Wine lists are operating documents: they show what a restaurant actually chooses to carry, how much breadth it gives guests, and which producers earn repeated placements across markets. This analysis covers 901 Michelin-starred restaurants with live wine-list coverage in AllWineLists. Of those, 897 have extractable wine rows suitable for list-level analysis.
The cohort contains 745 one-star, 132 two-star, and 24 three-star restaurants. Every producer named in the rankings below links to its canonical ELID producer page.
One-star restaurants supply most of the breadth
The 745 one-star restaurants make up 82.7% of the covered cohort. That scale matters: the higher tiers can show a different list profile, but their smaller samples should be read as directional rather than universal.Star level and list size are related—but the spread is wide
Across the 897 extractable lists, the median is 340 entries and the arithmetic mean is 525. The gap between the mean and median reflects a long tail of very large lists.| Tier | Covered | Extracted | Entries | Mean | Median | Middle 50% | Median matched producers |
|---|---|---|---|---|---|---|---|
| One star | 745 | 743 | 346,813 | 467 | 318 | 142–595 | 109 |
| Two stars | 132 | 130 | 101,172 | 778 | 508 | 173–1,015 | 176 |
| Three stars | 24 | 24 | 23,036 | 960 | 888 | 333–1,335 | 260 |
Distribution of list sizes
What “entry” means
An entry is one live row extracted from a restaurant wine list. It is not necessarily one unique cuvée: multiple vintages, formats, or service options can appear as separate rows. The comparison therefore describes operational list depth, not cellar inventory.
The producers with the widest restaurant distribution
Canonical name resolution links 345,332 entries (73.3%) to a producer ELID. At least one such match is present in 835 of the 897 extracted lists. Reach counts distinct restaurants, so a ten-wine placement and a one-wine placement each count once.| # | Producer | ELID | Restaurants | Reach | 1★ | 2★ | 3★ | Matched entries |
|---|---|---|---|---|---|---|---|---|
| 1 | Krug | FR-KRUG | 439 | 48.9% | 337 | 85 | 17 | 2,071 |
| 2 | Louis Roederer | FR-LORO | 353 | 39.4% | 277 | 60 | 16 | 1,478 |
| 3 | Bollinger | FR-BOLL | 342 | 38.1% | 266 | 60 | 16 | 1,513 |
| 4 | Dom Pérignon | FR-DOMP | 341 | 38.0% | 264 | 63 | 14 | 1,389 |
| 5 | Ruinart | FR-RUIN | 326 | 36.3% | 256 | 57 | 13 | 1,078 |
| 6 | Gaja | IT-GAJA | 320 | 35.7% | 248 | 58 | 14 | 2,221 |
| 7 | Chateau d'Yquem | FR-DYQU | 308 | 34.3% | 230 | 63 | 15 | 1,421 |
| 8 | Billecart-Salmon | FR-BISA | 306 | 34.1% | 240 | 54 | 12 | 1,321 |
| 9 | Tenuta San Guido | IT-TSGO | 297 | 33.1% | 237 | 50 | 10 | 1,421 |
| 10 | Laurent Perrier | FR-LAPE | 266 | 29.7% | 207 | 48 | 11 | 916 |
| 11 | Ornellaia | IT-ORNW | 265 | 29.5% | 202 | 51 | 12 | 1,464 |
| 12 | Palmer | FR-PALR | 262 | 29.2% | 192 | 55 | 15 | 778 |
| 13 | Vega Sicilia | ES-VESI | 255 | 28.4% | 198 | 46 | 11 | 1,226 |
| 14 | Chateau Latour | FR-LATR | 252 | 28.1% | 182 | 55 | 15 | 1,085 |
| 15 | Domaine Didier Dagueneau | FR-DIDA | 250 | 27.9% | 185 | 50 | 15 | 1,126 |
| 16 | Egly-Ouriet | FR-EGOU | 245 | 27.3% | 181 | 51 | 13 | 754 |
| 17 | Chateau Cos d'Estournel | FR-COSD | 244 | 27.2% | 186 | 45 | 13 | 763 |
| 18 | Château Mouton Rothschild | FR-MOUT | 243 | 27.1% | 180 | 50 | 13 | 899 |
| 19 | Domaine de la Romanée-Conti | FR-DRCO | 241 | 26.9% | 173 | 53 | 15 | 2,837 |
| 20 | Margaux | FR-MARX | 235 | 26.2% | 175 | 46 | 14 | 905 |
Producer references use canonical ELID producer identifiers. Resolution accepts unique normalized canonical names and repeated raw forms supported by at least two canonical matcher examples with at least 85% agreement. Unresolved raw strings are excluded.
A global restaurant sample, with a European wine center of gravity
Restaurant locations
Producer-origin placements
A producer-origin placement is one distinct restaurant × canonical producer pair. It measures distribution, not the number of bottles or list lines.
What the data supports—and what it does not
This dataset is strongest for distribution questions: how deep the live lists are, where the covered restaurants are, and how widely canonical producers travel across restaurant programs. It is not a market-size estimate. Online wine-list availability differs by country and restaurant, and a restaurant without a public list is outside the analytical sample.
Prices are deliberately excluded from cross-market comparisons here. The source rows span currencies, bottle sizes, by-the-glass service, pairings, and occasional nonstandard formats; comparing them without product- and format-level normalization would imply precision the source does not support.
A frozen cohort with reconcilable totals
The cohort was frozen on 2026-09-27 by reconciling the restaurant-searcher Michelin snapshot and Michelin ingress records to the paired AllWineLists establishment snapshot. Inclusion requires at least one live source-media record or live extracted wine row. The identity manifest contains 901 restaurants: 745 one-star, 132 two-star, and 24 three-star.
Coverage status: 894 restaurants have live source media; 897 have live extracted rows; 1,085 live media records and 471,021 live wine rows fall inside the cohort. Canonical full-wine matches at score ≥ 0.75 cover 19,574 entries (4.2%). Producer rankings use normalized canonical names plus evidence-backed raw-name aliases learned from accepted producer matches; aliases require at least two examples and 85% agreement.
The analysis is backed by michelin-starred-wine-lists.manifest.json (the 901 establishment identities, star tiers, match method, and source hashes) and build_michelin_wine_list_study.py (the upstream cohort and article generator). Shared-property Google Place IDs are explicitly excluded where the Michelin identity and wine-list establishment differ.