which door to use. A supplier who knows both stops pitching a product and starts addressing a use that a specific segment already has.